Basic

Photographic
Materials and
Processes

Basic Photographic Materials and Processes describes the Nanette L. Salvaggio is the Lecturer responsible for the
three crucial stages of creating the perfect photograph— Photographic Technologies class at the Rochester Institute
capture, processing and output—by providing a thorough of Technology. Her career in image quality, color science
technical investigation of modern, applied photographic and the photography industry includes work with Eastman
technologies. Kodak, Hughes Aircraft Corporation, TASC Inc. and Imagery
This new edition has been fully revised and updated Solutions, which she co-owned. She holds a Bachelor of
to explore digital image capture, processing and output. Science degree in Imaging and Photographic Technology
It covers a wide range of topics including: the scientific from the Rochester Institute of Technology.
principles of measuring and recording light, the inner work-
ings of digital cameras, image processing concepts, color Josh Shagam is a camera image quality engineer and
management and photographic output to screen and photographer working in Silicon Valley. He previously
print media. With these topics come in-depth discussions worked on image quality for aerial imaging systems at
of extending dynamic range, image histograms, camera Harris Corporation and before that taught photography at
characterization, display capabilities, printer and paper the Rochester Institute of Technology. He holds a Bachelor
technologies. It also includes applied exercises that provide of Science degree in Biomedical Photographic Communi-
the reader with a deeper understanding of the material cations and a Master of Professional Studies degree from
through hands-on experiments and demonstrations, con- the School of Visual Arts.
necting theoretical concepts to real-world use.
This comprehensive text provides photography stu-
dents, educators and working professionals with the
technical knowledge required to successfully create images
and manage digital photographic assets. It is an essential
resource for mastering the technical craft of photography.
Basic
Photographic
Materials and
Processes
Fourth Edition

Nanette L. Salvaggio
Josh Shagam
Fourth edition published 2020
by Routledge
52 Vanderbilt Avenue, New York, NY 10017

and by Routledge
2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

Routledge is an imprint of the Taylor & Francis Group, an informa business

© 2020 Taylor & Francis

The right of Nanette L. Salvaggio and Josh Shagam to be identified as

authors of this work has been asserted by them in accordance with
sections 77 and 78 of the Copyright, Designs and Patents Act 1988.

All rights reserved. No part of this book may be reprinted or reproduced

or utilised in any form or by any electronic, mechanical, or other
means, now known or hereafter invented, including photocopying and
recording, or in any information storage or retrieval system, without
permission in writing from the publishers.

Trademark notice: Product or corporate names may be trademarks

or registered trademarks, and are used only for identification and
explanation without intent to infringe.

First edition published by Focal Press 1990

Third edition published by Focal Press 2008

Library of Congress Cataloging-in-Publication Data

Names: Salvaggio, Nanette, author. | Shagam, Josh, author.
Title: Basic photographic materials and processes / Nanette L. Salvaggio
and Josh Shagam.
Description: Fourth edition. | London ; New York : Routledge, 2020. | Rev.
ed. of: Basic photographic materials and processes / Leslie Stroebel …
[et al.]. c2000.
Identifiers: LCCN 2019017064| ISBN 9781138744363 (hbk : alk. paper) |
ISBN 9781138744370 (pbk : alk. paper) | ISBN 9781315181097 (ebk :
alk. paper)
Subjects: LCSH: Photography.
Classification: LCC TR145 .B25 2020 | DDC 576.8/8--dc23
LC record available at https://lccn.loc.gov/2019017064

ISBN: 978-1-138-74436-3 (hbk)

ISBN: 978-1-138-74437-0 (pbk)
ISBN: 978-1-315-18109-7 (ebk)

Typeset in Myriad Pro by Alex Lazarou

 Josh dedicates
this book to his wife of 3 years, Taylor.

Nanette dedicates
this book to her husband of 33 years, Carl.

Thank you both for all of your patience and support

during the time it took to write this book.
Contents
Abbreviations and Acronyms xix
Introduction xxiii

Section 1: Capture

1 Light and Photometry 1

The Nature of Light 2
Measuring Light Waves 2
The Electromagnetic Spectrum 3
The Visible Spectrum 4
Spectral Power Distributions 6
Solid-Object Sources and the Heat-Light Relationship 7
Vapor Sources and the Effect of Electrical Discharge 7
Luminescence, Fluorescence and Phosphorescence 9
Methods of Measuring Light 10
Intensity 10
Flux 11
Illuminance 12
Luminance 14
Color Temperature 16
The Mired Scale 18
Color Rendering Index 20
The Color Quality Scale 21
Describing the Appearance of Light 22
The Polarization of Light 23
Commonly Encountered Light Sources 24
Sunlight 24
Tungsten-Filament Lamps 26
Tungsten-Halogen Lamps 27
Fluorescent Lamps 27
Compact Fluorescent Lamps 28
High-Intensity Discharge Lamps 28
Electronic Flash 29
Light-Emitting Diodes 30
Proof of Concept: Visualizing the Spectral Output of Light Sources 31

vii
 c o n t e n ts

2 Photographic Optics 33
Image Formation with a Pinhole 34
Image Formation with a Lens 36
Lens Focal Length 40
Normal Focal Length 42
Telephoto Focal Lengths 44
Wide-Angle Focal Lengths 44
Variable Focal Length 45
The Relationship Between Focal Length and Perspective 46
Focal Length and Image Size 46
Object Distance and Image Size 46
Focal Length and Object Distance 48
Drawing Lens Diagrams 49
Depth of Field 53
Depth of Field Controls 56
Depth of Field and F-Number 56
Depth of Field and Object Distance 56
Depth of Field and Focal Length 57
Depth of Focus 57
Specialty Lenses 58
Macro Lenses 58
Perspective Control Lenses 59
Catadioptric Lenses 60
Reversed Telephoto Wide-Angle Lenses 60
Supplemental Lenses 61
Extension Tubes and Teleconverters 61
Lens Shortcomings and Aberrations 62
Chromatic Aberration 62
Spherical Aberration 63
Diffraction 63
Vignetting 64
Bokeh 65
Barrel and Pincushion Distortion 66
Flare 66
Proof of Concept: Focal Length, Subject Distance and Portraits 68

3 Digital Camera Technology 71

The Single Lens Reflex Camera Design 72
Recording Light with Digital Image Sensors 73
Bayer Pattern Color Filter Array 74
X-Trans Color Filter Array 74
Camera Sensor Technologies 75

viii
 c o n t e n ts

Charge-Coupled Device Sensor 75

Sparse Color Filter Array Sensor 76
Complementary Metal Oxide Semiconductor Sensor 76
Backside Illumination Detectors 78
Foveon Sensor 78
Image Sensor Sizes 78
Camera Sensor Resolution 81
Pixel Pitch 81
Image Bit Depth 81
Aliasing and Moiré Patterns 83
Apertures 84
Shutters 85
Leaf Shutters 85
Focal Plane Shutters 86
Electronic Shutters 87
Automatic Exposure and Metering 88
Automatic Focus 88
Phase Detection Autofocus 88
Contrast Detection Autofocus 89
Active Autofocus 89
Autofocus Points 89
Image Stabilization 90
Lens-Based Optical Image Stabilization 90
Camera-Based Image Stabilization 91
Electronic Image Stabilization 91
Camera Viewfinders 91
Optical Viewfinders 91
Electronic Viewfinders 92
Electronic Flash 93
Camera Form Factors and Characteristics 94
Single Lens Reflex Cameras 94
Medium Format Cameras 94
Rangefinders 95
View Cameras 95
Mirrorless Cameras 96
Compact Cameras 96
Mobile Phone Cameras 96
Action Cameras 97
Camera Storage Media 98

4 Photographic Exposure 101

Exposure Determination 102

ix
 c o n t e n ts

Dictating Exposure Time with Shutter Speed 102

Dictating the Amount of Light with F-Number 104
Lens Speed 105
Dictating Effective Sensitivity with Gain and ISO 105
Exposure and Dynamic Range 107
Blooming 108
The Utility of Image Histograms 109
Histogram Interpretation 110
In-Camera versus Desktop Software Histograms 113
Exposure Value 113
Equivalent Exposure 113
The Image Capture/Exposure Sequence 114
Using a Neutral Reference Target 115
In-Camera Exposure Metering 115
Matrix or Evaluative Metering 116
Center-Weighted Metering 116
Spot Metering 116
Exposure Compensation 116
Automatic Exposure Behaviors 116
Handheld Light Metering 117
Alternative Metering Methods 117
Synchronizing Flash and Camera Exposure 118
Exposing to the Right 118
Bracketing Exposure 119
Setting White Balance 119
Proof of Concept: Optimizing Photographic Exposure with Low Light Shooting 122

5 Lens Filters 125

Transmittance, Opacity and Density 126
How Light is Filtered 127
The Filter Factor 128
Color Filters 129
Neutral Density Filters 130
Polarizing Filters 131
Ultraviolet and Infrared Filters 132
Spectrophotometric Absorption Curves 133
Proof of Concept: Constructing Color Images Using Color Separations 134

6 Motion Video Fundamentals 137

Motion Perception and Memory 138
Persistent Images 138

x
 c o n t e n ts

Afterimages 139
Visual Memory 139
Defining Motion Video and its Parameters 140
Frame Rate 140
Shutter Angle 142
Resolution 143
Aspect Ratio 144
Video Formats and Encoding Strategies 145
Intraframe and Interframe Encoding 146
Chroma Subsampling 148
Video Codecs 148
Motion JPEG 149
MPEG-4 Advanced Video Coding (H.264) 149
High Efficiency Video Coding (H.265) 149
VP9 149
AOMedia Video 1 (AV1) 150
Audio in Video 150
Bitrate and File Size 150
Video Quality Artifacts 151
Blocking 151
Packet Loss 152
Ringing 152
Mosquito Noise 152
Color Bleeding 152
Jerkiness 153
Emerging Trends in Motion Video 153
Wide Color Gamut and High Dynamic Range 153
Short Form or Micro Video 153
Immersive Video 153

Section 2: Process

7 Demosaicing and Interpolation 157

Contents of a Raw File 158
Image Interpolation 158
Demosaicing Image Data 160
Processing Steps Following Demosaicing 161
Exposure Adjustment 161
Color Space Mapping 161
Gamma Correction 162
Noise Reduction 162

xi
 c o n t e n ts

Anti-Aliasing Filter 163

Sharpening 163
Reinterpreting Raw with New Algorithms 163
Resampling an Image 163
Nearest Neighbor Interpolation 164
Bilinear Interpolation 165
Bicubic Interpolation 165
The Life Cycle of Image Pixel Data 166

8 Digital Asset Management 167

Defining Image Data 168
Image Data and Quality 168
Image File Formats 169
Raw 169
DNG 170
JPEG 170
TIFF 171
BMP 172
GIF 172
PNG 172
HEIF 172
Proprietary Editing Software Formats 173
Why Raw Works for Digital Asset Management 173
The Photographic Workflow 175
Image Ingestion 175
Naming Files 176
Cataloging Software 177
Image Editing 178
Metadata and Keywords 179
Image Data and File Sizes 180
File Storage Options 181
Optical Disc Media 182
Hard Disk Drives 182
Solid State Drives 183
Disk Configurations 183
Backup Solutions 184
Remote Server File Hosting and Storage Services 185
Data Recovery 185
Avenues for Data Transfer 185
Data Transfer Connection Standards 186
Computer Workstation Hardware 188
Central Processing Unit 188

xii
 c o n t e n ts

Graphics Processing Unit 188

Random Access Memory 188
Storage Drives 189
Proof of Concept: Considering File Types and Image Content 189

9 Digital Tone Reproduction 193

Objective and Subjective Tone Reproduction 194
The Evolution of Tone Reproduction 195
Revisiting the Image Histogram 196
Interpreting the Tone Curve 199
Using the Curves Tool to Modify Tones 199
Inverting Tones for Negative Film Scans 201
The Zero-Sum Game of Tone Reproduction 202
Using the Levels Tool to Modify Tones 203
Common Tone Reproduction Software Controls 204
Color Correction and the ColorChecker 205
Color Balancing Using Color Channel Curves 206
Split Toning 208
Proof of Concept: Color Correcting Film Scans using Curves 208

10 Software Filters 213

Sharpening an Image 214
Blurring an Image 218
Blending Images 219
Black and White Conversion 221
Digital Filtering for Special Effect 222
Filter Masking 223
Proof of Concept: Reproducing an Aesthetic Filter Process 225

11 Camera Characterization and Image Quality 229

The Imaging Pipeline 230
The Many Tasks of the Image Signal Processor 230
ISP Task: Auto White Balance 231
ISP Task: Black Level Subtraction 232
ISP Task: Stuck Pixels 232
ISP Task: Demosaicing 232
ISP Task: Compressing and Saving 232
ISP Task: Video Recording 232
Characterizing a Camera 233
Testing Sensor Linearity 233

xiii
 c o n t e n ts

Evaluating Bright and Low Light Performance 234

Categorizing Image Noise 234
Fixed Pattern Noise 235
Dark Noise 236
Read Noise 236
Photon Noise 236
Signal to Noise Ratio 236
Delineating Noise from Graininess 237
Considering Sensor Dynamic Range 238
Identifying Non-Uniformities 239
Evaluating Lens Quality: Sharpness and Detail 241
Modulation Transfer Function 241
Interpreting an MTF Plot 241
Shooting to Evaluate Image Definition 242
Assessing Color Accuracy 243
Calculating Color Difference 243
White Balance Accuracy 244
Calibrating Autofocus 245
Considering Your Imaging Tools 246
Proof of Concept: Comparing Noise Characteristics of Two Cameras 246

12 Image Compression 249

Image Compression and Information Theory 250
The Basic Stages of Image Compression 251
Finding Redundancies in Image Data 251
Spatial or Interpixel Redundancy 251
Coding Redundancy 252
Psychovisual Redundancy 252
Temporal Redundancy 253
Compression and Scene Content 253
Lossless Compression Approaches for Removing Spatial Redundancy 253
Run-Length Encoding 253
Bit-Plane Encoding 254
Differential Encoding 255
Lossless Compression Approaches for Removing Coding Redundancy 255
Huffman Coding 255
Lempel–Ziv–Welch Compression 256
The YCbCr Color Space 257
Compressing Color 258
Using Discrete Cosine Transforms to Compress 259
JPEG Compression 260
Common Image Compression Artifacts 261

xiv
 c o n t e n ts

Blocking Artifacts 262

Ringing Artifacts 262
Posterization 262
Color Bleeding 263
Generation Loss 263

Section 3: Output

13 Color Vision and Systems 265

Sensing Light with the Human Eye 266
Color Vision Theories 268
Testing Color Vision 269
Color Vision Deficiencies 270
Methods of Creating Color 271
Additive Mixing 271
Subtractive Mixing 272
Metamerism’s Role in Color Reproduction 274
Partitive Mixing 275
Defining Color with Color Specification Systems 276
The Pantone System of Color Specification 277
The Munsell System of Color Specification 277
The CIE System of Color Specification 279
The CIE Chromaticity Diagram 280
CIELAB Space 281
The Roles of Light and Color on Perception and Perspective 282
Light Adaptation 282
Color Contrast 283
Linear Perspective 284
Color and Perceived Distance 285
Shape and Dimension 285
Size and Relative Size 287
Depth of Field 288
Lighting 288
Overlap 288
Aerial Haze 288
Texture 288
Proof of Concept: Looking Closely at Additive Color Mixing on Displays 288

14 Color Management 291

Why Color Management Matters 292

xv
 c o n t e n ts

Who Governs Color Management? 292

Reviewing Color Data Concepts 292
Color Space Types 293
sRGB, Adobe RGB and ProPhoto RGB 294
Pointer’s Gamut 294
Revisiting the ColorChecker 294
Defining Characterization and Calibration 295
Color Management Systems 295
The ICC Profile 296
Canned Profiles 297
The iccMAX Profile 297
CMYK Profiles 297
Converting or Assigning Profiles 298
Creating Profiles with Color Measurement Tools 298
Input Profiles for Cameras or Scanners 298
Display Device Profiles 299
Printer Profiles 300
Gamut Handling and the Profile Connection Space 300
Rendering Intents 301
Evaluating Color Reproduction 302
Measuring and Comparing Color 302
Color Correction 303
Establishing an Appropriate Workspace Setup 304
Soft Proofing with Profiles 305
A Soft Proofing Workflow 305

15 Display Technology 307

Relevant Display Technologies for Photographers 308
Cathode Ray Tube 308
Liquid Crystal Display 308
Organic Light-Emitting Diode 311
Quantum Dot 311
Display Connection Standards 312
Display Resolution 312
Aspect Ratio and Screen Size 312
Refresh Rates and Response Time 313
Display Dynamic Range, Brightness and Bit Depth 314
Display Color Gamut 315
Defining the Display Viewing Conditions 316
Viewing Angle 316
Viewing Distance 317
Calibration and Configuration for Color Management 318

xvi
 c o n t e n ts

Recommended Settings and Conditions 318

Display Profile Characteristics 318
Emerging Display Trends and Applications 319
Night Modes and Reduced Blue Light Exposure 319
Electronic Viewfinders 320
Head Mounted Displays 320
Proof of Concept: Documenting the Growth in Mobile Device Display Capabilities 320

16 Printing and Physical Media 323

The Fundamentals of Printing 324
Pixels Per Inch and Dots Per Inch 324
Continuous Tone and Halftone Printing 325
Printer Technologies 326
Inkjet Printing 326
Photosensitive Light Writing Printing 328
Electrophotographic Printing with Dry Toners 329
Electrophotographic Printing with Liquid Toners 329
Thermal Dye-Diffusion Printing 330
Dye Sublimation Printing 331
Ink Technology 331
Aqueous Inkjet Inks 331
Solvent Inks 332
Ultraviolet Curable Inks 332
Latex Inks 333
Dye Sublimation Inks 333
Printing Substrates 333
Aqueous Inkjet Substrates 334
Chromogenic Substrates 335
Liquid Toner Substrates 336
Dry Toner Substrates 336
Dye Sublimation Substrates 337
Photo Paper Characteristics 337
Paper Weight and Thickness 337
Surface Sheen 338
Surface Texture 338
Brightness and Optical Brightening Agents 338
Mineral Whiteners 339
Paper White Point 339
Understanding Printer Settings 339
Photo Black and Matte Black Inks 339
Printhead Height 340
Printing Bit Depth 340

xvii
 c o n t e n ts

Super MicroWeave and Finest Detail 340

High-Speed Printing 340
Output Resolution 341
Printer Drivers 342
Print Size and Viewing Distance 342
Raster Image Processing 343
Hard Proofing 343
Printing Issues 344
Evaluating Paper and Print Quality 345
Testing for OBAs 346
Measuring Color Gamut and White Point 346
Measuring Black Point Density and Gloss 347
Critically Evaluating Color and Contrast 348

Index 351

xviii
Abbreviations
and Acronyms
AAC Advanced Audio Coding CMOS Complementary Metal Oxide Semiconductor
ACE Adobe Color Engine CMS Color Management System
A/D Analog to Digital CMYK Cyan, Magenta, Yellow, Key (Black)
AE Automatic Exposure CPU Central Processing Unit
AEB Automatic Exposure Bracketing CQS Color Quality Scale
AF Automatic Focus CRI Color Rendering Index
AM Amplitude Modulation CRT Cathode Ray Tube
ANSI American National Standards Institute CT Color Temperature
APS Active Pixel Sensor
AR Augmented Reality D50 CIE Standard Illuminant Daylight 5500K
ARTS Absorbed, Reflected, Transmitted, Scattered D65 CIE Standard Illuminant Daylight 6500K
AS-IPS Advanced Super In-Plane Switching DAM Digital Asset Management
AV1 AOMedia Video 1 DCT Discrete Cosine Transform
AVC Advanced Video Coding DIB Device Independent Bitmap
AWB Automatic White Balance DICOM Digital Imaging and Communications in
Medicine
B Byte DLP Digital Light Processing
BD Blu-ray Disc DNG Digital Negative
BID Binary Ink Developer DOD Drop-on-Demand
BMP Bitmap Image File DOF Depth of Field
BRDF Bidirectional Reflectance Distribution Functions DPI Dots Per Inch
BSI Backside Illuminated DSLR Digital Single Lens Reflex
DSNU Dark Signal Non-Uniformity
CCD Charge-Coupled Device DVD Digital Versatile Disc
CCFL Cold Cathode Fluorescent Lamp DVI Digital Visual Interface
CCM Color Correction Matrix DWT Discrete Wavelet Transform
CCT Correlated Color Temperature
CD Compact Disc EIS Electronic Image Stabilization
CF Compact Flash eSATA External Serial Advanced Technology
CFA Color Filter Array Attachment
CFL Compact Fluorescent Lamp ETTR Expose To The Right
CIE Commission Internationale de l'Éclairage EV Exposure Value
(International Commission on Illumination) EVF Electronic Viewfinder
CMM Color Management Module or Color Matching EXIF Exchangeable Image File Format
Module

xix
 A b b r e v i at i o n s a n d A c r o n y m s

F2 CIE Standard Illuminant Fluorescent JPEG Joint Photographic Experts Group

FAT File Allocation Table JPEG-LS Joint Photographic Experts Group - Lossless
FLAC Free Lossless Audio Codec Compression
FM Frequency Modulation
FPS Frames Per Second K Kelvin
KB Kilobyte
GB Gigabyte
GeoTIFF Geostationary Earth Orbit Tagged Image File LAN Local Area Network
Format LCD Liquid Crystal Display
GHz Gigahertz LED Light-Emitting Diode
GIF Graphics Interchange Format LPI Lines Per Inch
GIS Geographic Information Systems LUT Lookup Table
GOP Group of Pictures LZW Lempel–Ziv–Welch
GPS Global Positioning System
GPU Graphics Processing Unit MB Megabyte
GRACoL General Requirements for Applications in MCS Material Connection Space
Commercial Offset Lithography MJPEG Motion Joint Photographic Experts Group
GU Gloss Unit MK Matte Black
MPEG Moving Picture Experts Group
HD High Definition MTF Modulation Transfer Function
HDD Hard Disk Drive mSATA Mini-Serial Advanced Technology Attachment
HDMI High-Definition Multimedia Interface
HDR High Dynamic Range NAS Network-Attached Storage
HEIF High Efficiency Image File Format ND Neutral Density
HEVC High Efficiency Video Coding NIST National Institute of Standards and Technology
HFR High Frame Rate NTSC National Television Standards Committee
HMD Head Mounted Display
HTML Hypertext Markup Language OBA Optical Brightener Agent
Hz Hertz OIS Optical Image Stabilization
OLED Organic Light-Emitting Diode
ICC International Color Consortium OLPF Optical Low-Pass Filter
IPS In-Plane Switching OS Operating System
IPTC International Press Telecommunications Council
IR Infrared PB Petabyte
IS Image Stabilization PCIe Peripheral Component Interconnect Express
ISO International Standards Organization PCS Profile Connection Space
ISOBMFF ISO Base Media File Format PIE Parametric Image Editing
ISP Image Signal Processor or Internet Service PIP Photo Imaging Plate
Provider PK Photo Black
ITU International Telecommunication Union PRNU Pixel Response Non-Uniformity
PNG Portable Network Graphics
JBOD Just a Bunch of Disks POD Print On Demand
JND Just Noticeable Difference PPI Pixels Per Inch

xx
 A b b r e v i at i o n s a n d A c r o n y m s

PSB Photoshop Big USB Universal Serial Bus

PSD Photoshop Document UV Ultraviolet

QD Quantum Dot VA Vertical Alignment

QE Quantum Efficiency VM(λ) Photopic luminosity function of the human eye
VR Virtual Reality
RAID Redundant Array of Independent Disks VVC Versatile Video Coding
RAM Random Access Memory
RC Resin-Coated WCAG Web Content Accessibility Guidelines
RGB Red, Green, Blue WCG Wide Color Gamut
RGBAW Red, Green, Blue, Amber, White WORM Write-Once-Read-Many
RIP Raster Image Processor
RLE Run-Length Encoding XMP Extensible Metadata Platform
RPM Revolutions Per Minute
RW Rewritable YCbCr Luminance, Chroma Blue, Chroma Red

SaaS Software as a Service

SATA Serial Advanced Technology Attachment
SD Standard Definition or Secure Digital
SDHC Secure Digital High Capacity
SDUC Secure Digital Ultra Capacity
SDXC Secure Digital Extended Capacity
SDR Standard Dynamic Range
S-IPS Super In-Plane Switching
SLR Single Lens Reflex
SNR Signal to Noise Ratio
SoC System on a Chip
SPD Spectral Power Distribution or Silicon
Photodiode
SSD Solid State Drive
SWOP Specifications for Web Offset Publications

TB Terabyte
TIFF Tagged Image File Format
TN Twisted Nematic
TRC Tone Response Curve
TTC Tone Transfer Compensation
TTL Through the Lens

UDMA Ultra Direct Memory Access

UHD Ultra High Definition
UHS Ultra High Speed
UI User Interface

xxi
Introduction

It’s imperative that the modern image maker understands workings of digital cameras, image processing concepts,
the workings of photography’s latest tools and processes: color management and photographic output to both
this was our mindset when writing this textbook. This book screen and print media. With these topics come in-depth
is not a collection of step-by-step guides on photographic discussions of extending dynamic range, image histo-
composition, studio lighting or editing in Adobe Photo- grams, camera characterization, display capabilities, printer
shop. Instead, it’s a deep dive into the medium’s complex and paper technologies and more. We’ve also created sec-
and interrelated workings. We’ve organized the information tions at the end of select chapters called Proof of Concept
according to the basic pillars of a photographic workflow that provide the reader with a deeper understanding of
(capture, processing and output) but recognize that its the material through hands-on experiments, exercises and
content may not be read in a linear or single-sitting fashion demonstrations. These illustrative sections work to connect
(we’d be totally impressed if you did, though). theoretical concepts to real-world use-cases.
Furthermore, this book does not attempt to be a It’s our belief that a comprehensive book is needed to
comprehensive or historical guide to all of photographic teach the science and technology that makes modern
technology. It’s instead a technical investigation of modern, digital photography possible. This is not knowledge for its
applied photographic technology. We made the choice own sake—everything included on the following pages is
to write a forward-thinking resource not because we connected to the applied practice of image-making. All of
think “film is dead” (it’s not!) but because many, many it is in the service of empowering your photographic vision
resources exist concerning analog processes and tech- and technical craft. Photography is a medium in constant
nologies, including previous editions of this book. While flux with changes in aesthetics, technology and applica-
photographers of all types continue to explore film-based tions coming so frequently and fluidly that it is hard to pin
image-making, the reality is that wet process photography down its epochs. In overhauling our first-year undergrad-
is more of a niche than a norm. Modern photographic uate course at Rochester Institute of Technology’s School
technology inherited many core principles, terms and the of Photographic Arts and Sciences, we learned a great deal
insights made from hundreds of years of optical and mate- about what newcomers must know to be proficient with
rial science. We are no longer in a transition from analog the latest hardware and software tools. We hope that the
to digital and young photographers are exposed to film content explored on the following pages proves insightful
cameras and darkrooms only if they seek them out, not as and empowering for your creative practice.
ubiquitous cultural touchstones.
This new edition provides students studying pho-
tography for the first time or professionals working in this
paradigm with the technical background required to create
images and successfully manage digital photographic
assets. This book covers a wide range of topics: the scien-
tific principles of measuring and recording light, the inner

xxiii
Section 1
Capture
We begin our exploration of photographic processes and technologies that detect, record and digitize light including
technologies with image capture. An image may start as camera apertures, focusing systems and sensors. A great
a concept, a plan or idea; a photograph is truly born into camera can take a terrible photograph, so it’s important
existence at its moment of capture. There is no photo- to break down the mechanics of proper photographic
graphic capture without light, so we start this textbook exposure once we’re familiar with the camera itself. The
section with photometry: the science of measuring light. behaviors and interrelated effects of shutter speed, aper-
Photometry establishes our definitions of visible light, light ture and ISO are demystified in their own chapter.
source technologies and the characteristics of light such as Finally, we’d be remiss if we ignored motion video as
appearance and behavior. a capture method. While not photography in the strictest
Understanding light helps us to control and cap- sense, video comes with the territory due to the overlap of
ture it. The chapters following photometry start with image-making and light-recording equipment.
photographic optics and follow the path of light as it is There’s a lot of ground to cover in this section as we take
transformed from photons to bits. We look at the designs a comprehensive look at everything that goes into record-
and functions of lenses, perspective, distortions and ing light to make photographs. Forge ahead to expand
approaches for evaluating image quality. your knowledge of photographic technology and don’t
Forming an image on a plane is made permanent forget to take photographs as you go—test the concepts
by recording it. We lay out the relevant digital camera by experimenting and exploring with your eyes and your
camera.
 1 Light and
Photometry

A smartphone camera LED flash viewed through a diffraction grating, an optical material that acts like a prism. The multicolored streaks
radiating from the light source reveal its spectral energy components. Photograph by Rochester Institute of Technology photography student
Hank Tilson

The potential for a photograph can begin as an idea in one’s imagination or a confluence
of circumstances and awareness, yet it is made real only when the materials and processes
of photography are brought together to capture it. Light is the material that has defined
photography since its invention; it is the basic ingredient for making photographs. For this
reason, we start with photometry: the science of measuring and describing light. There’s a
lot of ground to cover: first we define light and its properties, then the many ways in which
light is produced, described and measured. Finally, we highlight the light sources typically
encountered by photographers. By chapter’s end, you will know and understand light to
confidently wield it for photographic ends.

1
 s ecti o n 1: Captu r e

The Nature of Light In 1905, Albert Einstein proposed a return to the corpus-
cular theory of light where light consists of photons, each
Light is defined as the radiant energy that our visual system photon containing a quantum of energy. These sugges-
is sensitive to and depends upon for the sensation of tions, along with others, developed into what is known
vision. Light is fundamental to photography. The function today as quantum theory or quantum electrodynamics.
of digital sensors and light-sensitive photographic mate- Quantum theory combines aspects of the corpuscular and
rials is to capture light and record the patterns it creates. wave theories and satisfactorily explains all of the known
We use sunlight, lamps or electronic flash to produce light; behaviors of light. It states that light acts like both a wave
exposure and color temperature meters to measure and and a particle: light exhibits a wave-particle duality. Unfor-
characterize the light; and lenses, shutters, apertures and tunately, this theory is difficult to conceptualize and can be
filters to control light falling on our light-sensitive recording rigorously explained only through sophisticated mathe-
medium. Our study of photography must start with under- matics. As a result, the corpuscular and wave theories are
standing light. still used where simple explanations of the behavior of light
Light is the subject of an enormous amount of experi- are required. Sometimes our observations of light behavior
mentation and research spanning many centuries through are best explained by its wave-like character. Other times,
to today. Isaac Newton was one of the first to make signif- light behavior is better characterized by considering it as
icant steps in understanding the nature of light. Newton packets of energy.
performed a series of experiments in the seventeenth
century and proposed that light is emitted from a source in
straight lines as a stream of particles; this became known as Measuring Light Waves
the corpuscular theory.
Photographers know that light bends when it passes Accepting that light moves as a wave function, the next
from one medium to another and that light passing step is determining the nature of the waves and the
through a very small aperture spreads out. These behav- relationship of light to other forms of radiation. The light
iors are not easily explained by the corpuscular theory. As experienced with our visual system is a fractional part of
a result, Christiaan Huygens proposed the wave theory, a wide spectrum of radiant energy that exists in the uni-
also in the seventeenth century, stating that light and verse, all of which can be thought of as traveling in waves.
similar forms of electromagnetic radiation are transmitted These forms of energy travel at the tremendous speed of
as a waveform in a medium. Thomas Young performed approximately 186,000 miles (3x108 meters) per second.
a number of experiments in the nineteenth century that Radiant energy waves vibrate at right angles relative to
supported Huygens’ characterization. The wave theory their path of travel. The distance from the crest (the peak
satisfactorily explained many of the phenomena associated height) of one wave to the crest of the next is termed
with light that the corpuscular theory did not, though it still the wavelength represented by the Greek letter lambda
did not explain all of them. (λ). Figure 1-1 illustrates this concept with a longitudinal
One of the notably unexplained effects is the behavior cross section of a light wave; in reality, the wave vibrates
of blackbody radiation. Blackbody radiation is radiation in all possible right angles to the direction of travel. The
produced by a body that absorbs all of the radiation that number of waves passing a given point in 1 second is the
strikes it and emits radiation by incandescence (heat frequency of vibration; the symbol f specifies this wave
energy) depending on its temperature. In 1900, Max Planck characteristic. The wavelength multiplied by the vibration
hypothesized the quantization of energy to explain the frequency results in the speed or velocity (v) of the radia-
behavior of blackbody radiation. This theory states that the tion. Thus, v = λ × f .
only possible energies that can be possessed by a ray of Since the wavelength of radiant energy can be meas-
light are integral multiples of a quantum of energy. ured with far greater accuracy than the frequency, it is

2
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Table 1-1 Units of length

Unit Symbol Length

Meter m 1 m (38.6 in.)

Centimeter cm 0.01 m (10–2 m)

Millimeter mm 0.001 m (10–3 m)

Micrometer/Micron µ (mu) 0.000001 m (10–6 m)

Nanometer nm 0.000000001 m (10–9 m)

Angstrom Å 0.0000000001 m (10-10 m)

Figure 1-1 A simple form of a light wave illustrated in a longitudinal

cross section. The second wave has a wavelength one-half that of the
first and, therefore, a frequency twice as great.

common practice to specify a type of radiation by its wave-

length. The unit of measure used for wavelength is the
nanometer (nm) due to the extremely short wavelengths
emitted by electromagnetic energy sources. A nanometer
is equal to one-billionth of a meter. Table 1-1 summarizes
units of length and their symbols to offer context for the
scale of radiant energy wavelengths.

The Electromagnetic Spectrum

When the various forms of radiant energy are placed along

a scale of wavelengths, the resulting continuum is called
the electromagnetic spectrum. Although each form of
radiant energy differs from its neighbors by an extremely
small amount, it is useful to divide this spectrum into the
generalized categories shown in Figure 1-2. All radiation is
believed to be the result of electromagnetic oscillations.
In the case of radio waves, the wavelengths are extremely
long (on the order of 109 nm) and are the result of long
electrical oscillations. Proof that such energy permeates
our environment is easily demonstrated by turning on a
radio. Radio waves are customarily characterized by their Figure 1-2 The electromagnetic spectrum.
frequency, expressed in cycles per second or Hertz (Hz).

3
 s ecti o n 1: Captu r e

The infrared region is the portion of the electromagnetic The Visible Spectrum
spectrum that we experience as heat. The origin of this
type of radiant energy, which is shorter in wavelength than Light energy is located near the middle of the electromag-
radio waves, is the excitation of electrons by thermal dis- netic spectrum called the visible spectrum. The boundaries
turbance. When these electrons absorb energy from their of this region are solely dictated by the response charac-
surroundings, they are placed in an elevated state of activ- teristics of the human eye. In fact, the standard definition
ity. When they suddenly return to their normal state, they of the visible spectrum specifies that light is radiant energy
emit electromagnetic radiation. Experiments show that experienced by a human observer as visual sensations via
any object at a temperature greater than –459°F (–273°C) the retina. Simply put, light is the energy that permits us
gives off infrared radiation. Thus, all the objects we come to see. By definition, all light is visible and for this reason
into contact with emit some infrared energy. In general, the the word visible is an unnecessary (and perhaps confusing)
hotter an object, the more total energy it produces and the adjective in the common expression visible light. This defi-
shorter the peak wavelength. nition also may be interpreted to mean that energy that is
If an object is heated to a high enough temperature, the not visible cannot and should not be called light. It’s also
wavelength of the energy emitted becomes short enough more accurate to talk about ultraviolet radiation (UV) and
to stimulate the retina of the human eye and cause the infrared radiation (IR), rather than ultraviolet light or infrared
sensation of vision. This region of the electromagnetic light. The popular use of such phrases as “blacklight” and
spectrum is labelled as light or visible light. Notice in Figure “invisible light” to describe such radiation makes it impos-
1-2 that the visible region occupies a relatively narrow sible to determine what type of energy is emitted and
section of the spectrum between approximately 380 and should be avoided.
720 nm. Because the sensitivity of the human visual system It helps to be familiar with how the human eye
is low at both extreme ends of this range, 400 and 700 nm responds to light to better understand it. Figure 1-3
are generally considered to be more realistic values. represents the photopic luminosity function of the human
Producing radiant energy shorter than 10 nm in wave- eye, VM(λ), as defined by the International Commission on
length requires that fast-moving electrons bombard an Illumination (abbreviated as CIE from its French name,
object. The sudden collision of these rapidly moving Commission Internationale de l’Éclairage). The plot
electrons striking the object produces extremely shortwave illustrates the eye’s sensitivity to different wavelengths
energy called X-radiation, or more commonly, x-rays. Still of light. The data indicates that the sensitivity of the eye
shorter wavelengths are produced if the electron bom- drops to near zero at 380 nm and at 720 nm. It also shows
bardment intensifies, as occurs in a cyclotron (a type of that the response of the eye is not uniform throughout
particle accelerator used in nuclear medicine). In addition, the visible spectrum. Human vision is most sensitive to
when radioactive material decomposes, it emits energy green light. If equal physical amounts of different colors
shorter in wavelength than x-rays. In both cases, the energy of light are presented to an observer in turn, the green
is called gamma rays, which are 0.000001 nm (10-6 nm) in portion of the spectrum appears the brightest while the
wavelength and shorter. Gamma rays are the most ener- blue and red parts appear dim. This uneven response
getic, penetrating radiation known. Our Earth’s atmosphere across the spectrum is key to working with light, color and
blocks a great deal of the gamma ray and x-ray radiation digital image information as a whole; storage, transmis-
from space. sion, compression and output of photographs are built
We begin to appreciate the usefulness of classifying on this foundational knowledge of our vision as we see in
the known forms of radiation using wave theory and the later chapters of this text.
portion of the spectrum that we categorize as light in the Figure 1-3 plots the international standard response
context of human vision and photography. function for the measurement of light. This means that
any light measurement device must possess a sensitivity

4
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

100

80
Relative Sensitivity

60

40

20

0
300 400 500 600 700 800
Wavelength (nm)

Figure 1-3 The photopic luminosity function or, more simply, the Figure 1-4 The spectral power distributions of a tungsten light source
sensitivity curve of the human eye. The plot indicates the relative (CIE Illuminant A) and daylight at 6500K (CIE Illuminant D65).
brightness of the energy at each wavelength.

function identical to it. Most photoelectric meters used in judging the color quality of the prevailing illumination for
photography have response functions significantly dif- the purposes of color photography. To prove it to yourself,
ferent from the standard and are not properly called light walk around with white notebook paper, moving from
meters, although the international standard curve can be indoors to outdoors or between rooms with a variety of
approximated by using filters with some meters. Note that light sources. Your eyes and brain communicate that the
determining the proper f-number and shutter speed for a paper is white in spite of the changes to the environment’s
given photographic situation does not require a meter with dominant light source and the composition of the light
this response function. It is more important for the meter to energy reflected by it.
match the sensitivity of the image sensor or film emulsion When a beam of white light passes through a glass
used than that of the human eye. prism as illustrated in Figure 1-5, the light is dispersed into
When all of the wavelengths between approximately its component colors (plus some energy we can’t see on
400 and 700 nm are presented to the eye in nearly equal either end, ultraviolet and infrared). This separation of the
amounts, the light is perceived as white. There is no colors occurs because wavelengths of light are bent by
absolute standard for white light because the human different amounts and is termed dispersion. Blue light, the
visual system easily adapts to changing conditions in shortest wavelength range in the visible spectrum, bends
order to obtain the perception of a neutral “whiteness.” to a greater extent than the longer wavelength green and
For example, Figure 1-4 shows that the distribution of red light. The result is the rainbow of colors that spans
wavelengths of energy present in daylight are significantly deep violet to a deep red. Experiments indicate that human
different from those of tungsten light; however, both can observers can distinguish nearly 100 different spectral
be perceived as white due to physiological adaptation and colors. A spectral color is the result of a visual response to a
the psychological phenomenon known as color constancy. single wavelength only. Such colors are the purest possi-
Our eyes easily adapt to any reasonably uniform amounts ble because they are unaffected by mixture with light of
of red, green and blue light in the prevailing illumination. other wavelengths. It is also possible to specify a certain
Color constancy means that our eyes are not reliable for region or color of the spectrum by the bandwidth of the

5
 s ecti o n 1: Captu r e

Figure 1-5 The dispersion of white light into the spectral colors. Figure 1-6 Spectral power distributions for D65 and D55 standard
illuminants.

wavelengths. For example, the red portion of the spectrum Planck’s equation is used to determine the SPD of a
is the region from 600 to 700 nm. It’s estimated that we blackbody radiator at any temperature (see Equation 1.1).
can see approximately 10 million colors when combining All of the terms in this equation are constant except for
wavelengths of energy. However, the visible spectrum is wavelength.
often arbitrarily divided into the seven colors labeled in
Figure 1-5: red, orange, yellow, green, blue, indigo and 2hc2
M= (Eq. 1.1)
violet (ROY G BIV is commonly taught in grammar school). λ (ehc/λkT–1)
5

Our description of the visible spectrum is further simpli-

fied in the photography world into just three regions: red, where
green and blue. h is Planck’s constant of 6.6262 x 10-34 J-s
c is the speed of light with a value of 2.9979 x 108 m/s
λ is the wavelength of energy in meters
Spectral Power Distributions k is Boltzmann’s constant of 1.3807 x 10-23 J/K, and
T is the absolute temperature of the light source in
The ability to separate white light into its individual degrees Kelvin.
wavelengths via dispersion means that we can measure a
light source by wavelength using a spectroradiometer. The Standard illuminants are established models referenced
resulting measurements, plotted as a function of energy for artificial light sources to describe their spectral power
at each wavelength, comprise a spectral power distribution distribution relative to sunlight, our primary natural light
(SPD). The amount of energy emitted across wavelength source. An illuminant is another way to refer to a light
bands is plotted on an SPD plot to accurately describe a source. The standard illuminant called D65 defines the SPD
light source’s energy emission. Figure 1-4 shows the SPD for noon daylight in the shade. D55 is another standard
for two example tungsten and daylight sources. Notice that illuminant which defines noon daylight; we break down
the tungsten line is smooth and continuous (without any the difference between daylight and sunlight later in this
breaks or gaps). As a result, tungsten sources are called con- chapter. Figure 1-6 shows these standard illuminants on a
tinuous sources and their SPDs simulate that of a blackbody combined SPD plot.
radiator.

6
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Solid-Object Sources and

the Heat-Light Relationship

Light is a form of energy that can only be produced from

some other form of energy. The simplest and perhaps most
common method is from heat energy, a process called
incandescence. Whether the source is the filament in a
tungsten bulb, a candle flame, or anything heated until it
glows, incandescence is always associated with heat. The
amount and color of light produced by an incandescent
source is directly related to its temperature. Consider,
for example, an iron poker with one end placed in a fire.
Holding the opposite end feels cold, but as it is left in the
fire its temperature rises and it begins to feel warm. By
increasing the temperature of the poker, we become aware
of a change in its radiating properties through our sense of Figure 1-7 Cross section of a simple blackbody radiator consisting of
touch even as it looks the same. Soon the poker becomes an enclosure surrounded by boiling or molten material.

too hot to touch and we sense its radiation as heat from a

short distance. As the temperature rises even higher, the
poker reaches its point of incandescence and begins to radiation. When such an experiment is performed and the
emit a deep red glow. If the poker is allowed to get hotter blackbody is heated to a variety of temperatures, the spec-
still, the light grows brighter and becomes more yellow- tral power distribution changes.
ish. At extremely high temperatures, the end of the poker
looks white and ultimately blue while becoming brighter
still. This example illustrates that a solid object heated to Vapor Sources and the
its point of incandescence and higher produces light that Effect of Electrical Discharge
varies in color as a function of its temperature. We use the
absolute or Kelvin scale when describing the temperature A fundamentally different method for producing light
of such sources because all objects emit some infrared involves passing an electrical current through gases.
energy at temperatures above absolute zero. Sources that employ this method are called discharge or
The best solid-object radiator is a blackbody as it absorbs vapor lamps and generally consist of a glass tube contain-
all of the energy that strikes it. All light sources are radia- ing an inert gas with an electrode at each end. An electrical
tors, though some are more efficient than others. Since current is passed through the gas to produce light and
a perfectly black object would absorb and emit energy ultraviolet energy. This energy may be used directly or to
but not reflect it when heated, it is the most efficient excite phosphors coated on the inside of the glass tube, as
source. A blackbody source is achieved in practice by the in a fluorescent lamp.
design shown in Figure 1-7 consisting of an enclosure The basic process involved in the production of light is
surrounded by a boiling material. Since the interior surface the same for all vapor lamps. Light emission from the vapor
of the object is darkened and concave, any entering light is caused by the transition of electrons from one energy
is absorbed (immediately or after one or more reflections). state to another. When the electrical current is applied to
Consequently, the hole appears perfectly black. As the the lamp, a free electron leaves one of the electrodes at
walls are heated, they emit radiant energy in all directions. high speed and collides with one of the valence electrons
The energy that escapes through the hole is blackbody of the vapor atom. The electron from the vapor atom is

7
 s ecti o n 1: Captu r e

1 1

0.8 0.8
Relative Energy

Relative Energy
0.6 0.6

0.4 0.4

0.2 0.2

0 0
400 450 500 550 600 650 700 350 400 450 500 550 600 650 700
Wavelength (nm) Wavelength (nm)
Figure 1-8 Spectral power distribution of a low-pressure sodium- Figure 1-9 Spectral power distribution of a low-pressure mercury-
vapor source. This source appears yellow. vapor source. This source appears violet.

bumped from its normal energy level to a higher one and illuminating streets and parking lots at night. Low-pressure
exists for a short time in an excited state. After the collision, mercury-vapor sources are used in greenhouses as plant
the free electron is deflected and moves in a new direc- lights because the ultraviolet energy they emit is beneficial
tion at a reduced speed. However, it excites several more to plant growth.
electrons before it completes its path through the lamp. Perhaps the most common vapor sources in daily life are
The excited electron eventually drops back to its former fluorescent lamps. These are typically low-pressure mer-
energy level and emits some electromagnetic radiation in cury-vapor tubes with phosphors coated on the inside of
the process. the glass envelope (another name for the bulb enclosure).
Each type of vapor atom, when excited, gives off energy When bombarded by the large amount of ultraviolet radia-
at wavelengths determined by its structure. Some gases tion emitted by the mercury vapor, these phosphors begin
emit radiation only at a few wavelengths while others emit to glow and give off visible energy at all wavelengths in
energy at many different wavelengths. These sources are the visible spectrum. Thus, the light emitted by a fluores-
said to show a discontinuous spectrum represented by a line cent lamp is the result of both the discontinuous energy
on a spectral-energy plot. For example, the spectrum of emitted by the vapor and the continuous energy emitted
sodium vapor shows a bright yellow line near 600 nm, as by the fluorescing phosphors.
shown in Figure 1-8, while mercury vapor produces energy There are many classes of phosphors that can be used
at many different wavelengths, both in the ultraviolet for this purpose. Each phosphor emits its own color of
region and in the visible region, shown in Figure 1-9. The light. Figure 1-10 illustrates the spectral-energy distribution
spectral characteristics of the emitted light are primarily for a typical cool white fluorescent lamp. The light that
dictated by the vapor in the tube. corresponds to the discontinuous line spectrum produced
The pressure under which the vapor is dispersed in by the mercury vapor may not be apparent to a human
the tube has a significant effect on the amount of energy observer because of color constancy, described in Chapter
emitted. Vapor lamps are categorized as low pressure or 13.
high pressure; low pressure indicates some small fraction Fluorescent lamps excite gas molecules within the
of atmospheric pressure while high pressure describes tube with electrons to produce energy of discrete wave-
sources working above atmospheric pressure. High-pres- lengths, largely in the blue and ultraviolet regions (primarily
sure sodium-vapor lamps are traditionally used for dependent on the gas elements in the tube). Some of this

8
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

The blood glows when it comes in contact with the iron

in luminol. When chemiluminescence happens in a living
organism, it is called bioluminescence. Bioluminescence is
the result of biochemical reactions and is seen in fireflies
and dinoflagellates (a type of plankton).
Triboluminescence occurs when rubbing breaks the
asymmetrical bonds in a crystal, crushing or breaking the
material. Examples include grinding sugar with a mortar
and pestle or the sparks seen when biting into a Wint-O-
Green Life Savers Mint in a dark room. Solid substances that
luminesce are called phosphors.
Photoluminescence includes fluorescence and phospho-
Figure 1-10 Spectral power distribution of a cool white fluorescent rescence and is caused by a chemical compound absorbing
lamp. photons, which elevates it to a higher energy state. Here
the light photons are given off until the compound returns
to its original energy state. The length of time this takes is
energy is absorbed by phosphors coated on the inside of typically around 10 nanoseconds.
the tube and is converted to longer-wavelength, visible Fluorescence is an emission of electromagnetic radiation
radiation. The color of the fluorescent emission is highly in the visible region that occurs during the time of excita-
dependent on the phosphor material and the activators tion. Thus, phosphors that are irradiated with ultraviolet
incorporated in it. energy fluoresce. The fluorescence emission ceases quickly
when the excitation is removed: somewhere between 10–8
and 10–1 seconds depending on the activator. Fluorescence
Luminescence, Fluorescence and imaging is a key tool for imaging microscopic structures
Phosphorescence and biological processes.
Phosphorescence is similar to fluorescence but with a
Luminescence is the emission of light from a substance by slower rate of decay. An example of phosphorescence can
means other than incandescence. Luminescence is a form be found in glow-in-the-dark children’s toys. Phosphores-
of cold body radiation as it usually occurs at low temper- cence emission occurs after the excitation stops and lasts
atures. It is generally caused by the emission of photons anywhere from 10–8 seconds to several hours. Its duration is
from atoms excited by energy of relatively short wave- largely dependent on temperature.
lengths. As the electrons return to a lower energy level, Some dyes fluoresce, including fluorescein, eosin,
energy of longer wavelengths is released. The rate at which rhodamine and a collection of materials used as optical
this occurs can be affected by the presence of an activator brighteners in materials like paper and fabric. Many modern
made up of ionized atoms that trap and release electrons photographic and digital printing papers are treated with
slowly for recombination. The exciting energy is usually in brighteners to give them cleaner and more brilliant white
the ultraviolet region but can be caused by energy in the appearances. This is a distinguishing characteristic when
visible and infrared regions as well. There are many forms of comparing modern prints to those made 40 or 50 years
luminescence. ago that did not use fluorescing agents. Fluorescence and
Chemiluminescence is the result of a chemical reaction. A phosphorescent dyes are also used in organic light-emitting
common example of chemiluminescence is found in foren- diode (OLED) technologies that are becoming common-
sics laboratories that use luminol, a chemical solution that place as display and lighting solutions.
indicates the presence of blood on a material or surface.

9
 s ecti o n 1: Captu r e

Methods of Measuring Light

Candles were the standard light source when the field

of photometry was established in the early 1900s. Con-
sequently, many of the definitions and units of light
measurement were based on the candle. It wasn’t until
1940 that the international standard unit for light meas-
urement became defined based on the light emitted by
1 square centimeter of a blackbody radiator heated to the
temperature of solidification of platinum. This unit is called
the candela and it’s more exact and reproducible than
relying on candles.
Starting with a clearly defined light source is important
Figure 1-11 Polar coordinate plot of intensities for a lamp-reflector
but we also need to know exactly how to measure that combination in candelas.
light source. Do we take a single measurement directly
next to the source? Should we take multiple measurements
all around a room lit by that source? What is the end goal Since the intensity of a light source changes with direc-
for taking this measurement? Let’s review the options. tion, it helps to know the distribution of intensities. This
information is often provided by the lamp manufacturer
in the form of a two-dimensional graph based on polar
Intensity coordinates, an example of which is shown in Figure 1-11.
Intensity (also called luminous intensity) is a measure of the Intensity is plotted against the angle using a special graph
rate at which the source emits light in a given or single called a polar-coordinate plot. The zero angle is head-on
direction. Initially, the intensity of a source could be deter- with the lamp, with the intensity in this direction known as
mined by comparison to the old standard candle, with the the beam intensity. A lamp’s intensity at any desired angle is
result expressed as candlepower. A 50-candlepower source derived from this plot.
emits light equivalent to amount of light from 50 standard The lamp-reflector combination illustrated in Figure 1-11
candles. The modern equivalent is the candela. One candela indicates that the beam intensity is approximately 800 can-
is one-sixtieth the light from 1 square centimeter blackbody delas. This intensity is nearly uniform within 25° on either
heated to the freezing temperature of platinum. Admittedly side of the beam position, indicating that this lamp-re-
more of a mouthful, but decidedly more precise. flector combination provides nearly uniform illumination
Tungsten lamps and fluorescent tubes always vary over a 50° angle of projection. At 65° off the beam position,
in their intensity with direction and therefore no single the intensity drops to nearly 400 candelas. The same lamp
measurement of intensity can completely describe them. equipped with a more narrowly curved reflector produces
Perhaps the easiest way to understand the concept of a distribution of intensities much narrower than what is
intensity is to think of a ball-shaped lawn sprinkler with shown in Figure 1-11. Understanding the distribution of
many holes through which the water sprays. Measuring the intensity is particularly helpful when using artificial light
intensity of a light source is similar to measuring the rate at sources in photography because it may mean the differ-
which water is sprayed through one of the sprinkler holes ence between spotlighting a subject’s face in a portrait and
in a specific direction. Such information is of limited value lighting an entire set with an even glow.
since it does not describe the variation in intensities around There are three ways in which the intensity of a light
the ball nor the total amount of water being emitted by the source can be reported as a single value illustrated
sprinkler. in Figure 1-12. In the simplest case, the intensity is

10
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Figure 1-12 Three ways to measure light source intensity. Figure 1-13 The relationship between candlepower or candelas
and lumens.

measured in a single direction and reported as a single Flux

candela value. When a number of readings are taken at Another method of measuring light uses flux (sometimes
uniform intervals on a horizontal plane around the source called luminous flux), the rate at which a source emits light
and then averaged, the result is the mean horizontal inten- in all directions. The flux of a light source is calculated
sity (candlepower) of the light source. Instead of taking a from intensity measurements and is closely related to
large number of individual readings, this result is obtained the measurement of mean spherical intensity. The unit of
by rotating the source rapidly upon its vertical axis while measurement for flux is the lumen. The lumen is defined
repeated readings are taken. The intensity of light in all as the amount of light falling on a surface 1 square foot in
directions is determined by measuring intensities at uni- area, every point of which is 1 foot from a uniform source of
form intervals around the light source. An average of these 1 candela. It is a three-dimensional concept as it concerns
readings gives the mean spherical intensity (candlepower) the light output in all possible directions in space. The rela-
of the illuminant. Note that this value is related to the total tionship between the candela and the lumen is illustrated
light output of the lamp. In each of these cases, the inten- in Figure 1-13.
sity is determined through comparison to a standard lamp If the opening indicated by A, B, C, D in Figure 1-13 is
at a variety of distances. 1 square foot of the surface area of a sphere with a 1-foot
radius, the light escaping is 1 lumen. The light escaping
becomes 2 lumens if the area of the opening is doubled.

11
 s ecti o n 1: Captu r e

The total surface area of a sphere with a 1-foot radius

is 12.57 square feet (that is, 4πr where the radius is 1 foot).
It follows that a uniform 1-candela source of light emits
a total of 12.57 lumens. An area of 1 square foot on the
surface of a sphere with a 1-foot radius subtends a unit
solid angle (1 steradian) at the center of that sphere. The
lumen is also defined as the amount of light emitted in a
unit solid angle by a source with an average intensity of 1
candela throughout the solid angle. All of this is to say that,
when considering a point source that emits light equally in
all directions, there are 12.57 (4π) lumens of flux for every
candela of intensity.

Illuminance
Candelas and lumens are both measurements of the light
source but neither one considers the effect of the light on
a surface or environment, only the source of it. For this,
we turn to illuminance, the light incident upon a surface.
Illuminance is measured in footcandles. A footcandle is the
illumination at a point on a surface that is 1 foot from, and
perpendicular to, the light rays from a 1-candela source. For
Figure 1-14 The relationship between intensity and illuminance at a
example, if the light source in Figure 1-14 has an intensity constant distance of 1 foot.
of 1 candela, the illuminance at point A (1 foot distant from
the source) is equal to 1 footcandle. The illuminance at
points B and C is less because they are at distances greater is perpendicular to the rays of a 1-candela source placed
than 1 foot. Therefore, an illuminance reading applies only 1 foot away. This illustrates an important relationship
to the particular point where the measurement is made. between lumens and footcandles. A lumen is the light
We determine the average illumination by averaging the flux spread over 1 square foot of area that illuminates that
footcandle measurements at a number of locations at our area to a level of 1 footcandle. Therefore, 1 footcandle is
surface. equal to 1 lumen per square foot. This forms the basis of a
The footcandle is the unit of measure most closely asso- simplified approach to lighting design known as the lumen
ciated with the everyday use of light. To get an idea of how method. When the square footage of an environment is
much light 1 footcandle is, hold a lit candle 1 foot from a known and the desired level of illumination determined, it
print in a dark room. The result is approximately 1 foot- is simple to determine the number of lumens required.
candle of illumination. A full moon on a clear night emits With a 1-candela source, as shown in Figure 1-15, the
approximately 0.02 footcandles; a well-lit street at night level of illumination on point A 1 foot away is 1 footcandle.
gives approximately 5 footcandles; a well-lit classroom If plane A is removed and the same beam of light passes
has nearly 50 footcandles of illumination. Direct sunlight is on to plane B, which is 2 feet away, this same beam of light
approximately 12,000 footcandles. now covers an area four times that of plane A. The average
Referring again to Figure 1-13, the surface A, B, C, D illumination on plane B is one-quarter as great as that on
fulfills the conditions for a surface illuminated to a level plane A, which is equal to one-quarter of a footcandle. If
of 1 footcandle. Every point on this square foot of surface the beam of light is allowed to fall upon plane C, 3 feet

12
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Figure 1-15 The relationship between intensity and illuminance for a constant
intensity and varying source-to-surface distances, illustrating the inverse square law.

away from the source, it’s spread over an area nine times as this law. Equation 1.2 defines the formula for the inverse
great as plane A, and so on. square law:
From this we conclude that illumination falls off (or
I
decreases) not in proportion to the distance but in pro- E= (Eq. 1.2)
d2
portion to the square of the distance. This relationship is
known as the inverse square law. Put another way, this law Where E is the illuminance in footcandles, I is the inten-
shows us that doubling the distance of a light source from sity of the source in candelas and d is the distance of the
our subject results in one-quarter of the light. We instinc- surface from the source in feet. Illuminance measurements
tively know that moving a light farther away decreases the can be used to determine the intensity (expressed as
light available to make an exposure, though we might not candelas) of a source by solving the above equation for I as
intuitively grasp just how dramatic and rapid that decrease shown in Equation 1.3:
is as the distance grows. The light travels farther and also
wider in its scope (i.e. the area illuminated increases with I = E x d2 (Eq. 1.3)
distance). The inverse square law is based on a point source
of light from which the light rays diverge. In practice, it An alternative unit for measuring illuminance in photogra-
applies with close approximation when the diameter of the phy is the metercandle. The definition of a metercandle is
light source is no greater than approximately one-tenth the similar to that of a footcandle except the distance from
distance to the illuminated surface. In other words, a light the source; S to point A in Figure 1-14 is 1 meter. One
behaves according to the inverse square law assuming it metercandle is the amount of light falling on a surface at a
approximates a point source: the illumination must derive point 1 meter from a 1-candela source. There are approx-
from a very small point (a tungsten bulb is one example). If imately 3.28 feet in 1 meter, making 1 metercandle equal
a light source appears to emit light from a broad area, such to 0.0929 footcandles. Or, 1 footcandle is equal to 10.76
as a softbox in the studio, it does not behave according to metercandles. The term metercandle is not used with great

13
 s ecti o n 1: Captu r e

Table 1-2 Standard illuminance levels for viewing reflection prints and
luminance levels for viewing transparencies.

ANSI PH2.30-1989

Comparison viewing/critical appraisal 2,200 ± 470 lux

Display, judging and routine inspection 800 ± 200 lux

Transparency viewing 1,400 ± 300 cd/m2

ISO 3664:2009

Prints and Proofs 2,000 ± 250 lux (preferred)

2,000 ± 500 lux (required)

Transparencies 1,270 ± 160 cd/m2(preferred) Figure 1-16 Measuring luminance as a function of intensity and
1,270 ± 320 cd/m2 (required) surface area.

frequency today; the preferred term is lux. One lux is equal in the direction of point A. This lamp’s projected area is
to 1 metercandle—an easy conversion to remember. 5 square inches. The luminance is equal to 100 candelas
When exhibiting photographic prints, the level of illu- divided by 5 square inches, or 20 candelas per square inch.
mination in the environment can have a significant impact Luminance is the photometric quantity that relates
on their appearance. Consequently, a number of standard closely to the perceptual concept of brightness. Brightness
conditions are specified for viewing purposes and are is used exclusively to describe the perceived appearance
summarized in Table 1-2. The variety of levels suggested of a source to the human eye and cannot be directly
is due in part to the differing visual tasks being performed. measured.
For example, the viewing of prints for comparison purposes Generally, as the luminance of a source increases, so
requires a higher level of discrimination and, therefore, does the brightness of that source. If two 60-watt tungsten
a higher illuminance level than does the viewing of a lamps are placed side by side—one lamp is frosted while
pictorial display. Further, since print judging and display the other is clear—the clear lamp looks much brighter.
involve subjective opinions, it should not be surprising that Both bulbs use the same amount of electrical energy and
a variety of standards exist. they both use a tungsten filament, so the intensities of the
two lamps should be the same. However, the luminance
of the clear bulb is greater because the projected area
Luminance of the filament is smaller than the projected area of the
The rate at which the unit area of a source emits light in glass envelope on the frosted lamp. We conclude from
a specific direction is called luminance. If a source is not a this observation that knowing a light source’s intensity is
point source but has an appreciable size (as all real sources less valuable than knowing its luminance. Consequently,
do), it is less useful to describe the intensity of such a luminance data for real sources is always preferred over
source than to specify its luminance. Luminance is derived intensity data when we want to better understand their
from intensity measurements, which are then related to the visual appearance.
projected surface of the source. Measuring luminance applies to reflecting and transmit-
Luminance is expressed in candelas per unit area (can- ting surfaces as well, since it makes no difference whether
delas per square foot, candelas per square inch or candelas the surface being considered is emitting the light or merely
per square centimeter depending upon the size of the reflecting or transmitting it. In this respect, if all of the light
surface area considered). For example, Figure 1-16 shows falling on a perfectly diffusing surface were re-radiated by
a frosted tungsten lamp with an intensity of 100 candelas the surface, the luminance would be numerically equal to

14
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

the illuminance. This does not happen in practice because The footlambert was invented to avoid the need to
real surfaces never reflect 100% of the light that strikes divide by π all of the time. The footlambert is defined as
them. For this reason, it is necessary to determine the one divided by π candelas per square foot. Thus, the previ-
reflection factor of the surface, the ratio of reflected light to ous relationship reduces to:
incident light. The following formula is used:
L=KxE (Eq. 1.6)
Reflected Light
Reflective Factor (K) = (Eq. 1.4)
Incident Light where L is now expressed in footlamberts.
The footlambert is actually defined as the luminance of a
Applying this formula to a perfectly diffusing surface, 100% reflecting surface illuminated by 1 footcandle of light.
luminance equals illuminance multiplied by the reflection Therefore, the luminance of the previously described sur-
factor. The most commonly used unit of luminance when face is calculated by multiplying 0.18 by 100 footcandles,
considering reflecting surfaces is candelas per square foot which equals 18 footlamberts. Although the calculations
and the formula is: are simpler for footlamberts, modern light meters read
luminance directly in candelas per square foot. Candelas
KxE
L= (Eq. 1.5) per square foot is the more commonly used unit of light
π
measurement in photography.
where L is the surface luminance in candelas per square Note that these examples assume that the illuminated
foot, E is footcandles incident on the surface and K is the surfaces are perfectly diffusing or reflecting equally in all
reflection factor of the surface. directions. This is approximately true for matte surfaces.
As illustrated in Figure 1-17, the product of the However, shiny surfaces give highly directional reflections
reflectance and the illuminance must be divided by π and do not follow these formulas.
(3.14) since the light is emitted into a hemisphere of Luminance is particularly useful in photography since
unit (1-foot) radius and IT is the ratio of the radius to the it provides a way of describing the light reflected from the
surface area of the hemisphere (IT = A/(2r)2). For example, surfaces of the subject being photographed. Whenever
if a perfectly diffusing surface with 18% reflectance is a reflected light meter reading is made, we’re measuring
illuminated with 100 footcandles of light, we calculate luminance. Luminance measurements have the unique
the luminance of that surface by multiplying 0.18 by 100 characteristic of being independent of the distance over
divided by 3.14, which equals 5.73 candelas per square which the measurement is made. For an example, if we
foot. use a handheld meter with a 50° angle of view to measure
the reflected light from a surface, the luminance value
obtained is identical to one taken with a spot meter from
the camera location (light meters are discussed further in
Chapter 4). The amount of light measured from the original
area decreases with increasing distance and the projected
surface area included in the angle of view increases in
direct proportion. Thus, the number of candelas per square
foot remains the same. This assumes a clear atmosphere
exists between the meter and the surface area being meas-
ured, which is usually the case. The various conversions
of illuminance data to luminance data are summarized in
Figure 1-17 A perfectly diffusing surface; sometimes referred to Table 1-3.
as a Lambertian surface.

15
 s ecti o n 1: Captu r e

Table 1-3 Illuminance–luminance conversions based on an 18%

reflectance neutral test card.

Illuminance to Luminance

Luminance = illuminance × reflectance

1. Footlambert = footcandle × reflectance

Luminance = 1.0 footcandle × 0.18 = 0.18 footlambert

2. Candela per square foot = footcandle × reflectance/π

Luminance = 1.0 footcandle × 0.18 = 0.18 / 3.1416 = 0.057 Cd/ft.2

3. Candela per square meter = metercandle × reflectance/π

Luminance = 1.0 metercandle × 0.18 = 0.18 / 3.1416 = 0.057 Cd/m2

Luminance to Illuminance

Illuminance = luminance / reflectance

1. Footcandle = footlambert / reflectance

Illuminance = 1.0 footlambert / 0.18 = 5.56 footcandles

2. Metercandle = apostilb / reflectance

Illuminance = 1.0 apostilb / 0.18 = 5.56 metercandles

3. Footcandle = candela per square foot × π/reflectance

Illuminance = 1.0 candela/ft.2 × 3.1416 / 0.18 = 17.45 footcandles

4. Metercandle = candela per square meter × π/reflectance

Illuminance = 1.0 candela/m2 × 3.1416 / 0.18 = 17.45 metercandles

Notes:
1. A perfectly diffusely reflecting surface (100%) illuminated by 1 footcandle
(1 lumen per square foot) reflects 1 footlambert (1 lumen per square foot or
1/π candela per square foot).
2. Metric: A perfectly diffusely reflecting surface (100%) illuminated by
1 metercandle (1 lumen per square meter or 1 lux) reflects 1 apostilb Figure 1-18 The range of color temperatures from
(1 lumen per square meter or 1/π candela per square meter). warm to cool appearances.

Color Temperature

Since the amount and color of radiation emitted by a meaning the plot of their spectral power distributions are
solid-object source is temperature-dependent, the color of continuous (see Figure 1-4). This presumption is correct for
light emitted from such a source can be completely spec- incandescent illuminants. However, the color of light emit-
ified by the Kelvin temperature (K) at which it is operating. ted by some sources, such as fluorescent lamps, has no
Such a rating is referred to as the color temperature of the relationship to the operating temperature (see Figures 1-8,
source. A color temperature is assigned to a light source by 1-9 and 1-10). In these cases, the measurement correlated
visually matching it to a blackbody radiator. The tempera- color temperature (CCT) is required to indicate that the color
ture of the blackbody radiator is raised until the color of its of light emitted by a blackbody radiator of that tempera-
light visually matches that from the light source. The Kelvin ture produces the closest visual match that can be made
temperature of the blackbody is then assigned as the color with the source in question.
temperature of the light. The use of color temperature in The color temperatures of common light sources are
photography presumes that the light source adheres to the shown in Figure 1-18. Counterintuitively, the color of a light
same heat-to-light relationship as the blackbody radiator, becomes cooler or colder (read: more blue, less yellow) as

16
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Table 1-4 Color temperatures of common light sources.

Light Source Temperature (K) Mired Value

Candle 1800–1900 555–526

Continuous LED studio lights 2700–6500 (variable) 370–153

Tungsten lamp 2800 357

CIE Tungsten illuminant or Illuminant A 2856 350

CIE Cool White Fluorescent 4100 244

Direct sunlight 4500 222

CIE D50 or horizon light 5000 200

Electronic flash 5500–6000 182–166

CIE D65 or noon daylight 6500 154

Sky light (overcast sky) 8000 125

North sky light ~15,000 67

the color temperature increases. This flies in the face of our reference image (an 18% gray card works well) from which
everyday experiences with temperature—usually hotter to calibrate a photograph’s accurate color reproduction.
things look more yellow or orange than blue. It is consist- Light-emitting diode (LED) panels and bulbs are
ent, though, with the property of light that tells us shorter increasingly popular for photographic use for a variety of
wavelength radiation has more energy. functional and practical reasons (weight, cost, minimal heat
Four CIE Standard Illuminants are listed in Table 1-4. output). LED sources are particularly interesting in this con-
Standard illuminants are theoretical light sources and do text because, unlike tungsten sources, varying the voltage
not represent specific consumer products. Illuminant A, applied (i.e. using a dimmer control) does not change their
introduced in 1931, represents a typical tungsten-filament color temperature. LED panels offer variable color tempera-
light with a color temperature of 2856K. CIE Cool White ture capabilities by using red, green and blue LEDs to make
Fluorescent or F2 is one of 12 defined fluorescent stand- white light. Varying the brightness of each RGB cluster pro-
ards. D50 and D65 represent daylight at 5000K and 6500K duces white light with a cooler or warmer color cast. This is
respectively. D50 is used in the printing industry while a handy feature when looking to match a light’s color tem-
D65 is more often used in photography as it is a practical perature to an environment’s existing light. Some panels
stand-in for average midday light. feature amber and white LEDs in addition to red, green and
Photographers are faced with a tremendous range of blue (referred to as RGBAW) to ensure a pure, neutral white
color temperatures, from the yellowish color of a candle and warmer color temperature output options.
at about 1800K to the bluish appearance of north skylight, Color temperature meters employ three silicon photo-
rated at 15,000K. Though Figure 1-18 shows an approxi- diodes filtered for red, green and blue that approximate
mate color temperature for a tungsten source, the actual the sensitivity of image sensors. These meters measure
color temperature encountered can vary significantly as a the amount of energy present in each of the three regions
result of reflector and diffuser characteristics, changes in of visible light and determine the color temperature by
the power supply and the age of the bulb. Consequently, finding the closest match to a blackbody curve. They work
it is more accurate to measure a light source directly with a best with continuous-spectrum light sources. If a color
color temperature meter or to photograph a neutral-tone temperature meter is used to measure a vapor source that

17
 s ecti o n 1: Captu r e

produces a discontinuous spectral power distribution, the the camera’s menu as a reference photograph. An 18% gray
results can be misleading and inaccurate due to significant card or neutral patch on a ColorChecker are ideal subjects
differences in the measured response compared to the because they are specifically created for colorimetric accu-
blackbody curves. racy. The camera uses this reference image to understand
All color photographs have some notion of a neutral the color temperature of the scene and assigns a custom
reference point in order to faithfully render the colors of value for use on all subsequent photographs. In effect, we
a scene. In order to do this, cameras estimate or assume help to calibrate the camera’s understanding of the color
the color temperature of the primary light in the scene. quality of our scene. A third option is to disregard the issue
Sometimes this works automatically without issue, but not altogether and wait until post-processing—not the most
always. Photographers have several options when the color efficient workflow but one that does not negatively impact
temperature of the light does not match the response of color reproduction or image quality in any way.
the camera’s sensor or if the software interpretation gets
it wrong. Note that while the term color temperature is
descriptive of the light source’s appearance relative to a The Mired Scale
blackbody radiator, color balance is used to describe the
color appearance of light rendered in a photograph. The Although color temperature provides a useful scale for clas-
term is often used interchangeably within the context sifying the light from continuous-spectrum sources, it has
of making images because we rarely care about a light limitations. A 500K change at the 3000K level, for example,
source’s characteristics for their own sake, we’re interested does not produce the same visual or photographic effect
in how it impacts any resulting photographs. Since a as a 500K change at the 7000K level. This is because of
balance of color necessitates a definition of a neutral tone, the nonlinear relationship between changes in a source’s
like white, it’s also called white balance. All three terms are color temperature and the changes in the color of light it
reasonably used when describing a photograph, though produces, which is illustrated in Figure 1-19.
color balance and white balance are not used to describe a This awkwardness is eliminated using the mired value. The
light’s character. mired value is the reciprocal of the color temperature and
Cameras have several built-in white balance settings that exhibits a nearly linear relationship between the numeric
correspond to commonly encountered color temperatures value and the visual change. Mired is an acronym for the
such as sunny daylight, cloudy daylight, indoor incan-
descent or indoor fluorescent. We can choose the white
balance setting that appropriately corresponds to the scene.
This may or may not give satisfactory results, especially
knowing that there is considerable variability in the color of
outdoor sunlight or indoor light bulbs. Take the example of
an indoor incandescent light: the published color tem-
perature for a light bulb in a living room is 2850K and the
incandescent setting on our camera is calibrated to 3000K
(an averaged value). This discrepancy may cause a color cast
in the image. The second option available is establishing
a custom white balance setting. It requires an extra step,
but the results are much more likely to accurately render
a scene’s color. Setting a custom white balance involves
taking a photograph of a neutral-toned object (white or Figure 1-19 The relationship between color temperature and the
gray works best) in the scene and selecting this image in relative amount of blue light emitted by a solid-object source.

18
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

term micro-reciprocal degrees. The reciprocal of color tem-

perature value is quite small, so multiplying it by one million
gets us a number that’s easier to intuit. Convert color tem-
peratures to the mired scale using the following formula:

1,000,000
mired = (Eq. 1.7)
color temperature

Color temperature meters can read out directly to mired

values. Decamired is sometimes used for convenience;
each decamired containing 10 mireds. For example, a
mired value for a color temperature of 125,000K is 80, or 8
Figure 1-20 The relationship between mireds and the relative amount
decamireds.
of blue light emitted by a solid-object source.
Figure 1-20 illustrates the relationship between changes
on the mired scale and changes in the resulting effect.
It shows that equal changes in the mired scale produce
equal changes in the effect. The relationship between
color temperature and mireds is illustrated in Figure 1-21,
which indicates that the higher the color temperature, the
lower the mired value. Bluish sources with very high color Figure 1-21 A comparison of color temperature and mired scale.

Table 1-5 Light measurement terms with typical examples of each. Note: Concepts 1–8 apply to light sources. Concept 8,
Luminance, also applies to reflected and transmitted light. Concepts 9–10 apply to light falling on a surface.

Popular Concept Technical Term Symbol Unit Abbreviation Measurement (Practical)

1. Strength Luminous I Candela c Compute from illuminance and

intensity distribution

2. Strength Luminous flux F Lumen lm (Manufacturer data) or estimate watt ×

luminous efficiency

3. Strength/watt Luminous µ Lumens/watt lm/W (Manufacturer data)

efficiency

4. Total light Luminous energy Q Lumen-second lm-sec (Manufacturer data) or integrate area
under curve

Spectrometer (electronic) Spectroscope

5. Color Wavelength λ Nanometer nm
(visual) Spectrograph(photographic)

6. Color Balance Color CT Degree Kelvin K (Manufacturer data) or color-temperature

Temperature meter

7. Source conversion Mired shift (N/A) Mired MSV (Manufacturer data) or compute

8. Brightness Luminance L Candela/sq. foot c/ft.2 Reflected-light meter

9. Illumination Illuminance E Footcandle ftc (fc) Incident-light meter with flat surface
Metercandle Lux mc

10. Exposure Photographic H Footcandle- ftc-sec Compute H = illuminance × time

exposure second mc-sec
Metercandle-
second

19
 s ecti o n 1: Captu r e

temperatures have very low mired values, while red-

dish-appearing sources with low color temperatures have
high mired values.
In the past, photographers used the mired scale to
determine which correction filters to place over the lens
when the color balance of a light and the film were mis-
matched. In the digital realm, we use the mired scale to
make corrections to the in-camera white balance. Since we
discussed how color temperature values are not linear to
the appearance of color changes, the mired scale is the log-
ical and intuitive choice for user-controlled tweaks to color
balance. We’ve encountered cameras with color balance
menus allowing for a manual shift in color balance with
mireds as the unit of change (i.e. +5 Green, –10 Magenta).
Table 1-5 summarizes many of the light measurement
terms and presents typical values for comparison pur-
poses. It is obvious from this table that light sources can
be described in a wide variety of ways. It is important to
understand the differences in order to obtain appropriate Figure 1-22 Spectral power distributions for two fluorescent lamps
information. with the same correlated color temperature (4200K). Their CCT values
can be the same despite the differing appearance of objects lit by
them.

Color Rendering Index

Color temperature is a great way to quantify the visual the perceived color of an object when compared to an
appearance of a light source. It does not tell us anything ideal, standard light source. For light sources with a color
about an object’s color appearance under that same temperature between 2000K and 5000K, the standard is the
source. Since an object’s color is a function of the light blackbody radiator. For sources above 5000K, the stand-
it reflects, it stands that its color appearance can change ard is D65. Two evaluated sources can only be compared
based on the light source’s spectral emission. The following to each other if their reference sources are similar (within
example illustrates why this matters. approximately 200K). Spectral power distribution is the key
The spectral-energy distributions for two fluorescent characteristic dictating a light’s CRI: the more a light source
lamps of equal correlated color temperature are shown in differs from the spectral power distribution of natural day-
Figure 1-22. The Cool White Deluxe lamp produces more light, the lower its CRI is likely to be. Color rendering index
red light than the Cool White lamp, resulting in widely and color temperature are independent characteristics.
different color rendering for some objects. If a person’s The color rendering index was designed to evaluate
face is illuminated by these sources, the Cool White Deluxe continuous-spectrum tungsten or incandescent light
lamp casts more red light, some of which reflects back to sources. The reference source always has a CRI of 100. The
the camera. This results in a healthier, more natural skin higher the CRI for a given source, the more closely object
complexion appearance than that given by the Cool White colors appear to their counterparts under the reference
lamp. source and the better the color rendition. A CRI value of
The CIE developed the color rendering index (CRI) in 1965 100 indicates that there is no color difference between the
to quantify a light source’s ability to accurately reproduce standard and the evaluated light source, offering perfect

20
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Table 1-6 Color temperatures and color rendering indexes for a variety of light sources. There are often premium models alongside
standard versions offering higher CRIs.

Bulb Type Color Temperature Color Rendering Index

Tungsten: Warm White 3000K 53

Tungsten: Warm White Deluxe 2900K 75

Tungsten: White 3500K 60

Tungsten: Cool White 4200K 66

Tungsten: Cool White Deluxe 4200K 90

LED: Daylight 5000K 85

LED: Soft White 2700K 80

Tungsten Halogen 2850K 100

Compact Fluorescent 2700K–6500K 80

Mercury Vapor 4200K 20

color rendering. A score of 80–90 is very good; anything by color science experts today and the eight color samples
over 90 is an excellent color rendering light source. A have proven to work poorly under modern light sources
negative CRI value can exist, though this should not be sur- like LEDs. The color rendering index calculation was last
prising to anyone that’s stood out in a parking lot at night: revised in 1995. The CIE recently called for an additional
sodium vapor lights are one example.1 These illuminants revision, but one has not yet been established.2,3 The Illumi-
are particularly poor in producing accurate color appear- nating Engineering Society published a promising metric in
ances. Photographing under high CRI-scored light sources 2015 aimed at replacing CRI called TM-30; it is not officially
is not required for color photography in general, but it endorsed by CIE.4,5
is desirable when color accuracy counts. Table 1-6 lists
common light sources, their color temperatures and their
color rendering index values. The Color Quality Scale
Determining the color rendering index for a light source
requires knowing the spectral power distribution and the The National Institute of Standards and Technology (NIST)
color temperature (or CCT in the case of non-continuous recently developed a new method for evaluating a light
spectrum sources) so that an appropriate reference source source’s color rendering ability that addresses the short-
can be selected. First, we measure the CIE chromatici- comings of CRI called the Color Quality Scale (CQS).6 CQS
ties of eight Munsell color samples with low to medium and CRI are similar in that they use averages of color differ-
chromatic saturation under the evaluated source (more ences. They differ in that the CQS uses 15 color samples to
on what this measurement means in Chapter 14). Next, we the CRI’s eight.
do the same under the reference source. The differences NIST selected the 15 samples from the Munsell Book of
between the two sets of data indicate the color shift; the Color. All of them have the highest chroma available for
eight differences are averaged to arrive at the color render- their hues (remember, CRI uses low and medium chroma)
ing index. This method can be misleading because poor and sample the Munsell hue circle with approximately even
scores on a few of the samples can still yield a desirable CRI. spacing. They selected these samples to more accurately
The calculation also uses a color space considered obsolete span the color of normal objects and to include saturated

21
 s ecti o n 1: Captu r e

colors that include the peak wavelengths of modern light 3. Shadows: the darker areas blocked from direct light by
sources. These samples are also commercially available the subject. Note how this area is not devoid of light
from X-Rite Incorporated, the owner of Munsell. The CQS entirely but instead illuminated by indirect light bounc-
uses the 1976 L*a*b* (CIELAB) color space for its calcula- ing off of the environment.
tions because it is the CIE-recommended color space and
is considered visually uniform. The Munsell System and Identifying light in an environment or crafted studio
CIELAB systems are also discussed in Chapter 13. setup is important for two reasons. First, it helps avoid
At the time of publication, the CIE recognizes the flaws making poor metering decisions. Making light measure-
with CRI and is investigating an update to the CRI metric. ments from a specular light source can be misleading due
They do not yet endorse the CQS but published the fol- to its high contrast between bright and dark areas. Second,
lowing statement on revisions or new approaches to color the appearance of light and its characteristics relative to
rendering metrics: “The CIE . . . recommends that important the illuminated subject influences our perception of that
lighting metrics such as the Colour Rendering Index require subject. Since specular highlights are a consequence of
formal international agreement. New metrics introduced reflectivity, we’re highly attuned to seeing them when a
at the regional level could cause confusion in the global surface has a glossy or shiny texture. This could be due
lighting market.”7 to the material (metal, mirrors) or the subject’s condition

Describing the Appearance of Light

We describe the appearance of light as being either specular

or diffuse. Point sources (bare light bulbs, bright sun on a
clear day) produce specular, directional light. This type of
light is characterized by having hard-edged, well-defined
shadows, bright highlights and specular highlights in
reflective areas or surfaces. Reflective surfaces reflect a large
percentage of the incident light rays in an orderly fashion
which leads to such bright highlights. Broad sources (studio
softboxes, overhead office lighting) produce diffuse, non-di-
rectional light. This type of light is characterized as having
soft, ill-defined shadows (if it produces any at all) and low
contrast between highlights and shadows.8 Rough, uneven
or textured surfaces act to further scatter incident light and
lead to diffuse reflections even if that light is specular.
Figure 1-23 shows a pair of tomatoes photographed
with a point light source and with a diffuse light source in
a studio. In addition to identifying that the primary light
source is specular, we also identify:

1. Highlights: the bright, directly illuminated portion of the

subject.
2. Specular highlights: the brightest and most reflective Figure 1-23 The same subject can look different when lit primarily by
parts of the subject, often rendered as pure white. a point light source (top) or diffuse light source (bottom).

22
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

(freshly polished, wet). We subconsciously infer a sub-

ject’s texture or material composition by the presence or
absence of these specular highlights.
Earlier, we defined skylight as the light scattered by
molecules in the Earth’s atmosphere. This indirect light
is important because it adds a diffuse quality to sunlight.
Direct sunlight acts as a point source because it is one small
spot in the sky. When some of that light bounces around in
our atmosphere, especially on an overcast day where it scat- Figure 1-24 The polarization of light.
ters through thick clouds, sunlight takes on a completely
different appearance. Instead of a single, intensely bright
source, the sun effectively becomes a dimmer, spread out The polarization of light occurs naturally in a number of
one. You can block the sun from your eyes by putting a ways:
hand up in front of your face, but consider the scale of the
light source when it’s spread across the entire visible sky on 1. Light emitted from a portion of clear blue sky, at right
a cloudy day. Think of this the next time you’re looking to angles to a line connecting the viewer (or camera) and
do an outdoor portrait shoot—an overcast day might be the sun, is highly polarized. The scattering of the light
more desirable than one with clear skies. rays is caused by very small particles in the atmosphere
The sun also illustrates that a light source’s effective size such as dust, molecules of water vapor and other
is a defining characteristic of light quality. There is no light gases. As the angle decreases, the effect also decreases,
source in our solar system physically larger than the sun (a illustrated in Figure 1-25. A portion of the light blue sky
star with a diameter more than 100 times that of Earth). Yet encountered in typical outdoor scenes is polarized.
it’s distance from us is so great that it’s effective size in the 2. Light becomes polarized when it is reflected from a flat,
sky is actually quite small. The quality of light provided by glossy, non-metallic surface like glass, water and shiny
the sun on a clear day (acting as a point source) falling on plastics. This effect is maximized at an angle whose
a building is comparable to a lamp illuminating a pair of tangent is equal to the refractive index of the reflecting
playing dice on a desk. The effective size of a light source material; this is known as the Brewster angle. For surfaces
considers the size and position of the light relative to the
subject. Bringing the lamp close to the dice dramatically
changes the light quality and renders them just as it would
if the Earth were moved much closer to the sun. Both sub-
jects get considerably warmer, too.

The Polarization of Light

As discussed at the start of this chapter, light appears to

travel in a wave-like motion. The vibrations of these light
waves occur in all directions at right angles to the path of
travel, as illustrated by the light beam to the left of the filter
in Figure 1-24. This is an unpolarized light beam. When light
waves vibrate exclusively in one direction it is polarized. The
light beam on the right of the filter illustrates this condition. Figure 1-25 The polarization of skylight.

23
 s ecti o n 1: Captu r e

property is demonstrated by rotating a second slab of

tourmaline across the direction of polarization as illus-
trated in Figure 1-27. When the second slab is rotated to
a position of 90° from that of the first, no light is trans-
mitted through the second filter. Commercially available
polarizing filters are made from materials whose
chemical composition calls for the grouping of parallel
molecular chains within the filter. This is the nature of
the filters illustrated in Figures 1-24 and 1-27.
Figure 1-26 The polarization of light reflected from a glossy surface.
Polarized light becomes depolarized when it strikes
a scattering medium. This most frequently occurs when
like glass and water, this angle is approximately 55° from the light reflects off of a matte surface, such as a sheet of
the perpendicular (35° from the surface) and is illus- white photo paper. It can also become depolarized upon
trated in Figure 1-26. transmission through a translucent material like a sheet of
3. Light becomes polarized when it is transmitted through acetate.
certain natural crystals or commercially manufactured
polarizing filters. Tourmaline is an example of a dichroic
mineral that polarizes the light it transmits. Dichroic Commonly Encountered Light
means that the material exhibits multiple refractive Sources
indices, causing light to move through it differently
than standard materials. When a beam of unpolarized So far, our discussion of light has centered on methods
light passes through a thin slab of tourmaline, the only of production, its properties and measurement. Pre-
light transmitted is oriented along a single plane. This sumably, you’d like to take a photograph at some point!
Let’s review the light sources most often encountered
by photographers. Broadly speaking, light sources come
in two varieties: artificial or natural. The primary natural
light source is, of course, the sun. Artificial sources can be
subcategorized as continuous or discontinuous in duration
(try not to mix this up with their spectral emission continu-
ity—here we’re describing emission as a function of time).
Sunlight, tungsten bulbs and LEDs are all continuous as
they provide illumination for as long as we may wish to
take a photograph, even if that’s a duration spanning many
minutes. Studio strobes or handheld flash units, on the
other hand, are generally considered to be discontinuous
because they illuminate for brief fractions of a second. Let’s
dive into our light sources in greater detail.

Sunlight
The sun produces enormous amounts of radiation in all
Figure 1-27 The polarizing effect of a tourmaline crystal. directions through space. The radiation that we see travels

24
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

through the vacuum of space, through our planet’s atmos- While you’re thinking about the moon, consider that it’s
phere, onto the surfaces of everything around us. Luckily, got the same spectral power distribution as daylight—it’s
that atmosphere mostly protects us from the forms of a D65 illuminant—albeit at a fraction of the intensity, since
radiation our bodies find harmful while allowing quite a lot moonlight is simply sunlight reflecting off of the moon’s
of visible light radiation through. We often consider the sun surface.9 It’s a moot point for our vision since we don’t per-
as a fairly simple light source: it’s always the same size in the ceive color at extremely low light levels, but it may come in
sky, we learn from an early age to draw it as a yellow circle handy when shooting a nighttime exposure lit exclusively
and we have interactions with it many hours a day (unless by the moon.
you retreat to a dark room with shut curtains in the hopes Back to the sun: its energy distribution outside of Earth’s
of completing a textbook). However, its qualities vary with atmosphere closely approximates that of a blackbody
a complex combination of variables. The time of year, your source at 6500K. A considerable amount of this light is
elevation relative to sea level, air quality, weather conditions lost as it travels through the atmosphere, particularly in
and time of day all change the appearance of sunlight. the blue region of the spectrum. The color temperature
Daylight is composed of at least two different sources of of sunlight combined with the scattered skylight varies
light: direct sunlight, which travels directly through Earth’s between 5000K and 6000K measured from the ground at
atmosphere; and skylight, the light scattered and reflected midday (noon). Daylight white balance presets on cam-
within the atmosphere. Skylight is blue because the small eras are typically tuned to a color temperature of 5500K.
particles in the atmosphere selectively scatter light—espe- Variations in color temperature with the time of day are
cially the short wavelengths—away from the direction illustrated in Figure 1-28, illustrating how a one-size-fits-
of travel of the sunlight. This effect is known as Rayleigh all setting for daylight is not ideal. This is especially true
scattering and it varies inversely with the fourth power of when taking photographs at sunrise or sunset because the
the wavelength. In other words, the scattering at 400 nm is sun’s oblique angle relative to the viewer causes light to
9.4 times greater than at 700 nm. The color temperature of travel through more of the Earth’s atmosphere, illustrated
blue skylight ranges between 10,000 and 20,000K. Addi- in Figure 1-29 (overleaf). During these brief but beauti-
tional light may also reach the subject by reflecting off of ful times of the day, that additional atmosphere in the
objects. Mie scattering is an effect caused by water droplets light’s path scatters much of the blue and green bands,
suspended in clouds that are similarly sized to light wave- rendering warm red and orange hues. Photographers
lengths (indiscriminately across the visible light wavelength refer to this as the golden hour, though its true duration
range) and explains why clouds are white. and probability of occurring is largely dependent on
The amount of direct sunlight is at a maximum when
the sky is clear. This results in 88% of the total light on the
ground, producing a lighting ratio of approximately 8:1.
The ratio decreases as the amount of overcast increases.
Skylight supplies significant “fill” in the sense that shad-
owed areas not receiving direct sunlight are still very well
illuminated. We can appreciate the contribution of skylight
in illuminating our world by observing its absence: the
moon has no atmosphere and thus no skylight, only direct
sunlight illumination. Photographs taken on the moon
show a world with harsh, contrasty light and deep, impene-
trable shadows. Even if it had air to breathe, photographers
would have a considerable challenge making flattering Figure 1-28 The variation in light color temperature with time of day at
glamour portraits on the moon. the surface of the Earth (no cloud cover).

25
 s ecti o n 1: Captu r e

Figure 1-29 The relationship between position of the sun

and the length of the sunlight’s path of travel through Earth’s
atmosphere.

weather and geographic location. Morning light tends to

be less intense and often cooler in color temperature com-
pared to sunsets due to the air’s humidity in transitioning Figure 1-30 Relative spectral power distributions for three
incandescent sources.
from cooler nighttime temperatures.

Tungsten-Filament Lamps
Incandescent lamps emit light when the tungsten-filament
wire reaches a high temperature as the result of its resist-
ance to the passage of electricity. Electricity is converted
to heat which is then converted to light. The filament is
housed in a glass envelope (the bulb) containing an inert
gas such as nitrogen or argon and it evaporates as it heats
up, ultimately settling on the inside surface of the enve-
lope—this is why a dead tungsten bulb appears blackened
or opaque. The upper temperature limit for this type of Figure 1-31 The variation in light output, color temperature
source is 3650K: the melting temperature of tungsten. The and life of a tungsten lamp at different voltages.
color temperature of light emitted by tungsten-filament
lamps ranges from 2700K for a 15-watt household light
bulb to 3200K for photographic studio lamp. Figures 1-30,
1-31 and 1-32 highlight the relationships of spectral-en-
ergy distribution, voltage, color temperature and energy
efficiency.
Tungsten bulbs are waning in popularity due to their
low energy efficiency relative to alternatives like LED bulbs;
they can waste up to 90% of the input energy through
heat emission rather than light emission. Many countries
have already started the phase out of these bulbs in favor
of more efficient modern bulbs for basic infrastructure Figure 1-32 The average relationship between luminous
projects like street lamps. efficiency and color temperature of tungsten-filament lamps.

26
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

Figure 1-33 Typical lamp depreciation for a conventional

tungsten lamp and a tungsten-halogen lamp.

Tungsten-Halogen Lamps
Tungsten-halogen lamps are sometimes called quartz-io-
dine or simply halogen lamps and emit light through
incandescence. They differ from conventional tungsten
lamps in that iodine or bromine gas is added and the bulb
is made from quartz or a similar high-temperature glass.
The iodine combines with the tungsten evaporating from
the filament which, combined with the high temperature
of the filament, produces a decomposition that redeposits
tungsten on the filament. This chemical reaction increases
the life of the filament and prevents the deposit of tung-
sten on the bulb enclosure. Figure 1-34 Spectral power distributions of three different
fluorescent lamps.
The tungsten-halogen design offers high luminance
output consistently over a lifespan that exceeds that
of a traditional tungsten lamp. The filament also burns to re-emit it as light. The color of light emitted by a fluores-
hotter, meaning that its color temperature more closely cent tube depends largely on the mixture of fluorescent
approaches daylight. Figure 1-33 plots the loss in total light materials used in the phosphor coating.
output for conventional tungsten lamps and tungsten-hal- The light that reaches the eye or camera from such a
ogen lamps over time. lamp, therefore, consists of the light given off by these fluo-
rescent compounds plus part of the light from the mercury
vapor that gets through them without being absorbed. The
Fluorescent Lamps result is a continuous spectrum produced by the fluores-
Fluorescent lamps produce light by establishing an cent material superimposed upon the line spectrum of
arc between two electrodes in an atmosphere of very energy produced through the electrical discharge of the
low-pressure mercury vapor contained in a glass tube. mercury vapor. The spectral-energy distributions for three
This low-pressure discharge produces ultraviolet radiation commonly encountered fluorescent lamps are shown in
at specific wavelengths that excite crystals of phosphors Figure 1-34.
lining the wall of the tube. Phosphors such as calcium Fluorescent lamps with a correlated color temperature
tungstate have the ability to absorb ultraviolet energy and of 5000K (such as the General Electric Chroma 50®) are

27
 s ecti o n 1: Captu r e

believed to give a better match to daylight and are now built-in ballasts that produce the necessary electrical
specified in American National Standards Institute (ANSI) current to excite the gas molecules, the result of which
standards for transparency illuminators and viewers. Fluo- is the emission of ultraviolet radiation. The ultraviolet
rescent lamps find widespread use because they generate radiation stimulates the fluorescent material coated on
minimal heat and are less expensive to operate (that is, the inside of the tube, which absorbs and re-emits the
they have higher luminous efficiency) than tungsten lamps. energy as visible light. CFL bulbs are more energy effi-
Additionally, they are low luminance sources due to their cient than tungsten-filament bulbs, produce less heat as
larger surface areas relative to tungsten lamps. The result is a by-product and have longer lifespans. Some CFL bulbs,
that these lamps have less glare and produce more diffuse particularly older ones, take several minutes to achieve
illumination. Fluorescent lamps are nearly always used for full brightness when switched on. This trait is worsened
commercial lighting and, consequently, photographers when operating in colder temperatures. Once looked at
working under such light conditions should be familiar as a potential successor to traditional tungsten, CFLs are
with their characteristics. taking the backseat to LED as the desired light source
The luminous efficiency of fluorescent lamps is generally of the future. Figure 1-35 is a sample spectrum for a CFL
higher than that of tungsten sources, ranging between 40 bulb.
and 60 lumens/W. Additionally, the average life of a fluores-
cent lamp is approximately 5,000 hours, which is nearly five
times as long as that of a conventional tungsten lamp. High-Intensity Discharge Lamps
Similar in operation to fluorescent lamps, high-intensity dis-
charge lamps produce light by passing an arc between two
Compact Fluorescent Lamps electrodes that are only a few inches apart. The electrodes
Compact fluorescent lamps (CFLs) operate similarly to a are located in opposite ends of a small sealed transparent
traditional fluorescent lamp but are designed in form fac- or translucent tube. Also contained within the tube is a
tors that allow it to take the place of tungsten light bulbs. chemical atmosphere of sodium and/or mercury. The arc
CFL bulbs take on a variety of shapes, always consisting of electricity spanning the gap between the electrodes
of a tube containing argon and mercury vapor. They have generates heat and pressure much greater than in fluores-
cent lamps and for this reason these lamps are also referred
to as high-pressure discharge sources. The heat and pressure
1
generated are great enough to vaporize the atoms of
0.9 the various metallic elements contained in the tube. This
0.8
vaporization causes the atoms to emit electromagnetic
energy in the visible region. Since the physical size of
0.7
the tube is small, it allows for the construction of optical
0.6
sources that have excellent beam control. Such sources are
Relative Output

0.5 frequently employed for nighttime illumination of sports

0.4 stadiums, highways, exteriors of buildings and the interiors
of large industrial facilities.
0.3
As in the operation of low-pressure discharge sources,
0.2
high-intensity discharge lamps produce spikes of energy at
0.1
specific wavelengths. These peaks are the result of specific
0 jumps of the electrons within the atomic structure of the
350 400 450 500 550 600 650 700 750
Wavelength (nm) metallic elements. Energy is emitted in peaks located in
Figure 1-35 A sample compact fluorescent lamp bulb spectrum. different positions of the visible spectrum for each element.

28
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

0.9

0.8

0.7
Relative Output

0.6

0.5

0.4

0.3

0.2

0.1

0
350 400 450 500 550 600 650 700 750 800
Wavelength (nm)

Figure 1-36 Spectral power distribution of a high-pressure metal

halide lamp, ConstantColorTM CMH by General Electric. It is a modern,
energy efficient replacement for high-pressure sodium bulbs.

0.9
Figure 1-38 Light output curve of an electronic flash unit.
0.8

0.7
Relative Output

0.6

0.5 time of operation is exceedingly short, with the longest

0.4 times being on the order of 1/500 second and the shortest
0.3 times approaching 1/100,000 second. The time-light curve
0.2 for a typical electronic flash unit is plotted in Figure 1-38.
0.1 The effective flash duration is typically measured between
0
300 350 400 450 500 550 600 650 700 750 800
one-third peak power points and the area contained under
Wavelength (nm) the curve between these limits represents nearly 90% of
Figure 1-37 Spectral power distribution of a high-intensity metal the total light produced.
halide discharge source, the General Electric Multi-VaporTM lamp.
The spectral power distribution of these sources shows
a line spectrum, the exact nature of which is determined
Thus, these lamps do not have a true color temperature, by the type of gas dispersed in the tube. Although the gas
since they are not temperature dependent for the color gives a line spectrum, there are so many lines and they
of light emitted. The SPDs of two commercially available, are so well distributed throughout the visible spectrum
high-intensity discharge sources are shown in Figures 1-36 that no serious error is involved in considering the spec-
and 1-37. trum to be continuous, as shown in Figure 1-39 (overleaf).
The spectrum from the gas in these tubes approaches
incandescence due to the high-current density at the
Electronic Flash time of discharge. The resulting light has a correlated
An electronic flash lamp consists of a glass or quartz tube color temperature of approximately 6000K, which is con-
filled with an inert gas such as xenon with electrodes veniently close to the 5500K used as the typical daylight
placed at either end. When a high-voltage current from the white balance setting on a camera. A color-correction
discharge of a capacitor passes between the electrodes, coating or filter is built-in to the flash unit to ensure an
the gases glow, producing a brilliant flash of light. The total output of 5500K.

29
 s ecti o n 1: Captu r e

from an input voltage through a p-n junction. One semicon-

ductor material called the p-type is the positive carrier and
another semiconductor material, the n-type, is the negative
carrier. Energy flows in one direction from the n-type to
the p-type. The junction that forms between them allows
electrons to change state, giving off energy in the form of
photons. The color of an LED is determined by the semicon-
ductor materials and the amount of energy flowing through
them. The resulting electroluminescent light produces
enough heat to necessitate a heatsink but not so much that
it can be felt by us except in extremely close proximity—per-
Figure 1-39 Relative spectral power distribution of a typical
fect for when a situation calls for lighting with heat-sensitive
xenon-filled electronic flash unit.
subjects like a portrait or an ice cream still life.
A single LED can be very bright, though it is more
The total light output for an electronic flash unit common today to see groups or arrays of LEDs to produce
depends upon the design of the flash lamp and reflector, brighter and larger light sources. LEDs may be two-dimen-
the voltage, and the capacity of the capacitor. In the early sional panels or more unique in shape to create a form
days of electronic flash photography, exposure was con- factor that mimics that of a tungsten-filament bulb (allow-
trolled by setting the f-number as the number obtained by ing for compatibility with existing light fixtures). LED panels
dividing a guide number by the flash-to-subject distance. arrange dozens or even hundreds of diodes in a circular or
Portable electronic flash units today have built-in sensors rectangular shape to create a broad source.
that monitor the light reflected from the subject. Alterna- Since the wavelength emission of LEDs is partially
tively, they may measure the light falling on the imaging dictated by the amount of energy set through the p-n
sensor and quench the flash when sufficient light produces junctions, some light sources offer variable color tem-
the desired exposure. perature capabilities. Some sources are comprised of
Electronic flash units have proven to be an incredibly red, green and blue LEDs that visually combine to create
useful source of light for photographers. Their consistency white light, allowing for tunability to different colors when
in color output, quantity of light output and their extremely desired. Others encase the diode in a material that filters
short duration are all reasons for their widespread use. the light emission to the desired wavelength. Specialized
LED sources exist that emit in the UV and IR portions of the
spectrum for scientific applications.
Light-Emitting Diodes Most LED-based studio lighting is designed for record-
Relatively new to the light source scene is the light-emit- ing video, as studio strobe lighting does not produce
ting diode or LED. An LED is unique in that it does not use continuous duration light and tungsten lights tend to be
phosphors, gases or filament to create light. Instead, it hot and physically large. See Figure 1-40 for one example
produces light through electroluminescence when an elec- of a studio light using an LED array. These lighting solu-
trical current is passed through semiconductor materials. tions have the additional benefit of being portable and
This process also produces heat as a by-product, though power efficient. Primary characteristics to consider for these
much less so than tungsten sources. A heatsink plate is solutions are maximum light output, degree of control over
incorporated into the design of an LED to move heat away color and brightness and form factor or shape.
from the diode and prevent damage. A form of LED technology is used to backlight tel-
The electroluminescence created by an LED happens evision, computer monitor and smartphone screens.
when electrons move within the semiconductor, introduced While this incarnation of LEDs is not frequently used as a

30
 c h a p t e r 1 : L i g h t a n d P h oto m e t r y

tungsten-filament does, though they do get dimmer over

their lifetime. Products on the market claim CRI values
around 80 and 90; this is one characteristic that is hard to
match with tungsten’s typical CRI of 100. LEDs have been in
use for about 50 years but saw rapid adoption as a primary
light source technology in homes and in photo and video
production in the last decade as costs have fallen. Given
the efficiency, considerable lifespan and degree of control
offered by LED bulbs, the future is destined to be lit by
LEDs in their many form factors.

Proof of Concept: Visualizing the

Spectral Output of Light Sources
Figure 1-40 An LED panel for photography and videography.

We’ve explored the numerous ways of characterizing light

in this chapter. The following demonstration showcases a
photographic light source, we do live in a world illumi- unique way in which we observe the spectral emissions of
nated by screens. Chapter 15 details more on LEDs and common artificial light sources. The spectral emission of a
display technologies. light source tells us about the composition of the visible
LED lights have a very long life of up to 50,000 light radiation produced and comparing multiple sources
hours (compared to between 1,000 to 2,000 for a tung- reveals that seemingly similar bulbs can make highly varied
sten bulb) and do not burn out in the same way that light.

Figure 1-41 Five light bulbs photographed through a diffraction grating to reveal their spectral output.

31
 s ecti o n 1: Captu r e

We gathered a set of light bulbs: traditional tungsten, Notes

halogen, compact fluorescent and two LED designs. The 1 Whelan, M., and R. DeLair. History and Types of Speakers, 2013,
specifications listed in Figure 1-41 come from the bulbs’ www.edisontechcenter.org/SodiumLamps.html.
advertised intensity and color temperature. If you walk past 2 Whitehead, Lorne. “Prof. Lorne Whitehead: ‘Improving
the displays in the light bulb aisle of a home improvement the CIE Color Rendering Index – How This Can Be Done
and Why It Matters.’” SCIEN Blog Posts RSS, Stanford
store, you can quickly spot differences in color. This is
University, Feb. 21, 2012, scien.stanford.edu/index.php/
harder to do when viewing a single light source illuminat-
lorne-whitehead/.
ing an environment and the variety of bulb technologies
3 “Position Statement on CRI and Colour Quality Metrics.”
makes it less obvious why it appears the way it does. If we
Position Statement on CRI and Colour Quality Metrics (October
can see the light energy emitted by each bulb in relation 15, 2015) | CIE, International Commission on Illumination,
to the complete visible spectrum, it’s easier to understand Oct. 15, 2015, cie.co.at/publications/position-statement-cri-
why some sources appear “warm” and others “cool,” i.e. and-colour-quality-metrics-october-15-2015.
why they vary in their color temperatures. 4 Wright, Maury. “IES Publishes TM-30 Defining New Color
To make the photograph shown in Figure 1-41, we lined Metrics for Characterizing Lighting.” LEDs, Aug. 19, 2015,
the bulbs up on a table and photographed them in a dark www.ledsmagazine.com/articles/2015/08/ies-publishes-tm-
room. A special optical material called a diffraction grating 30-defining-new-color-metrics-for-characterizing-lighting.
was positioned in front of the camera lens. This material html.
splits white light into its components similar to a prism. The 5 Royer, Michael, and Kevin Houser. “WEBINAR:
result is the visual equivalent of what is shown in a spectral UNDERSTANDING AND APPLYING TM-30-15.” Department
of Energy, Sept. 15, 2015. www.energy.gov/eere/ssl/
power distribution plot—varying energy emission across
webinar-understanding-and-applying-tm-30-15.
the visible spectrum.
6 Davis, Wendy L., and Yoshihiro Ohno. “Development of
Consider the following questions and see if you can
a Color Quality Scale.” NIST, Feb. 6, 2006, www.nist.gov/
answer them based on the image:
publications/development-color-quality-scale-0.
7 “Position Statement on CRI and Colour Quality Metrics.”
1. One of the five lights does not produce a continuous Position Statement on CRI and Colour Quality Metrics (October
spectrum of white light. Which bulb is it? 15, 2015) | CIE, International Commission on Illumination,
2. What visual evidence supports the provided information Oct. 15, 2015, cie.co.at/publications/position-statement-cri-
that one LED bulb is 5000K and another is 3000K? and-colour-quality-metrics-october-15-2015.
3. Which light source exhibits the least red energy 8 Hunter, Fil, et al. Light – Science & Magic: An Introduction to
emission? Does this result in a warm or cool color tem- Photographic Lighting. Focal Press, 2015.
perature appearance? 9 MacEvoy, Bruce. “Light and the Eye.” Handprint : Modern
Color Theory (Concepts), Aug. 1, 2015, handprint.com/HP/
WCL/color1.html.

32
 2 Photographic
Optics

Photograph by Rochester Institute of Technology photography student Erin Percy

If understanding light is the first step toward the photographic image, step two is con-
trolling light to form bright, well-resolved recordings of the world. Controlling light is best
accomplished by directing its path and its behavior as it travels through the camera. The
simplest way to start controlling light is with a small opening—a pinhole—but that’s just
the starting point. Photographic optics provide the technical means to make sharp, detailed
images of our world from many distances, environments, speeds, magnifications and light-
ing conditions. This chapter covers the principles of image formation, camera lenses, depth
of field and considerations for evaluating image quality from an optical imaging system. The
last section brings many of the chapter’s concepts together in an applied shooting exercise
to better appreciate the relationships between lens choice and camera position.

33
 s ecti o n 1: Captu r e

Image Formation with a Pinhole A handheld, light-tight box with a tiny opening at
its front creates a pinhole camera when a light sensitive
The pinhole as an image-forming mechanism played an medium is placed opposite the opening. Figure 2-1 is an
important role in the evolution of the modern camera. example of a homemade pinhole camera. We typically
The observation of images formed by a small opening in think of pinhole cameras using film or photographic paper
an otherwise darkened room goes back at least to Aris- and darkroom chemistry. These materials were cheap and
totle’s time, about 350 B.C.—and the pinhole camera still accessible for many decades. It is increasingly difficult
fascinates photography students because of the simplicity today to find a darkroom within which we can easily load
with which it forms images. The camera obscura, Latin for and process the photographs from these simple cameras.
“darkened room,” evolved into a portable room or tent Pinhole photography lives on in the digital realm with the
large enough to accommodate a person. In fact, any space same mechanical and optical principles applied. Swap-
can be turned into a camera obscura by restricting light ping the light-sensitive film or paper with a digital sensor
exclusively to a small opening in a given space, though replaces the need for chemistry and safe lights while retain-
the image that results can be quite dim depending on ing the simplicity and craft that pinhole imaging allows. We
the scene outside. The light that travels through the small recommend browsing pinhole.org, a community of active
opening projects onto an image plane: the location at photographers and the creators of an annual Worldwide
which an image is formed. The portable room concept Pinhole Photography Day.
shrank to a portable box with a small pinhole opening and We described the behavior of light in terms of the
tracing paper, used as a drawing aid. By 1570, the pinhole corpuscular theory in Chapter 1. Pinholes form images
was replaced by a simple glass lens to produce a brighter, because the corpuscular theory tells us that light travels in
easier to trace image. a straight line and for each point on an object, a reflected
ray of light passing through a pinhole falls on exactly one
spot on the image plane. Because light rays from the top
and bottom of the scene and from the two sides cross
at the pinhole, the image is reversed vertically and hori-
zontally so that objects in a scene appear upside-down
and backward at the location of the sensor as illustrated
in Figure 2-2. Thankfully, our camera viewfinders flip the

Figure 2-1 This homemade pinhole camera uses sheet film Figure 2-2 Reversal of the image vertically and horizontally by
and electronic components to control when and for how long the crossing of the light rays at the pinhole produces an image
the pinhole lets light through. Photograph by Joe Ziolkowski of the letter L that is upside-down and backwards.

34
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

image of our scene to appear as we perceive it—right- a large format view camera (which uses bellows to extend
side-up and forward. Optical viewfinders accomplish this the image distance) results in a telephoto effect. The
with mirrors; electronic viewfinders use software to make angle of view, on the other hand, decreases as the image
the correction. distance increases (see Figure 2-3). A pinhole controls the
A pinhole camera offers two adjustable parameters for amount of light but it does not focus it, so changing the
dictating the character of its image: the pinhole-to-image- image distance has little impact on the apparent sharpness
plane distance and the diameter of the pinhole opening. of the image. That said, critical examination of the image
Image size increases in direct proportion to the pinhole-to- reveals an optimal pinhole-to-image-plane distance for a
image-plane distance, meaning that using a pinhole with pinhole of a given diameter.
Increasing the size of a pinhole beyond the optimal size
allows more light to pass, increasing illuminance at the
image plane and reducing the exposure time required. It
also reduces image sharpness. When the pinhole size is
decreased below the optimal size for the specified pin-
hole-to-image plane distance, diffraction causes a decrease
in sharpness (see Figure 2-4).
Diffraction is the behavior of waves disrupted by an
object or opening; it causes light waves to bend and
spread out. Too little diffraction (like not using a pinhole at
all) and an image fails to resolve at all. Too much diffraction
introduced into an imaging system leads to destructive
interference of the bent light waves in a way that degrades
resolved detail. Thus, a Goldilocks set of pinhole size
Figure 2-3 As the pinhole-to-sensor distance increases, the image and image distance parameters helps produce the best
size increases and the angle of view decreases.

Figure 2-4 Three photographs made with pinholes of different sizes: (A) one-half the optimal size, (B) the optimal size and (C) twice the optimal
size. These represent small sections cropped from the full-sized photographs.

35
 s ecti o n 1: Captu r e

possible image. The optimal pinhole size is calculated with The speed of light travelling in a vacuum is 186,000
the formula: miles per second. Other than photographing stars from a
space telescope, we’re concerned with light as it travels
√
D= f (Eq. 2.1) through air and other materials. When it does, its speed
141
decreases. All photographic lenses rely on a material’s abil-
where D is the diameter of the pinhole in inches and f is the ity to bend light rays as they travel from subject to sensor.
pinhole-to-image plane distance in inches (see Table 2-1). Refraction is the behavior of light rays changing direction,
When photographing relatively close subjects, the value or bending, due to a speed change as they move through
of f is found by substituting the object distance (u) and the from one medium to another. Snell’s Law of Refraction is
image distance (v) in the formula 1/f= 1/u+ 1/v. Note that used to calculate how much a medium changes the direc-
the formula changes when measuring pinhole diameter, D, tion of light rays depending on their angle of incidence.
in millimeters: Any material that transmits some amount of light—glass,
plastic, water—has an index of refraction that indicates how
√
D= f (Eq. 2.2) quickly light travels relative to its behavior in air. It’s this
28
change from one medium to another, say from air to glass
To take one example, using an image distance of 8 and back to air again, that introduces the path changes of
inches dictates that the optimal pinhole diameter is about light rays.
1/50 inch. A No.10 sewing needle produces a pinhole of
approximately this size, a helpful fact for photographers Speed of light
in a vacuum c
looking to make use of their sewing kits. Refractive index (n) = = (Eq. 2.3)
Speed of light v
in a medium
Table 2-1 Optimal pinhole diameters for different pinhole-to-image
plane distances. We approximate air’s refractive index as 1.0, not far off
f (Distance) D (Diameter) F-number
from that of a vacuum where light does not bend or slow
at all. Any material that refracts light more than air has a
1 in. 1/140 in. f/140
refractive index greater than 1. Index of refraction is unitless
2 in. 1/100 in. f/200
because it’s a ratio of the speed of light in a vacuum rela-
4 in. 1/70 in. f/280 tive to its speed in a medium. Wavelength-independent
8 in. 1/50 in. f/400 refraction is called dispersion, mentioned in Chapter 1 to
16 in. 1/35 in. f/560 explain why a glass prism creates a rainbow of colors out of
white incident light.
Optical materials like glass and plastic have refractive
indices higher than air. Light traveling through them bends
Image Formation with a Lens toward the perpendicular or normal as illustrated in Figure
2-5. We’ve all seen an example of refraction when looking
Making images with pinholes is a great way to observe the at a straw in a glass of water: it appears broken or in two
principles of image formation and the parameters that con- places at the same time, as part of the straw is submerged
trol its behavior. Pinholes are limiting, however, because in the liquid while the rest is surrounded by air. This is the
they do not efficiently collect and focus light to the degree same phenomenon that makes it challenging to spear a
that we have come to expect from our image-making tools. fish in a stream (a traditional rite of passage for all pho-
The rest of this chapter focuses on lenses as a consequence tographers). Our eyes perceive the bent light rays and
of this limitation. incorrectly estimate the actual position of the fish where it
exists below the water’s surface.

36
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-6 Three types of positive lenses: double convex (left), plano-
convex (center) and positive meniscus (right).

Figure 2-5 Light travelling through a medium at an angle

with a refractive index greater than 1.0 bends toward the
perpendicular.

The refractive index of glass is 1.5, sometimes higher for Figure 2-7 The lens axis is a straight line through the centers of
curvature of the lens surfaces.
particular types of glass. This makes it a great candidate for
bending light to form images, which is exactly why we use
glass for photographic lenses. Certain plastics are also used
in smartphone camera system optics in the interest of size, one of the surfaces is flat, the lens axis is a straight line
cost and weight. Water’s refractive index is 1.333, making it through the one center of curvature and perpendicular to
a viable but less effective candidate for image formation. the flat surface.
The shape of a lens controls how it bends light to form A positive lens produces an image by refracting or
an image. A positive lens is easily recognized as thicker at bending light such that all of the light rays from an object
the center than the edges. It must have one convex sur- point falling on its front surface converge to a point behind
face. The other surface can be convex (double convex), flat the lens (see Figure 2-8, overleaf). If the object point is at
(plano-convex) or concave (positive meniscus) as illustrated infinity or a large distance, the light rays enter the lens trav-
in Figure 2-6. The curved surfaces of positive lenses are eling parallel to each other. The location where they come
typically spherical like the outside (convex) or inside (con- into focus is the principal focal point. We determine the
cave) surface of a hollow ball. If the curvature of a spherical focal length of the lens by measuring the distance from the
lens surface is extended to produce a complete sphere, principal focal point to the back nodal point—roughly the
the center of that sphere is identified as the center of middle of a single-element lens (see Figure 2-9, overleaf).
curvature of the lens surface. A straight line drawn through Reversing the direction of light through the lens with a dis-
the two centers of curvature of the two lens surfaces is tant object on the right produces a second principal focal
defined as the lens axis or optical axis (see Figure 2-7). If point and a second focal length to the left of the lens, as

37
 s ecti o n 1: Captu r e

Figure 2-8 Image formation with a positive lens.

Figure 2-11 Image formation with a negative lens.

Let’s pause here because most of us grow up thinking

that images, by definition, are the things we see with our
own eyes. Though accurate, this is an incomplete under-
standing of image formation and optics. When we talk
about a positive lens forming a real image, it’s an image
Figure 2-9 The back focal length is the distance between the back that can be captured by a sensor or projected onto a
nodal point and the image of an infinitely distant object.
surface where we can see it for ourselves. Positive lenses
are converging lenses because they bring light rays together
to form the real image we see. Their counterpart, negative
lenses, are diverging lenses because they never allow light
rays to focus at a point in front of them. Instead, a negative
lens forms an image behind itself—we call this a virtual
image because it cannot be captured or projected.
The optical center of a lens is a point on the lens axis
where an off-axis, undeviated ray of light crosses the lens
Figure 2-10 The front focal length is found by reversing the direction
of light through the lens.

well as a second nodal point (see Figure 2-10). The two sets
of terms are distinguished using the designations of object
(or front) and image (or back). For example: front principal
focal point and back principal focal point.
A negative lens cannot form an image alone and is easily
identified as thinner in the center than at its edges. Light
rays entering a negative lens travelling parallel to each
other diverge and never form an image. Its focal length is
determined by extending the diverging rays back toward
the light source where it forms a virtual image (see Figure
2-11). Note that light rays traveling through the center of Figure 2-12 All light rays passing through the optical center leave
the lens traveling parallel to the direction of entry. Nodal points are
a lens (negative or positive) are undeviated and continue identified by extending the entering and departing portions of an
behind the lens the same as they entered it. undeviated ray of light in straight lines until they intersect the lens axis.

38
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

axis. All rays of light that pass through the optical center are
undeviated—that is, they leave the lens traveling parallel to
the direction of entry (see Figure 2-12). Object distance and
image distance are measured to the optical center when
great precision is not needed. When precision is required,
the object distance is measured from the object to the
front (or object) nodal point and the image distance is
measured to the back (or image) nodal point.
We locate the front nodal point on a lens diagram by
extending an entering, undeviated ray of light in a straight Figure 2-13 A simple nodal slide. The line on the side identifies the
line until it intersects the lens axis. We locate the back location of the pivot point.

nodal point by extending the departing ray of light back-

ward in a straight line until it intersects the lens axis.
Manufacturers do not mark nodal points on lenses,
however, we can find their locations with a simple experi-
ment. We first place the lens in a nodal slide, a device that
allows for pivoting the lens at various positions along its
axis. A simple trough with a pivot on the bottom like the
one in Figure 2-13 works for crude measurements. Profes-
sional optical benches like the one shown in Figure 2-14
are used when greater precision is required. We focus the
lens on a distant light source with the front of the lens
facing the light. Then we pivot the lens slightly from side to
side while observing the image it forms on a flat plane (a
board or piece of paper works well). If the image does not
Figure 2-14 A professional optical bench permits microscopic
move as the lens pivots, this pivot point is the back (image) examination of the aerial image formed by a lens.
nodal point. If the image does move, however, we have
to move the lens toward or away from the light source on
the nodal slide, focus the image on the paper by moving the front nodal point and the image distance from the
the whole nodal slide and try pivoting it again. We repeat image to the back nodal point. Accurate determination
this process until the image remains stationary when the of the focal length requires measuring the distance from
lens is pivoted. The front (object) nodal point is determined the sharp image of a distant object to the back nodal
the same way with the lens is reversed so that the back point. Measurements to the nodal points are also made
of the lens is facing the light source. Knowing the nodal when using lens formulas to determine image size and
point locations on lenses is particularly helpful with certain scale of reproduction (see Figure 2-15, overleaf). With
applications requiring optical precision. If it were up to us, conventional lenses that are relatively thin, little error is
lenses would be labeled with this information directly from introduced by measuring to the physical center of the
the manufacturer. lens rather than to the appropriate nodal point. With
Here are three practical applications of knowing the thick lenses and some lenses of special design—such as
nodal point locations: telephoto, reversed telephoto and variable focal length
lenses—considerable error is introduced by measuring
1. When accuracy is required, such as during lens design, to the center of the lens. Here, the nodal points can
the object distance is measured from the object to be some distance from the physical center of the lens

39
 s ecti o n 1: Captu r e

Figure 2-15 For an object point on the lens axis, object distance is
measured to the front nodal point and image distance is measured
from the image to the back nodal point.

Figure 2-16 For an object point off the lens axis, object and image
distances are measured in a direction that is parallel to the lens axis, to
the corresponding nodal planes.

and may even be in front of or behind the lens. When

distances are measured for objects or images that are
not on the lens axis, the correct procedure is to measure
the distance parallel to the lens axis to the appropriate
nodal plane rather than to the nodal point. The nodal
plane is a plane that is perpendicular to the lens axis and Figure 2-17 A tripod mount designed to place the point of
that includes the nodal point (see Figure 2-16). rotation at the back nodal point of the lens. The nodal points of the
2. Panoramic photography is easier with more seamless photographer’s favorite lenses are marked with tape.

and easily stitched results if the camera is rotated on

the back nodal point. Special tripod mounts like the one the lens in the drawing, regardless of the number of lens
shown in Figure 2-17 make this a practical and straight- elements. We adhere to this procedure in the following
forward technique. The setup allows for taking multiple section.
frames as the camera is rotated that are later blended
and stitched together.
3. Making lens diagrams to illustrate image formation is Lens Focal Length
greatly simplified using nodal points and planes rather
than lens elements. Whereas lens designers must con- Let’s step back and speak practically about photographic
sider the effect that each surface of each element in a optics. We believe it’s important to know how an image is
lens has on a large number of light rays, many problems formed and why, but a photographer just starting out is
involving image and object sizes and distances are illus- not going to wrestle with the details—rather, you’re going
trated and solved using the nodal planes to represent to pick up a camera with a lens attached and start shooting.

40
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

The key concept to understand from the moment you put Lens focal length is commonly described in millimeters.
the camera’s viewfinder up to your eye is lens focal length. Figure 2-18 illustrates a range of focal length lenses for a
We’ve defined focal length as the distance from the prin- full frame camera and the corresponding field of view of
cipal focal point to the back nodal point of the lens. It’s a each. Note that all of these are actually compound lenses in
measure of the distance it takes the optics to bend light to that they consist of many separate lens elements manu-
a point of focus at the image plane. The problem with this factured and arranged to achieve the desired focal lengths
definition is that you won’t find principle focal point or back with high image quality. Focal length and field of view
nodal point labeled on the lens barrel. Instead, photogra- have an inversely proportional relationship: the longer the
pher’s associate lens focal lengths with their resulting field focal length, the smaller the field of view.
of view, the amount of a scene that is projected to fit onto It’s tempting to take a mental shortcut and memorize a
the image sensor. Focal length also dictates subject mag- given focal length in millimeters with a corresponding field
nification as it appears on the sensor and perspective (the of view. However, there is no single measurement for each
relative sizes of objects in a scene). In other words, pick a category of lens: the focal length that achieves a desired
lens based on how much of the scene you wish to capture, field of view is different depending on the sensor size.
the distance from which you can capture it, how subjects Lenses of different focal lengths are required to achieve
at different distances appear relative to each other, or some equivalent fields of view between smartphone camera sen-
combination of the three. The right lens choice won’t leave sors, crop sensors, full frame sensors and medium format
you surprised or frustrated when putting an eye up to the sensors.
viewfinder. Focal length varies based on the degree or amount that
the lens bends light, a function of light refraction. Short
focal length lenses bend light very sharply, bringing light
to a point of focus just a short distance away from the back
of the lens. Long focal length lenses, in contrast, bend the
light much less and over a longer distance, bringing light
to a point of focus farther from the back of the lens. Longer
focal length lenses always necessitate additional physical
distances between the front lens elements and the sensor,
which is part of the reason that portable camera options
shy away from long focal lengths.
Professional photographers gravitate toward camera
equipment featuring the ability to change lenses, com-
monly known as interchangeable lens camera systems.
Swapping out the optics provides creative and functional
flexibility with regard to image formation and is much
cheaper than buying a set of cameras with unique focal
length lenses on each one. That said, we doubt you’ll fail
to capture a fleeting, special moment simply because you
had an 80mm lens instead of an 85mm. In this way, small
value differences in focal lengths are not nearly as impor-
tant as their general type: photographers work with normal,
wide-angle, telephoto and variable focal length (zoom) lenses
Figure 2-18 A range of focal lengths and the fields of view that result
when using a full frame sensor. Field of view decreases as focal length
and tend to group them as such. We’ll step through each
increases. category and identify their particular qualities. The numbers

41
 s ecti o n 1: Captu r e

we use apply to full frame cameras, so a wide-angle lens measurements and calculation. First, we determine the
described later does not function as such on a medium diagonal dimension of the rectangular sensor. A stand-
format sensor. ard, full frame sensor measures 24mm by 36mm and the
Pythagorean theorem tells us:

Normal Focal Length a2 + b2 = c2 (Eq. 2.4)

The go-to standard focal length on any camera system
is one that achieves an angle of coverage, or field of view, or
approximating 55°. Where did we come up with this
number? Studies show that the functional field of view of c = √ a2 + b2 (2.4)
the human eye is 55°. We use the designation of normal
because photographs taken with this field of view appear c = √ 24mm2 + 36mm2 (2.4)
natural and in-line with our perceptual experience of the
world. That doesn’t necessarily make it the best or most
√ 1,872 = 43.3mm (2.4)
effective focal length for a given photograph, it’s simply
helpful to establish as a baseline from which to compare.
Anything wider than normal renders a scene with a greater The full frame sensor diagonal dimension is 43.3mm. With
field of view than we experience ourselves while anything this number in-hand, we calculate a normal focal length
narrower than normal renders a scene with a lesser field of providing a desired field of view of 55°:
view.
If 55° doesn’t seem like very much for our vision, keep sensor diagonal
Focal length = (Eq. 2.5)
in mind that our eyes are different from a camera capture
2 tan ( field of2 view (
in one important dimension: time. Our eyes can change
shape, iris dilation and position many times in just frac-
tions of seconds, with the perceived visual data being 43.3
Focal length = (2.5)
continually processed by the brain. A still image is locked 2 tan ( 552 (
into one combination of lens focal length, exposure time,
aperture setting and focus distance. Additionally, the
Focal length = 41.6mm (2.5)
camera contains only a single sensor whereas we have two
eyes and two light-sensing retinas. We do perceive the
world beyond the 55° field of view range, just not with the Our target focal length is 41.6mm to achieve the
same level of visual acuity or attention and not all in one standard 55° field of view. With some rounding out of
singular instance in time. Field of view can be measured convenience, 50mm or 55mm are traditional normal focal
horizontally, vertically or diagonally—having a single value lengths, though the range from 35mm to 85mm is broadly
instead of three makes our lives easier. In light of all of this, considered to qualify as well.
we specify that 55° is equivalent to the eye’s visual cone Alternatively, algebraically rearranging the previous
of attention.1 You can sense someone approaching out of equation allows us to calculate the field of view if sensor
the “corner of your eye” (an expression, obviously, as our size and lens focal length are known:
spherical eyeballs don’t have corners), though it’s more of
a light-difference sensation than full visual acuity. More on Field of View = 2 tan-1 ( sensor
2 x focal length (
diagonal
(Eq. 2.6)
human vision and its role in photography in Chapter 13.
Determining what focal length provides a field of
view equivalent to 55° on a given sensor requires some

42
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Normal focal length lenses are invaluable and often lens considered against the same subject photographed
inexpensive compared to other lens types that require using telephoto and wide-angle lenses.
more complex optical engineering or more material. Figure In the past, it was safe to assume that a typical consumer
2-19 shows a photograph taken with a normal focal length camera with a non-interchangeable lens used a focal
length of approximately 35mm, producing a slightly wide
field of view, or a zoom lens that included this focal length
in its zoom range. However, phone cameras in particular
ushered in an era of wider focal length lenses based on
how the average user photographed: with a particular
interest in capturing a wide field of view. We continue to
define a normal focal length as equivalent to 50mm on a
full frame system because it offers minimal distortion, rep-
resents our first person visual experience closely and serves
as a convenient reference point for all focal lengths wider
or longer. Having said that, it may very well be true that the
majority of cameras out in the world shoot with wide-angle
lenses.
This is a good time to point out a common oversimpli-
fication regarding focal length. Stand in one place, take a
photograph with a 50mm lens, swap it for a 25mm lens
and take another photograph. You might conclude that the
lenses always change subject magnification. By only chang-
ing focal length, you’ve made two photographs of a subject
rendered at different sizes on the sensor. Repeat this exper-
iment while adding a second variable: distance. By moving
closer to the subject and taking that second photograph
at 25mm, you achieve the same subject magnification as
the 50mm at the original distance. This shows us that focal
length dictates working distance, the distance between the
front of the lens and the subject photographed. Both 50mm
and 25mm lenses can achieve the same subject magnifica-
tion; they differ in the working distance required to achieve
it. Longer focal length lenses mean photographers can be
a considerable distance from the subject (a critical detail
for venomous spider photographers). Shorter focal length
lenses mean photographers can get right in the middle
of the action (a boon for puppy photographers). Some-
times you’ll choose focal length not because of its field of
view but rather for its ability to solve a problem in working
Figure 2-19 A scene photographed with a full frame SLR using distance, either when you can’t get closer or when you
20mm, 50mm and 200mm focal length lenses, respectively. The specifically can’t get away. The secondary effect of changing
photographer stayed in the same location for all three. Photographs by
Rochester Institute of Technology photography student Serena Nappa
working distance is changes in perspective, which we’ll talk
about a little later in this chapter.

43
 s ecti o n 1: Captu r e

Telephoto Focal Lengths locate the position of the back nodal plane in a ray tracing
Telephoto lenses offer increased subject magnification and of a telephoto lens focused on a distant object point, as in
decreased field of view when object distance is held con- Figure 2-20 (top), by reversing the converging rays of light
stant. They also tend to compress perspective. Telephoto in straight lines back through the lens until they meet the
focal lengths are anything above the normal 50mm on a entering parallel rays of light. To determine the position of
full frame sensor. There’s overlap in each of these catego- the back nodal point with a telephoto lens, we pivot the
ries, though, so most photographers don’t consider a lens lens in various positions along the lens axis on a nodal slide
to be a telephoto until it’s 85mm or greater. The overlap is until the image of a distant object remains stationary. Alter-
also due to the fact that true telephoto lens designs have natively, the lens and camera can be focused on infinity,
unique lens element arrangements. Photographers are whereby the back nodal point is located exactly one focal
not usually concerned with the actual design and instead length in front of the image plane. The position of the back
simplify telephoto as a function of focal length. Generally nodal plane is noted in relation to the front edge of the
speaking, 80mm to 300mm is described as telephoto and lens barrel, since this relationship remains constant.
anything greater is a super telephoto focal length.
The basic design for a telephoto lens consists of a
positive element in front of (but separate from) a negative Wide-Angle Focal Lengths
element as illustrated in Figure 2-20. When a telephoto Wide-angle lenses are broadly defined as any focal length
lens and a normal-type lens of the same focal length are below than the normal focal length. Again, though, there’s
focused on a distant object point, both images come to a wiggle room in these categories and photographers
focus one focal length behind the respective back (image) typically consider lenses wider than 35mm to be truly
nodal planes. However, the lens-to-image distance is wide-angle. These focal lengths offer decreased magnifi-
smaller with the telephoto lens. This reduced lens-to-im- cation and increased field of view at a constant working
age distance is due to the fact that the back nodal plane distance relative to a normal lens. They can exhibit exag-
is located in front of the lens with telephoto lenses rather gerated perspective, sometimes to the point of distortion.
than near the center, as with normal lenses. It is easy to Ultra wide-angle lenses like 8mm or 12mm on a full frame

Figure 2-20 The lens-to-sensor distance is shorter for a telephoto Figure 2-21 The covering power of a wide-angle lens compared
lens (top) than for a normal-type lens (bottom) of the same focal to that of a normal-type lens of the same focal length. The images
length. formed by the two lenses are the same size.

44
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

sensor are called fisheye because of their characteristic compromise. Their designs allow the photographer to
bulging, circular image distortion. Fisheye lenses can change the distances between lens elements inside the
achieve a field of view approaching 180° by foregoing the lens barrel to achieve different focal lengths. Focal length
formation of a rectilinear image and instead rendering is altered continuously while the image remains in focus.
straight lines as curved, particularly near the image frame The relationship is illustrated with a simple telephoto lens
edges. where the distance between the positive and negative ele-
A wide-angle lens is optically defined as a lens having ments is varied, as illustrated in Figure 2-22. This change in
an angular covering power significantly larger than the 55° position of the negative element changes the focal length
angle of coverage provided by a normal-type lens, or as (as well as image size and angle of view), but the image
having a circle of good definition with a diameter consider- does not remain in focus. Other problems include aberra-
ably larger than the focal length when focused on infinity tion correction and keeping the relative aperture constant
(see Figure 2-21). Most modern wide-angle lenses have a at all focal length settings.
considerable number of elements and they generally pro- Two methods are used for the movement of elements in
duce good definition even at the maximum aperture, with zoom lenses. One method is to link the moveable lens ele-
much less fall-off of illumination toward the corners than ments so that they move the same distance. This is called
the earlier lenses. optical compensation. The mechanical movement is simple;
the optical design is complex and requires more elements.
The other method is mechanical compensation and involves
Variable Focal Length moving multiple elements by different amounts and
Having the right lens at the right moment is both logis- requires a complex mechanical design. For the problem
tically challenging and potentially expensive (there are of maintaining a constant f-number as the focal length
a lot of lenses to own and then you have to carry them is changed, an optical solution is incorporating an afocal
all). An ideal solution is to have a single, versatile lens that attachment at or near the front of the lens so that the aper-
adjusts to multiple focal lengths. Variable focal length ture diameter and the distance between the diaphragm
lenses offer us this functionality, though not without some and the sensor remains fixed. An alternative mechanical
solution adjusts the size of the diaphragm opening as the
focal length is changed.
There are also mechanical and optical methods for
keeping the image in focus. The mechanical method
consists of changing the lens-to-sensor distance, as with
conventional lenses. The optical method involves using a
positive element in front of the afocal component that is
used to keep the relative aperture constant.
The primary disadvantage of variable focal length lenses
concerns their ability to resolve sharp images. Compared to
fixed focal length lenses (also called prime lenses), variable
focal length lenses are less optimized for sharpness at any
one setting. If you find that you frequently photograph
at 85mm, invest in an 85mm prime lens; an 85mm prime
produces a sharper photograph than a variable 35–135mm
zoom lens set to 85mm. Additionally, zoom lenses have
Figure 2-22 Changing the distance between the positive and negative
elements of a telephoto lens changes the focal length.
smaller maximum apertures (the impact of which is
described later on in this chapter).

45
 s ecti o n 1: Captu r e

The Relationship Between Focal included as a control. Two basic concepts important to an
Length and Perspective understanding of the control of linear perspective are:

Understanding how to control perspective in photogra- 1. Image size is directly proportional to focal length.
phy is essential yet often misunderstood. Perspective is 2. Image size is inversely proportional to object distance.
the perception of space and the relationships of objects
in that space, including scale, distance and size. We’ve
used the term in this chapter thus far to categorize lens Focal Length and Image Size
focal lengths. Perspective is also influenced by depth of Photographers like to have more than one focal length
field, lighting, subject overlap, color and aerial haze. Our lens available to control image size and the corresponding
brain uses details like shadows and color saturation to angle of view. Let’s examine our first concept, image size
get a sense of objects in three-dimensional space that is directly proportional to focal length, using the following
photographers can exploit to further their visual goals. example. We photograph a building that reproduces ¼
Stereoscopic vision, thanks to our pair of eyes, gives inch high on our sensor when using a 50mm focal length
us additional information about depth and distance; lens on a full frame camera. Now we switch to a 100mm
traditional, single-lens camera systems do not have stereo- focal length lens while keeping the camera in the same
scopic information. position. The building is now ½ inch high on the sensor.
Linear perspective in an image only changes when the The relationship between image size and focal length is a
camera location changes relative to your subject. Con- direct one: doubling the focal length doubles the image
sider the example of taking two photographs, framing size. It also doubles everything else in the scene, though
the subject similarly, first with a telephoto and second it does not change the ratio of image sizes for objects at
with a wide-angle lens. The perspective in the resulting different distances (see Figure 2-23).
two photographs is different. Many people assume that This relationship holds true with the image distance
the perspective change is due to the lens focal length approximately equal to the focal length of the lens, which
change because it is the obvious variable in the scenario. is the case when photographing objects that are at a
However, the change in perspective is actually due to the moderate to large distance from the camera. Photograph-
difference in distance between the camera and subject. ing objects that are close to the camera causes the image
To achieve similar framing, the two focal length lenses distance to be larger than the focal length, causing this
require the photographer to move closer or farther away relationship to fail.
to achieve similar framing. While the lens change is an
important part of the story, the true source of the per-
spective change is the distance from the photographer to Object Distance and Image Size
the subject. Consider an example of placing two objects of equal size at
Changes in linear perspective happen when there are distances of 1 foot and 2 feet from a camera lens. This pro-
changes in camera position and the focal length of camera duces images that vary in size in a 2:1 ratio with the object
lenses. If the definition of linear perspective were limited at 1 foot appearing twice as large as the object at 2 feet.
to the relative image size of objects at different distances Actual image sizes can be determined given the object
(or the angle of convergence of parallel subject lines), the size, focal length and object distances with either graphical
photographer’s control would be limited to the choice drawings or lens formulas. For now, we are only concerned
of camera position (object distance). If changes in the with relative image sizes.
appearance of linear perspective were included even when Linear perspective is based on the ratio of image sizes
there is no change in the relative image sizes of objects at for objects at different distances. Figure 2-24 shows two
different distances, then lens focal length would have to be objects of equal size at distance ratios of 1:2, 1:3 and 1:4.

46
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-23 Photographs made from the same position with 135mm (left), 55mm (middle) and 28mm (right) focal length lenses. Image size
changes in proportion to focal length. Relative sizes for objects at different distances remain constant.

The sizes of the resulting images are in ratios of 2:1, 3:1 and
4:1. If we have movable objects, the relative image sizes
and linear perspective are easily modified by changing the
object positions within the scene. Many times, however, we
photograph objects that cannot be moved and therefore
the options available to control perspective or alter image
size are either moving the camera or changing the lens
focal length.
Consider two objects at a distance ratio of 1:2 from the
camera (see Figure 2-25, overleaf). The ratio of the image
sizes is 1:2. Next, we move the camera away to double the
distance from the closer object. This does not double the
distance to the farther object. Therefore, the ratio of the
image sizes does not remain the same. The ratio of object
distances changes from 1:2 to 2:3 by moving the camera;
the ratio of image sizes changes from 2:1 to 3:2 (or 1.5:1).
Moving the camera farther away from the subjects reduces
Figure 2-24 Image size is inversely proportional to object distance. the size of both objects in the image and also makes them
The ratios of the object distances from top to bottom are 2:1, 3:1
and 4:1.
appear nearly equal in size. The two images will never be
equal in size no matter how far away we move the camera.
However, with very large object distances, the differences
in size becomes insignificant.

47
 s ecti o n 1: Captu r e

The linear perspective produced by moving the camera

farther from the objects is referred to as a weaker perspec-
tive than that produced with the camera in the original
position. Thus, weak perspective is an attribute in a picture
in which image size decreases more slowly with increasing
object distance. With weak perspective, space appears
compressed as though there were less distance between
nearer and farther objects than actually exists (see Figure
2-26).
Conversely, moving a camera closer to the two objects
increases the image size of the nearer object more rapidly
than that of the farther object, producing a stronger
Figure 2-25 Doubling the distance from the camera to the perspective. Strong perspective is especially flattering
near object changes the ratio of distances to the two objects to architectural photographs of small rooms because
from 1:2 to 2:3.
it makes the rooms appear more spacious. All of these
examples teach us that linear perspective is dictated by
the camera-to-subject distance, not focal length of the
lens.

Focal Length and Object Distance

Photographers commonly change focal length and object
distance simultaneously to control linear perspective and
overall image size. If the perspective appears too strong
and unflattering in a portrait shoot when taken with a
normal focal length lens, the photographer can substitute a
longer focal length lens and move the camera farther from
the subject to obtain about the same size image but with
a weaker and more flattering perspective. Because short
focal length, wide-angle lenses tend to be used with the
camera close to the subject and telephoto lenses are often
used with the camera at relatively far distances, strong
perspective is associated with wide-angle lenses and
weak perspective is conversely associated with telephoto
lenses. However, it is the camera position and not the focal
length or type of lens that produces the abnormal linear
perspective.
The change in linear perspective with a change in object
distance is seen most vividly when a corresponding change
Figure 2-26 Space appears compressed in the top photograph, is made in the focal length to keep an important part of the
made with a 150mm lens, compared to the bottom photograph
made with a 50mm lens from a closer position.
scene the same size. In Figure 2-27, the focal length of the
lenses and the camera positions were adjusted to keep the
images of the nearer object the same size. The difference in

48
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-27 Photographs made with a 50mm focal

length lens (top) and 135mm focal length lens
(bottom) with the camera moved closer to match the
image size. The change in the relative sizes between
the subject and background elements is due to the
camera position change. Photographs by Rochester
Institute of Technology photography students Daniel
Bacon and Jesse Wolfe

linear perspective is revealed by the difference in size of the Drawing Lens Diagrams
images of the farther object.
In situations where a certain linear perspective con- Lens diagrams of complex lens systems become quite clut-
tributes significantly to the photograph’s effectiveness, tered and difficult to follow. Luckily, they can be simplified
the correct procedure is to first select the camera position and serve as a helpful alternative to mathematical formulas.
that produces the desired perspective and then select the In an effort to simplify the drawings, we employ the thin
focal length lens that produces the desired image size. A lens treatment: the two nodal planes of a lens element are
zoom lens offers the advantage of providing a range of considered close enough to each other that they can be
focal lengths. With fixed focal length lenses, if the desired combined into a single plane without significant loss of
focal length is not available and changing the camera accuracy. The two drawings in Figure 2-28 (overleaf) show a
position reduced the effectiveness because of the change comparison between using lens elements and using nodal
in perspective, an alternative is to use the next shorter focal planes. Provided here are the instructions for creating a
length lens available and enlarge and crop the resulting scaled-down lens diagram using a single nodal plane like
photograph. the one in Figure 2-29 (overleaf).

49
 s ecti o n 1: Captu r e

1. Draw a straight, horizontal line to represent the lens axis.

2. Draw a line perpendicular to lens axis to represent the
nodal plane of the lens.
3. Place marks on the lens axis one focal length in front
and behind the lens. In this example, the focal length is
1 inch. These two marks represent the front and back
principal focal points.
4. Draw the object at the correct distance from the lens.
In this example, the object is 1 inch tall and located 2
inches from the lens.
5. Draw the first ray of light from the top of the object
straight through the optical center of the lens (that is,
the intersection of the lens axis and the nodal plane).
Figure 2-28 Lens diagrams showing image formation using the lens
6. Draw the second ray parallel to the lens axis, to the
elements (top) and the simpler thin-lens procedure (bottom).
nodal plane, then through the back principal focal
point.
7. Draw the third ray through the front principal focal point
to the nodal plane, then parallel to the lens axis.
8. The intersection of the three rays represents the image
of the top of the object. Draw a vertical line from that
intersection to the lens axis to represent the entire
(inverted) image of the object.

We determine the correct size and position of the image

from the original drawing using a ruler. The image size is 1
inch and the image distance is 2 inches. From this we can
generalize that placing an object two focal lengths in front
of any lens produces a same-sized image two focal lengths
behind the lens.
Changing the distance between the object and the lens
produces a change in the image formation location. The

Figure 2-29 Making an actual-size or scale lens diagram in eight

steps.

Figure 2-30 Moving an object closer to the lens results in an increase

in both image distance and image size.

50
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

relationship is an inverse one, so as the object distance two nodal planes, separated by the appropriate actual or
decreases, the image distance increases. Since the two scale distance. The three rays of light are drawn from an
distances are interdependent and interchangeable, they object point to the front nodal plane, as before, but they
are called conjugate distances. Image size also changes as are drawn parallel to the lens axis between the two nodal
the object and image distances change. Moving an object planes before they converge to form the image (see Figure
closer to the lens results in an increase in both the image 2-31).
distance and the image size. These relationships are illus- It’s not just image and object distances that we derived
trated in Figure 2-30. from lens diagrams. Angle of view, a measure of how much
The closest distance that an object can be placed in of the scene is recorded on the sensor, is also determined
front of a lens that yields a real image is, theoretically, through these drawings. Angle of view is typically meas-
slightly more than one focal length. Placing an object at ured for the diagonal dimension of the sensor, although
exactly one focal length from the lens causes the light rays two angles of view are sometimes specified: one for the
from an object point to leave the lens traveling parallel sensor’s vertical dimension and one for the horizontal.
to each other. As a result, we think of an image only as Angle of view is determined with the sensor placed one
forming at infinity. In practice, the closest an object can be focal length behind the lens which corresponds to focus-
placed in front of a lens while still obtaining a sharp image ing on infinity. We first draw a horizontal line for the lens
is determined by the maximum distance that the sensor axis and a vertical line to represent the nodal planes, just
can be placed from the lens. Problems like these are solved as we did with the thin-lens treatment. Next, we draw a
with lens diagrams. If the maximum image distance is 3 second vertical line one focal length (actual or scale dis-
inches for a camera equipped with a 2 inch focal length tance) to the right of the nodal planes. The second vertical
lens, an actual-size or scale drawing is made starting with line represents the sensor diagonal, so it must be the
the image located 3 inches to the right of the lens. Three correct (actual or scale) length.
rays of light are then drawn back through the lens, using The drawing in Figure 2-32 represents a 50mm lens on
the same rules as before, to determine the location of the a camera whose sensor diagonal is approximately 43mm.
object. Lines drawn from the rear nodal point (that is, the inter-
If we work with lenses too thick to generalize as thin section of the nodal planes and the lens axis) to opposite
lenses, or where greater accuracy is required, only small corners of the sensor form an angle that can be measured
modifications must be made to the thin-lens treatment. with a protractor. No compensation is necessary for the
If we know that the front and back nodal planes are drawing’s scale since there are 360° in a circle no matter
separated by a distance of 1 inch in a certain lens, two how large the circle is drawn. The angle of view in this
vertical lines are drawn on the lens axis to represent the example is approximately 47°.

Figure 2-31 The nodal planes are separated by an appropriate Figure 2-32 The angle of view of a lens-sensor combination is
distance for thick-lens diagrams. determined by drawing a line with the length equal to the sensor
diagonal at a distance of one focal length from the lens. The angle
formed by the extreme rays of light is measured with a protractor.

51
 s ecti o n 1: Captu r e

What happens if we don’t know the focal length of a

lens? No problem! Set the lens to form a sharp image of
an object on an optical bench (or equivalent low-tech
solution), measure object and image distances and use
Equation 2.7 to solve for f. For example, if the object and
image distances are both 8 inches, the focal length is
4 inches (or 101.6mm). This formula demonstrates the
Figure 2-33 Angles of view for 15mm and 135mm focal length lenses
on a full frame camera. inverse relationship between the conjugate distances
u and v, whereby moving a camera closer to an object
requires increasing the lens-to-sensor distance to keep the
Two other focal length lenses, 15mm and 135mm, are image in sharp focus.
represented in the drawing in Figure 2-33. The measured It also illustrates that u and v are interchangeable, mean-
angles of view are approximately 110° and 18°. It’s apparent ing that sharp images can be formed with a lens in two
from these drawings that using a shorter focal length lens different positions between an object and the sensor. For
on a camera increases the angle of view, whereas using a example, if a sharp image is formed when a lens is 8 inches
smaller sensor decreases the angle of view. When a camera from the object and 4 inches from the sensor, another (and
is focused on nearer objects, the lens-to-sensor distance larger) sharp image forms when the lens is placed 4 inches
increases and the effective angle of view decreases. from the object and 8 inches from the sensor (see Figure
Lens diagrams are especially helpful for new pho- 2-34). Exceptions to this rule: 1) when the object and image
tographers because they make it easy to visualize the distances are the same, which produces an image that is
relationships involved in image formation. With experience, the same size as the object; 2) when an image distance is
it’s more efficient to solve problems relating to distances larger than a camera body can realistically achieve between
and sizes using mathematical formulas. Most problems of
this nature encountered by practicing photographers can
be solved by using these five simple formulas:

1=1 +1 (Eq. 2.7)

f u v

R= v (Eq. 2.8)
u

R = (v – f ) (Eq. 2.9)
f

R= f (Eq. 2.10)
(u – f )

R = f(R + 1) (Eq. 2.11)

Where
f = focal length
u = object distance
v = image distance
Figure 2-34 Since the object and image distances are
R = scale of reproduction, which is image size/object interchangeable, sharp images can be made with a lens in two
size, or I/O. different positions between object and sensor.

52
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

the lens and the image plane to form the second sharp Depth of Field
image.
Photographers may encounter the problem of deter- Camera lenses focus on one object distance at a time. In
mining how large a scale of reproduction (R) is obtainable theory, objects in front of and behind the object distance
with a certain camera-lens combination. If a view cam- focused on are not imaged sharply on the sensor. In practice,
era’s maximum bellows extension (v) is 16 inches and is acceptably sharp focus is seldom limited to a single plane.
equipped with an 8 inch focal length (f ) lens, we use the Instead, objects somewhat closer and farther away appear
formula R = (v – f )/f = (16 – 8)/8 = 1, where the image is sharp. This is referred to as depth of field (DOF). Depth of field
the same size as the object. A longer focal length lens can is the range of object distances within which objects are
be used to obtain a larger image when the camera cannot imaged with acceptable sharpness. It is not limited to the
be moved closer (see Equation 2.4). However, a shorter plane focused on because the human eye’s resolving power
focal length lens is preferred in close-up work because the is finite; a circle up to a certain size appears as a point to our
maximum bellows extension is the limiting factor. Replac- visual system (see Figure 2-35). The largest circle that appears
ing the 8 inch lens with a 2 inch lens increases the scale of as a point is referred to as the permissible circle of confusion.
reproduction from R = 1 to R = 7. The size of the largest circle that appears as a point
Calculating focal length when an imaging system depends on viewing distance. For this reason, permissible
employs a combination of lenses requires the following circles of confusion are generally specified for a viewing
formula: distance of 10 inches and 1/100 inch is commonly cited
as an appropriate value for the circle diameter. Even at a
1=1 +1 (Eq. 2.12)
fc f1 f2 fixed distance, the size of the permissible circle of confu-
sion varies with factors including differences in individual
Where eyesight, the tonal contrast between the circle and the
fc = the combined focal length background, the level of illumination and the viewer’s
f1 = the focal length of the first lens criteria for sharpness.
f2 = the focal length of the second lens Although many photography books that include the
subject of depth of field accept 1/100 inch as appropriate,
If the lenses are separated by a physical space or dis-
tance (d) from the current lens, we use the following
formula:

1=1 +1 – d (Eq. 2.13)

fc f fs (f x fs )

Where
fc = the combined focal length
f = the focal length of the current lens
fs = the focal length of the supplementary lens
d = distance between f and fs

Figure 2-35 Depth of field is the range of distances within which

objects are imaged with acceptable sharpness. At the limits, object
points are imaged as permissible circles of confusion.

53
 s ecti o n 1: Captu r e

there is less agreement among lens manufacturers. A exposure time. A variation of this method is using three
study involving a small sample of cameras designed for movable objects, focusing on one and placing the other
advanced amateurs and professional photographers two at the near and far limits of the depth of field as indi-
revealed that values ranging from 1/70 to 1/200 inch were cated by the scale or table.
used—approximately a 3:1 ratio. We consider two different To judge the results, output 6x8 inch or larger prints to
methods for evaluating depth of field scales and tables for view from a distance equal to the diagonal of the prints.
specific lenses. The diagonal of a 6x8 inch print is 10 inches, which is con-
One method is photographing a flat surface with fine sidered to be the closest distance at which most people
texture or detail at an angle of approximately 45°, placing can comfortably view photographs or read. If the photo-
markers at the point focused on and at the near and far graph made at the f-number specified by the depth of field
limits of the depth of field as indicated by the depth of field scale has either too little or too much depth of field when
scale or table, as shown in Figure 2-36. The first photo- viewed at the correct distance, the photograph that best
graph should be made at an intermediate f-number with meets the viewer’s expectation should be identified from
additional exposures bracketing the f-number one and the bracketing series. A corresponding adjustment can be
two stops on both sides, with appropriate adjustments in made when using the depth of field scale or table in the
future.
The second method involves determining the diameter
of the permissible circle of confusion used by the lens man-
ufacturer in calculating the depth of field scale or table. No
photographs are needed for this approach. Instead, substi-
tute values for the terms on the right side of the formula:

f2
C= (Eq. 2.14)
(f-number x H)

Where C is the diameter of the permissible circle of

confusion on the sensor, f is the focal length of the lens,
f-number used and H is the hyperfocal distance at the
indicated f-number.
The hyperfocal distance is defined as: a) the closest dis-
tance that appears sharp when a lens is focused on infinity;
or b) the closest focus distance that renders an object at
infinity to appear sharp. When a lens is focused on the
hyperfocal distance, the depth of field extends from infinity
to one-half the hyperfocal distance (see Figure 2-37). If
f/16 is selected as the f-number with a 2 inch (50mm) focal
length lens, the hyperfocal distance can be determined
from a depth of field table, a depth of field scale on the
lens or by noting the near distance sharp at f/16 when the
lens is focused on infinity. If the near-limit marker on a DOF
Figure 2-36 Validating depth of field scales and tables by scale falls between two widely separated numbers, making
photographing an object at an angle with markers at the focus location accurate estimation difficult, set infinity opposite the
and the indicated near and far limits. Photograph by Rochester Institute
far-limit marker as shown in Figure 2-38 and multiply the
of Technology photography alumnus Oscar Durand
distance opposite the near-limit marker by two.

54
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-37 The hyperfocal distance is the closest distance that appears sharp when a lens is focused
on infinity (top), or the closest distance that can be focused on and have an object at infinity appear
sharp (bottom).

Since the circle of confusion is commonly expressed as

a fraction of an inch, the hyperfocal distance and the focal
length must be in inches. The hyperfocal distance at f/16
for the lens illustrated is 24 feet (288 inches). Substituting
these values in the formula from Equation 2.14 equates to
22/(16 x 288) or 1/1,152 inch. This is the size of the permissi-
ble circle of confusion on a full frame sensor, but this image
must then be magnified six times to make a 6x8 inch print
to be viewed at 10 inches. Thus, 6 x 1/1,152 = 1/192 inch, or
approximately half as large a permissible circle of confusion
as the 1/100 inch value commonly used.
Note that the size of the permissible circle of confusion
used by a lens manufacturer in computing a depth of field
table or scale tells us nothing about the quality of the lens
itself. The manufacturer can arbitrarily select any value and
in practice a size is selected that is deemed appropriate for
the typical user. If a second depth of field scale is made for
the lens in the preceding example based on a circle with
a diameter of 1/100 inch rather than approximately 1/200
inch, the new scale indicates that it is necessary to stop
down only to f/8 in a situation where the original scale
Figure 2-38 The hyperfocal distance can be determined from a depth
of field scale by focusing on infinity and noting the near-distance
specified f/16. Lens definition, however, is determined by
sharp at the specified f-number (top) or by setting infinity opposite the quality of the image for the object focused on and not
the far-distance sharp marker and multiplying the near distance sharp the near and far limits of the indicated depth of field.
by two (bottom). Photograph by Rochester Institute of Technology
photography alumnus James Craven

55
 s ecti o n 1: Captu r e

Depth of Field Controls

Photographers have three controls over depth of field:
f-number, object distance and focal length. Since viewing
distance also affects the apparent sharpness of objects
in front of and behind the object focused on, it is gener-
ally assumed that photographs are viewed at a distance
approximately equal to the diagonal of the picture. At this
distance, depth of field is not affected by making differently
sized prints from the same file. For example, the circles of
confusion at the near and far limits of the depth of field
are twice as large on a 16x20 inch print as on an 8x10 inch
print from the same image file. However, the larger print
would be viewed from double the distance, making the
two prints appear to have the same depth of field. If the
larger print is viewed from the same distance as the smaller
print, it appears to have less depth of field. Cropping when
enlarging decreases the depth of field because the print
size and viewing distance do not increase in proportion to
the magnification.

Depth of Field and F-Number

The relationship between f-number and depth of field
is a simple one: depth of field is directly proportional to
the f-number, or D1/D2 = f-number1 /f-number2. If a lens’ Figure 2-39 Depth of field varies in proportion to the object distance
squared. Photographs made at a scale of reproduction of 1:1 and
available f-numbers range is from f/2 to f/22, the ratio larger tend to have a shallow depth of field. Photograph by Rochester
of the depth of field at these settings is D1/D2 = (f/22)/ Institute of Technology photography alumnus Oscar Durand
(f/2) = 11/1. Changing the f-number is generally the most
convenient method of controlling depth of field, but focal lengths or less from the subject, the depth of field at
occasionally insufficient depth of field is obtained with a large aperture sometimes appears to be confined to a
a lens stopped down to the smallest diaphragm open- single plane (see Figure 2-39).
ing or too much depth of field is obtained with a lens At the other extreme, by focusing on the hyperfocal
wide open. Other controls must be considered in such distance, depth of field extends from infinity to within a
circumstances. few feet of the camera with some lenses (see Figure 2-40).
The mathematical relationship between depth of field
and object distance (provided the object distance does
Depth of Field and Object Distance not exceed the hyperfocal distance) is represented by the
Depth of field increases rapidly as the distance between formula D1 /D2 = u12/u22 where u is the object distance. For
the camera and the subject increases. For example, dou- example, if two photographs are made with the camera 5
bling the object distance makes the depth of field four feet and 20 feet from the subject, the ratio of the depths of
times as large. The differences in depth of field with very field is as follows:
small and very large object distances are dramatic. In pho-
D1 20 2
tomacrography, where the camera is at a distance of two = = 16 (2.15)
D2 5 2

56
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

If the object distance is increased to achieve a larger Depth of Field and Focal Length
depth of field when the camera lens cannot be stopped There is an inverse relationship between focal length and
down far enough, it is necessary to take into account depth of field: as focal length increases, depth of field
the enlarging and cropping required to obtain the same decreases. Using a 50mm focal length lens and a 300mm
image size as with the camera in the original position. focal length lens on the same camera produces a dramatic
There is still a net gain in depth of field in moving the difference in depth of field. The mathematical relationship
camera farther from the subject, even though some of the is D1 /D2 = (focal length2 )2/(focal length1 )2. This relationship
increase is lost when the image is cropped in printing. holds true only if the camera remains in the same place
The net gain is represented by the formula D1 /D2 = u1 /u2 , relative to the subject.
which is the same as the preceding formula with the
square sign removed.
Depth of Focus

Depth of focus is the focusing latitude present when pho-

tographing a two-dimensional subject. It is the distance
the sensor plane can be moved in both directions from
optimum focus before the circles of confusion for the
image of an object point match the permissible circle of
confusion used to calculate depth of field. Note that depth
of field calculations assume the image plane occupies a
single flat plane while depth of focus calculations assume
the subject occupies a single flat plane (see Figure 2-41).
If a three-dimensional object completely fills the depth of
field space, there is only one position for the sensor, there
is in effect no depth of focus and no tolerance for focusing
errors.

Figure 2-40 Focusing on the hyperfocal distance produces a depth of

field that extends from infinity to one-half the hyperfocal distance.

Figure 2-41 Depth of field calculations are based on the assumption

that the two-dimensional image plane is in the position of optimum
focus (top). Depth of focus calculations are based on the assumption
that the subject is limited to a two-dimensional plane (bottom).

57
 s ecti o n 1: Captu r e

Depth of focus for a two-dimensional subject is calcu- Macro Lenses

lated by multiplying the permissible circle of confusion by Lenses producing excellent quality images with objects at
the f-number by 2. Using 1/200 inch for the permissible moderate to large distances may not perform well when
circle of confusion on a 6x8 inch print or 1/1200 inch on a used at small object distances. Macro lenses are designed
full frame sensor, the depth of focus is C x f-number x 2 = specifically for photographing at small object distances
1/1200 x f /2 x 2 = 1/300 inch. From this formula we can also while rendering the subject at or greater than life size on
see that depth of focus varies in direct proportion to the the sensor. They are designed with additional aberration
f-number, as does depth of field. corrections that are otherwise apparent at such magnifi-
Depth of field decreases when the camera is moved cations such as focus fall off along image edges (shown in
closer to the subject, however at the same time the depth Figure 2-42).
of focus increases. As the object distance decreases, the It’s common to see lenses boasting macro capabilities
lens-to-sensor distance must be increased to maintain despite not being able to render the subject greater than
sharp focus of the image. This increases the effective life size. Here the use of the term is misleading, replac-
f-number. It is the effective f-number, not the marked ing its technical definition with one broadly describing
f-number, that is used in the formula above. an ability to focus at close distances (which we would
Although increasing focal length decreases depth of
field, it does not affect depth of focus. At the same f-num-
ber, both the diameter of the effective aperture and the
lens-to-sensor distance change in proportion to changes in
focal length. Consequently, the shape of the cone of light
falling on the sensor remains unchanged. It follows that as
focal length does not appear in the formula C x f-number x
2, it has no effect on depth of focus.
Although changing sensor size does not seem to affect
depth of focus, using a smaller sensor reduces the correct
viewing distance and, therefore, the permissible circle of
confusion. Substituting a smaller value for C in the formula
C x f-number x 2 reduces the depth of focus.

Specialty Lenses

So far we’ve covered normal, telephoto and wide-angle

lenses. That trio works as an excellent set for photogra-
phers that shoot a variety of subjects and scenes, allowing
us to shoot across a range of distances and fields of view.
Normal, telephoto and wide-angle focal length lenses
are a versatile combination in any camera bag. We turn
to specialty lenses for specific applications, effects or
image-making requirements. Figure 2-42 Photographs of a small object made with a standard
camera lens (top) and a macro lens (bottom), both at their
maximum aperture. Photograph by Rochester Institute of Technology
photography student Kristina Kaszei

58
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

presumably do with small subjects). These lenses tend to perspective convergence of rectilinear lines like the ones
produce a maximum scale of reproduction no larger than seen in Figure 2-43. Left uncorrected, the vertical building
1:2. If you need to occasionally photograph larger subjects elements appear to lean. Architectural photography is a
at very close distances, such a “macro” lens is perfectly smart fit for perspective control lenses as shifting helps
adequate—but you’ll need a true macro lens to photo- render subjects as they are without distortion.
graph particularly tiny subjects like insect eyes or pollen A normal lens allows us to focus in one plane perpendic-
grains on a flower. ular to the optical axis of the lens; the tilt function allows for
manipulation of the plane of focus from the perpendicular.
Carefully considered tilt-shift movements help to extend
Perspective Control Lenses the plane of focus to minimize focus fall-off.
Also known as tilt-shift, perspective control lenses intro- Figure 2-44 (overleaf) highlights the other outcome
duce perspective or distortion controls on SLRs that have whereby intentionally introducing extra distortion or
historically only been available with view cameras. The lens unusual focus fall-off. Interestingly, a unique side effect
housing of a tilt-shift lens introduces a variable position- of this imaging method is the illusion of miniaturization.
ing of the optics relative to the sensor plane, whereas a We’re used to seeing tiny things up close with limited
traditional lens uses optics in a fixed position and distance. depth as our eyes adjust to the close focusing distance
As a consequence, perspective control lenses can be a and small scale. When we see extremely shallow depth of
range of focal lengths and behave exactly like their non-tilt- field on scene—especially at elevated vantage points—
shift counterparts when set in a neutral position. Adjusting more commonly photographed with everything in
the lens housing positioning knobs shifts the optics up, focus, we start to perceive them as miniaturized. Unusual
down, left or right and can also tilt up or down (hence the placement of the in-focus sweet spot can be used for
nickname). other compositional effects, though this is also possible
There are two outcomes when employing a perspec- by holding any normal lens in front of the camera body
tive control lens. The first is counteracting the natural without mounting it.

Figure 2-43 Before (left) and after (right) perspective correction using a tilt-shift lens. Photograph by Rochester Institute of Technology student
Adam Montoya

59
 s ecti o n 1: Captu r e

Figure 2-44 A miniaturization effect is introduced with a tilt-shift lens by manipulating the focus plane to drastically decrease depth of field and
focus fall-off. Photograph by Rochester Institute of Technology photography student Joel Beckwith

Catadioptric Lenses mirror and again by the smaller mirror on the back of the
Notwithstanding the shorter lens-to-sensor distance glass element and it passes through the opening in the
obtained with telephoto lenses compared to normal lenses concave mirror to form the image at the sensor plane. The
of the same focal length, the distance becomes inconven- glass element and the opaque stop reduce aberrations
iently large with long focal length lenses. Catadioptric lenses inherent to the mirror system.
achieve a dramatic improvement in compactness through Catadioptric lenses are capable of producing images
the use of folded optics, combining glass elements and with excellent image quality with some drawbacks. Their
mirrors. The name catadioptric is derived from dioptrics design makes it impractical to contain a variable-diameter
(the optics of refracting elements) and catoptrics (the aperture, so exposure must be regulated by other means
optics of reflecting surfaces). Similar designs are employed and depth of field control is nonexistent. It also requires
for telescopes. very large lens diameters to achieve long focal lengths.
Figure 2-45 illustrates the principle of image formation Finally, these lenses have a signature (and not necessarily
with a catadioptric lens. A beam of light from a distant desirable) image artifact in out-of-focus areas that look like
point passes through the glass element, except for the rings or donuts.
opaque circle in the center; it is reflected by the concave

Reversed Telephoto Wide-Angle Lenses

Reversed telephoto wide-angle lens designs are used to
avoid interference between the lens and the mirror.
Problems may be encountered when using conventional
design, short focal length wide-angle lenses because of the
resulting short lens-to-sensor distances. The lens designer’s
solution is to reverse the arrangement of the elements in
Figure 2-45 Catadioptric lenses make use of folded optics to obtain
long focal lengths in relatively short lens barrels. a telephoto lens, placing a negative element or group of

60
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-46 The back nodal plane is behind the lens with reversed Figure 2-47 Positive supplemental lenses enable the camera to focus
telephoto wide-angle lenses, providing a larger lens-to-sensor on near object without increasing the lens-to-sensor distance.
distance than for a normal-type lens of the same focal length.

elements in front of and separated from a positive element supplemental lenses behave. Adding a positive supple-
or group of elements. This design places the image nodal mental lens lowers the focal length; it cannot convert
plane behind the lens (or near the back surface), which a normal lens into a wide-angle one. Therefore, using a
effectively moves the lens farther away from the sensor supplemental lens when photographing objects at far
(see Figure 2-46). Lenses of this type are called reversed distances may not provide the desired covering power.
telephoto, inverted telephoto and retrofocus wide-angle However, using a positive supplemental lens allows you to
lenses. focus on new objects without increasing the lens-to-sensor
distance (see Figure 2-47).

Supplemental Lenses
Photographers can use a lower-cost, single element Extension Tubes and Teleconverters
supplemental lens to give their general-purpose lens more Supplemental lenses are not the only way to modify the
functionality. A supplemental lens alters focal length when image produced by a camera lens. Extension tubes increase
attached to a lens. Adding a positive supplemental lens the image distance between the lens elements and the
produces the equivalent of a shorter focal length. These sensor. They do not contain any additional optics and
supplemental lenses are attached close to the front surface instead act as a physical separator to position the back
of the camera lens, making it possible to calculate the focal of the camera lens farther from the camera body mount.
length of the lens combination with reasonable accuracy They increase magnification and allow for closer focusing
using the formula 1/fc = 1/f + 1/fs, where fc is the combined distances. The downside to using extension tubes is light
focal length, f is the focal length of the camera lens and fs is loss. With additional distance to travel, the additional light
the focal length of the supplemental lens. fall-off raises the effective f-number of the attached lens.
Supplemental lenses are commonly calibrated in diop- Adding optical elements to further increase image mag-
ters. Diopter power equals the reciprocal of the focal length nification results in a teleconverter. These have the same
in meters of D = 1/f. To convert the diopter value to a focal drawbacks as extension tubes when it comes to reducing
length the previous equation becomes f = 1/D. Diopters are the amount of light reaching the sensor. Teleconverters
the measure of lens power used for corrective eyeglasses are often categorized by the amount of magnification they
and contacts, which may make it easier to remember how introduce, i.e. 1.4x or 1.5x. It’s possible to chain together

61
 s ecti o n 1: Captu r e

a long focal length lens and multiple teleconverters to Chromatic Aberration

achieve very long effective focal lengths, though the We understand now that positive lenses work by bringing
weight and physical length introduced by this assembly is light rays to a single point of focus at the location of our
unwieldy for handheld or portable use. camera’s sensor. Lenses accomplish this by the principle of
refraction whereby light changes direction when it travels
through the lens material. The catch is that light changes
Lens Shortcomings and Aberrations direction or bends to varying degrees based on its wave-
length (dispersion). Blue light rays do not come into focus
It’s difficult to make a single, perfect lens for making at the same position as red light rays—a problem when
photographs. Earlier in this chapter we defined concepts visible light consists of a whole variety of wavelengths.
of photographic optics using simple lenses due to their Unless actively corrected, shorter wavelengths (blues)
straightforward design. However, nearly every lens used bend more and come into focus in front of the image
for making pictures is actually a compound lens consisting plane while longer wavelengths (red) bend the least and
of groups or sets of different optical components in series. come into focus behind the image plane (see Figure 2-48).
Compound lens systems are necessary for minimizing the This is called chromatic aberration. We’re simplifying these
optical imperfections of any one lens. Photographic optics groupings, but the bottom line is that our full color image
have shortcomings that, unmitigated, result in aberra- is at risk of forming across multiple planes when our sensor
tions or departures from the ideal image that appear as can only capture one.
unwanted image artifacts or imperfections. Lens shortcom- Chromatic aberration appears as color fringing along
ings can be grouped into four categories relating to image subject edges in a photograph and it is especially notice-
definition, image shape, image illuminance and image able along edges with considerable contrast, such as
color. an object’s silhouette against a bright sky. They are not
Here we describe the aberrations most commonly necessarily red, green or blue. In fact, chromatic aberration
encountered by photographers. It is not a comprehensive in modern lenses is often yellow and/or magenta (see
list; some optical shortcomings are only problematic in Figure 2-49). Wide-angle lenses are particularly susceptible
specific use-cases like microscopy or telescopes, so we’ve to color fringing due to the extreme angles that light enters
put those to the side. Distortions, color fringing and focus and bends through them. Generally speaking, chromatic
aberrations tend to be more extreme and noticeable along aberration appears most prominently near the corners of a
frame edges. photograph where the optics are the least corrected.

Figure 2-48 Chromatic aberration as light travels through a positive lens. It is commonly corrected using an achromatic doublet lens.
Adapted from images by Andreas 06 and DrBob, released into the Public Domain under a Creative Commons license

62
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-49 Chromatic aberration visible along contrasty edges. Figure 2-50 Spherical aberration as light travels through a positive
Photograph by Ruben Vargas lens.

A common optical solution to counteract this aberration light to bring an image into focus. Light passing through
is achromatic doublet lens elements. This two-lens system the edges of the lens bend more than the light traveling
pairs a positive, low index of refraction crown glass with a through the center as illustrated in Figure 2-50. This results
negative, high index of refraction flint glass. The differing in an unsharp image and is called spherical aberration. As
indices of refraction effectively correct the shortcomings with chromatic aberration, combinations of additional lens
of the system and reduce chromatic aberration in the final elements in the optical system are introduced to mitigate
image. this image quality degradation. Aperture size can also reduce
Interestingly, our eyes also suffer from chromatic spherical aberration by blocking the light from the outside
aberration. There are a few things that help minimize its edges of the lens. Spherical aberration is reduced as the lens
appearance. There are no blue-sensitive light detectors or aperture is stopped down, giving the photographer some
S cones at our centralmost vision (the fovea), only red (L control over the degree of unsharpness.
cones) and green (M cones) sensitive ones. Additionally, We often get caught up in chasing perfectly sharp,
the brain mitigates the perception of chromatic aberration well-defined photographs. Sometimes artistic intent
when it processes the signals from our eyes. It’s believed dictates something decidedly less accurate or optically pre-
that the chromatic aberration inherent to human vision is cise. Soft-focus specialty lenses under-correct for spherical
interpreted to better understand depth and drive accom- aberration, leaving us with a unique effect sometimes used
modation, though we are not consciously aware of it.2 in portrait photography or pictorialist-style aesthetics.

Spherical Aberration Diffraction

Some light rays travel through the center of a lens while Diffraction is the only lens aberration that affects the
others travel through its periphery. This can mean that definition of images formed with a pinhole. The principles
the light rays travel through different amounts of the lens of geometrical optics tell us that image definition increases
medium; remember, positive lenses form images by bending indefinitely as a pinhole is made smaller. In practice, there

63
 s ecti o n 1: Captu r e

is an optimal pinhole size and image definition decreases the optical and mechanical elements of a complete lens
due to diffraction as the pinhole is made smaller. The system. We can’t remove the barrel, so we can’t entirely
narrow beam of light passing through the pinhole from an eliminate this type of vignetting. Optical vignetting
object point spreads out not unlike water coming out of a becomes more apparent as the aperture is adjusted to
hose nozzle. The smaller the pinhole, the more apparent larger, more open settings. Some of the light rays enter-
the effect. ing the lens from the edges of the scene are blocked by
Diffraction happens when light bends around the edge the lens barrel and are rendered with less exposure or no
of an object which, in our case, is an aperture. The amount exposure at all, causing the vignette. The effect is min-
of bending depends on the wavelength of the light and imized by stopping down the lens. This particular issue
the size of the opening. When the aperture is much larger can rear its ugly head with cheaper lenses that can suffer
than the wavelength, the diffraction is barely noticeable. from noticeable vignetting at wider apertures, leaving
When the wavelength of the light and the size of the aper- you with the choice between minimizing vignetting and
ture are nearly the same, diffraction is noticeable to the eye using a desired, stopped-down aperture setting. Holga
and renders the image soft. film cameras are known for—and sought out specifically
In practice, the definition of an image formed by an because of—their cheap design that frequently exhibits
otherwise perfect lens is limited by diffraction. Some lenses strong vignetting.
are diffraction-limited because under the specified condi- Natural vignetting also appears as a gradual darkening
tions they approach optical perfection. Using resolution or reduction in brightness from the center of the image to
as a measure of image definition, the diffraction-limited the edges. Its root cause is different from optical vignetting,
resolving power is approximated with the formula: however. Light rays entering the lens at its center travel in
a straight line and have a shorter distance to travel to the
1,800
R= (Eq. 2.15) sensor compared to rays that enter at an angle. The center
f-number
of a sensor or film plane receives more light and therefore
where: more exposure than at its periphery. The cosine fourth law
R = the resolving power in lines per millimeter quantifies this type of vignetting.
1,800 = constant for an average wavelength of light of Pixel vignetting appears the same as natural vignetting.
550 nm It’s caused by the physical photosite that receives the light
rays collected by the lens optics. The effect of the cosine
Applying this to a hypothetically perfect lens with min- fourth law is at play here because a photosite is not a
imum and maximum f-numbers of f/2 and f/22, this flat, two-dimensional light detector. Instead they can be
equation tells us that its diffraction-limited resolving power thought of like trenches or wells; as light rays travel toward
is 900 and 82 lines/mm, respectively. the photosites, those that are headed straight into the
opening make it with greater success than ones traveling
off-axis.
Vignetting Imagine the classic carnival game of ping pong toss.
Vignetting or light fall-off is the reduction in brightness from The player throws balls into a tightly arranged array of fish
the center to the outer edges of an image. There are sev- bowls and observes that many shots bounce off the bowl
eral causes and therefore categories of vignetting: natural, edges instead of landing inside them. The game would
optical, mechanical and artificial. Every lens exhibits some be much easier if the player is positioned directly above
optical and natural vignetting as a reality of lens design. the table of fish bowls, dropping ping pong balls straight
Optical vignetting appears as a gradual darkening from down. The odds of winning increase when there’s minimal
the center to the edge of the image and can be seen reliance on perfectly thrown, carefully calculated parabolic
in Figure 2-51. It’s caused by the lens barrel that houses arcs made from an oblique angle. The carnival game gets

64
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

harder when the player’s distance from the bowls is greater file metadata containing the lens focal length and model
and the ping pong balls’ angle of travel is less direct. against a lens characteristic database created by software
Similarly, light rays headed toward photosites have a lesser developers. These developers quantify the vignetting
probability of making it into the well if they approach at behavior of most major lens manufacturer offerings by
an angle. Blocked light rays become a form of light vignet- conducting controlled lab shoots. From this information
ting. It’s an unavoidable reality of image formation and our a counter-vignette is applied to the image, lightening
best strategy is mitigation. Microlenses are installed atop edge pixels that are known to be darker than their center
photosite wells to better collect or catch incoming light. counterparts. This same strategy is used by camera man-
If carnival game analogies aren’t helping to visualize the ufacturers, though the process happens in-camera when
problem, jump to Chapter 3 for digital sensor illustrations. saving to JPEG; often the proprietary lens characterization
Mechanical vignetting is a very abrupt artifact caused by information is readable only by their respective software for
a physical obstruction in the light path. A long lens hood raw captures. Third party image editing tools have to create
or a filter holder attached to the front of the lens barrel are the information on their own and do not necessarily have
likely culprits, especially if they are designed for focal length prepared vignette correction for obscure or rare camera
lenses different from the ones used. Mechanical vignetting lenses.
is eliminated by removing the obstruction.
The last type is artificial or post-processing vignetting.
Photo editing software offers tools that allow us to add a Bokeh
darkening effect that emulates natural or optical vignetting The blurry, out-of-focus areas that result from points of
for aesthetic effect (see Figure 2-51). It’s useful for drawing light not at the plane of focus are called bokeh. This term is
the eye toward the center or toward a brighter, highlight colloquially derived from boke, Japanese for “blur” or “haze”
area of the composition. On the right image and with a in the context of photography. While not an aberration per
delicate hand, the tool is a valuable one. However, it is easy se, bokeh is a characteristic inherent to a camera lens that
to overdo and risks becoming a cheap shortcut to better some photographers pay particular attention to.
composition and refined tone adjustments. The shape of out-of-focus areas, especially bright points
Optical vignetting correction is offered in professional of light, is a direct consequence of the aperture open-
image editing software. The software can leverage image ing shape. The aperture opening is created by a set of

Figure 2-51 Photographing a flat, uniformly lit surface is a quick way to spot vignetting (left). Darkened edges, either optical or artificial (post-
processing) means, can be an attribute to the final image’s appearance and visual focal point (right).

65
 s ecti o n 1: Captu r e

sides. The image distortion is greatest at the edges of the

frame and the least at its center. Super wide-angle lenses
exhibit barrel distortion, as it’s an inherent consequence
of refracting extremely oblique angles of light onto the
camera sensor (see Figure 2-53). Pincushion distortion
is the opposite effect where straight vertical lines bend
inward toward the center of the frame. Here, too, the
distortion is at its most extreme near the image frame
edges. Either type renders shapes differently from how
they appear in the world. Telephoto lenses can suffer from
pincushion distortion though not to the obvious degree
Figure 2-52 The aperture opening shape is seen in out-of-focus that a fisheye lens suffers from barrel distortion (see Figure
specular highlights. This photograph shows the bokeh quality for a
100mm macro lens set to f/8 with considerable distance between the
2-54). Both can be largely corrected in post-processing,
subject and background elements. particularly if reference data exists that document the exact
degree of distortion introduced by a lens—called lens pro-
files. Lens profiles are created by photographing reference
overlapping metal petals or blades attached to a mech- targets and mathematically determining the distortion
anism that can dilate or constrict its diameter. The petals introduced by a lens. They offer a type of engineering
themselves have straight edges and therefore create small instructions to undistort image content when applied by
angles at the points of overlap. The opening that results post-processing software. The only downside to using lens
is similar to drawing a perfect circle using a ruler instead profiles to correct for significant distortion is that cropping
of a compass: you have to draw a group of straight lines may be required to retain a rectilinear image frame.
at consistent angles to close the circle. Lens apertures are
designed using anywhere from 5 to 20 individual blade
segments. The more segments, the closer the opening Flare
shape resembles a true circle. The closer the opening is to a Flare is non-image forming light in our imaging system that
circle, the more circular the bokeh appears. degrades image quality at capture. With all photographic
Lens bokeh is most apparent when shooting at wider optics, our goal is to control and direct light to form an
apertures or when focused at close distances with a image of a scene to be recorded by the light-sensitive
distant background as evident in Figure 2-52. While the medium. For the most part, our lenses do this well, espe-
shape of the blurred points of light is closely related to the cially because lens manufacturing includes steps to coat
aperture shape, it is not the only influencing factor. The lens elements with a number of special optical materials to
optical design of the lens also dictates the way in which minimize internal reflections. However, there is inevitably
out-of-focus points are blurred. Some lenses have better some light that scatters or otherwise bends at angles that
correction at their center than at their edges, meaning that results in a washed out, low contrast image. The cinematic
bokeh near the frame corners can be further distorted or flare bursts that appear near the sun in scenic landscapes
blurred. have aesthetic appeal, but that’s the crowd-pleasing side
of flare. The not-so-desirable flare is the kind that brings
a general haziness to photographs and is often tough to
Barrel and Pincushion Distortion spot in the moment. This flare happens when a bright
Barrel and pincushion distortion are common image-for- light source is either directly within the lens’ line-of-sight
mation pitfalls. Barrel distortion is when straight lines bend or when one is at an extreme, oblique angle relative to the
outward, similar to how a wooden barrel bulges out at the lens’ front element. In both cases, light enters the lens and

66
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

Figure 2-53 Barrel distortion using a fisheye lens at 15mm (left). Even this extreme distortion can be counteracted
with distortion correction controls in post-processing software, though cropping from the original framing is
necessary (right).

Figure 2-54 Pincushion distortion using a telephoto lens at 70mm (left). Distortion correction helps to keep
the rectilinear architecture lines straight (right).

67
 s ecti o n 1: Captu r e

Unless you’re reading this from a cubicle at an optical

engineering lab, we assume this part is taken care of by
the time you buy a lens. You may be familiar with such
coatings if you wear eyeglasses since the same problem
exists there. These coatings help but they can never
entirely eliminate flare.
2. Use a lens hood attached to the front of the lens barrel.
These tube-like or petal-like attachments block extrane-
ous and oblique angle light from entering the lens. Lens
hoods can be low-tech or makeshift. As long as they
don’t creep into the field of view, anything that works to
minimize flare is worthwhile. Well-designed lens hoods
are coated on their inside surface with flocking material
Figure 2-55 A studio lighting setup with (left) and without (right)
a light aimed at the camera lens within its field of view. The light (think black velvet or felt) that absorbs light rays before
introduces significant flare that reduces image contrast. they have a chance to enter the lens at flare-inducing
angles. This same material is used inside the lens barrel
for the same functional purpose. Lens hoods exist for
bounces around in a random way, eventually reaching the most focal lengths.
sensor and adding exposure to areas that shouldn’t receive 3. Avoid pointing the camera right at a light source. It’s
it. This reduces contrast because dark areas become lighter, often unavoidable to have the sun in a landscape scene
as can be seen in the grayscale patches on the target in but we’re not suggesting that you photograph only
Figure 2-55. In a way, flare is like a kind of noise or static in in the shadows and scorn the daylight. It’s a matter of
the form of light hitting our sensor that isn’t contributing to awareness: in the studio, it’s easy to unwittingly position
forming an accurate image of our scene. a light source seemingly off-camera that introduces
Flare can also be introduced or exacerbated by the contrast-deteriorating effects as demonstrated in Figure
camera design. Light travels through the lens, out through 2-55.
the rear lens element, through air (a small distance within 4. Keep everything clean. Fingerprints, smudges, dirt
the camera body) and finally reaches the sensor. Reflective or any other debris that deposits on an exposed lens
materials in the camera body or a poorly light-sealed con- surface act as a light scattering medium. We’ve all seen
nection between the lens and camera can introduce flare. terrible smartphone images due to a smudged, finger-
This is why you won’t see cameras designed with anything print-oil covered lens. Keep a microfiber lens cleaning
other than black on the inside. Nor do you want a camera cloth handy and stay vigilant: when bright spots or light
body made out of transparent plastic. Apple learned this sources in a scene start showing up in photographs as
the hard way when designing the iPhone 4 in black and smeary streaks, it’s a strong indication that your optics
white models: it’s speculated that they delayed the white hygiene needs work.
model for months once they discovered that its camera
suffered from excessive flare due to some of the materials
used.3 Proof of Concept: Focal Length,
A multi-pronged strategy is the way to go when avoid- Subject Distance and Portraits
ing flare:
The principles of perspective in photography are observ-
1. Treat the lens elements with special coatings that limit able by taking a sequence of portrait images using varied
unpredictable light spreading and internal reflections. focal length lenses and subject distances. Portraits are

68
 c h a p t e r 2 : P h oto g r a p h i c O p t i c s

particularly important because we’re highly attuned to equivalent framing as the previous exposures. Keep
human faces and how ourselves and others appear in f-number the same and capture a third image.
images. Let’s look at the effect of altering both lens focal
length and subject distance simultaneously by taking a Questions to consider when reviewing the resulting
sequence of photographs with the following steps: photographs:

1. Obtain three focal length lenses: a lens in the range of • How does the relative distance between the subject
18 to 24mm, a 50mm lens and a lens in the range of 150 and the background appear to change when shooting
to 200mm. between the shortest to longest focal length lenses?
2. Find a location where your model can sit or stand in • How do the subject’s features appear to change
place. between the shortest and longest focal lengths? Which
3. Start with the shortest focal length lens on the camera version is the most flattering or most representative of
and fill the frame with your subject. We suggest a tight how you view this person in face-to-face interactions?
crop on the subject’s head and shoulders to maximize Can you imagine scenarios where you would pick the
the effect. Capture an image. focal length for its unique traits?
4. Switch to the 50mm lens. Work to maintain a similar • Which lens emphasizes the background? Which lens
framing as the previous image by moving farther away. minimizes the background?
The larger focal length lens means that a farther dis-
tance is needed to maintain an equivalent field of view. Figure 2-56 shows one set of results from this exercise.
Keep f-number the same and capture a second image. Focal length and camera position change in each image
5. Switch to the long focal length lens and reposition such that our subject, the young man, stays relatively the
yourself relative to your subject to again maintain an same. Notice how the background appearance changes

Figure 2-56 A sequence of portraits made by changing lens focal length and subject distance. Photographs by Rochester Institute of Technology
photography students Dorothy Marquet and Ben Morrow

69
 s ecti o n 1: Captu r e

dramatically from image to image. Though the frame is Notes

dominated by the subject, his apparent relationship to 1 Goolkasian, Paula. “Size scaling and spatial factors in visual
the background and environment is altered. Altering the attention.” American Journal of Psychology, vol. 110, 1997, p.
camera location and the focal length used allows us to 397.
either emphasize or diminish the background in the scene. 2 Cholewiak, S.S., G.D., Love, P.P. Srinivasan, R. Ng, and M.S.
Banks. “ChromaBlur: Rendering chromatic eye aberration
This applies our understanding of photographic perspec-
improves accommodation and realism.” ACM Transactions
tive to intentionally create a desired visual effect. Don’t
on Graphics, vol. 36, no. 6, 2017, Article 210.
blame your environment for poor photographs when you
3 Camm-Jones, Ben. “Apple explains white IPhone 4 delay.”
can solve it with smartly chosen photographic optics and
PCWorld, PCWorld, Apr. 28, 2011, www.pcworld.com/
positioning. Also consider the impact to facial features: article/226520/Apple_Explains_White_iPhone_4_Delay.
longer focal lengths are generally more flattering! The html.
added subject distance required at longer focal lengths can
make your subject more comfortable and at ease, while
short focal lengths require you to move into someone’s
personal space.

70
 3 Digital Camera
Technology

Photograph by Rochester Institute of Technology photography alumnus Dan Wang

The previous chapter concentrated on optics forming an image on the image plane, the
location at which a light-sensitive material is placed. The image-recording technology of
choice for more than 150 years was film: an analog and chemistry-dependent medium.
Today, the dominant light-sensitive material in cameras is the image sensor, a digital and
electronics-dependent medium. The goal is the same today as it was 150 years ago: to
record light information in a permanent way. If we consider the sensor as the heart of the
camera, we have to support and understand the additional anatomy that gives life to digital
photographs. This chapter dives into that anatomy: the sensors, apertures, shutters and

71
 s ecti o n 1: Captu r e

other advanced hardware and software mechanisms that out with different lenses. Professional photographers find
work in harmony to control and record light as electronic unmatched utility in being able to swap lens focal lengths
information. Camera technology is always evolving to meet at a moment’s notice. There’s much more going on inside,
new demands and expand image-making capabilities, yet though, that makes the SLR distinct from a smartphone
the primary components remain the same. with clip-on lens attachment.
Figure 3-1 is a cutaway illustration of a single lens reflex
camera’s internal components. Light reflected from an
The Single Lens Reflex Camera object in front of the camera enters the lens (1) and is
Design focused on the main reflex mirror (2). Here the light takes
two possible paths. Most of the light is directed through the
Studying the single lens reflex (SLR) camera design is a great focusing screen (7) and condenser lens (8) up to the pentap-
way to understand the components and functionality of a rism (9) which is located just in front of the viewfinder (10).
digital camera. Keep in mind that the fundamental design Chapter 2 explained that the image formed by a lens is
is adapted from its film predecessors, with the image plane inverted and reversed (or upside-down and backwards). The
swapping analog for digital. Solely going off of appearance, pentaprism is an arrangement of five mirrors that flip the
it’s difficult to appreciate how differently SLRs operate from object image back to normal so we see it in the viewfinder
point-and-shoots or other camera designs. The primary as it appears in front of us. The image in the viewfinder
difference at a glance is the large lens attached to the is the same as the one seen by the sensor, offering up a
camera body. One of the key aspects of the SLR design is its synchronized experience between framing the shot and
interchangeable lens system. A mount on the body allows the resulting capture. The reflex mirror also allows some
for the entire set of optical components to be changed of the light to pass through to a secondary mirror (4) which

Figure 3-1 A cross section view of a single lens reflex camera reveals the multiple paths traveled
by the image before, during and after capture.

72
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

directs the image to a light sensor that determines exposure, recorded as a voltage by the sensor. The reality, however,
a phase detection autofocus sensor, or both (3). When the is that some of those photons are left unrecorded. The
shutter button is pressed, the main reflex and secondary mir- ratio of incoming photons to electric charges collected
rors flip up out of the light path, allowing the image formed describes the quantum efficiency (QE) of the sensor. The
by the lens to reach the image sensor (6). The sensor detects higher the quantum efficiency, the less light is wasted in
and records the light. Once the shutter curtains (5) complete the process. QE is wavelength-dependent, meaning that a
their opening and closing event, the mirrors flip back down sensor’s efficiency is different if it’s recording mostly blue
to their original positions. During exposure, light is blocked light compared to red (just as one example). If you see a
from traveling to the viewfinder and the light and/or focus single value describing QE of a sensor, it’s likely an average
sensors as it is all redirected to the image sensor. performance across its sensitivity spectrum. Alternatively,
a photosite may receive zero photons if no light is present.
Photosites with no photoelectric activity during exposure
Recording Light with Digital Image end up as black pixels in the image file.
Sensors Two things have to happen next: first, the voltage is
amplified to a usable level (the initial voltage is too small to
Our image sensors record light using a two-dimensional create a viable photograph). The voltage amplification has
array or grid of light detectors called photodiodes, photo- shortcomings, as we’ll see when we take a deeper look at
sites or silicon photodiodes (SPDs). Photosites is preferred for photographic noise. Second, our analog input signal must
brevity’s sake and we use it throughout this text, though be converted to digital information—zeroes and ones, the
the other terms are accurate alternatives. Photosites are language that computers speak—this is called the analog
simple in that they detect light that varies from very dark to digital conversion. Each photosite’s electrical voltage
to very bright. They do not inherently understand or record becomes a digital value or digital count, one of a defined
anything about the color of a scene, only the range of set of values that represents a brightness. Furthermore,
tones in it. These light detectors are placed close together these digital values must be discrete, whole numbers,
yet there is always some miniscule space between them. whereas analog light signals in nature can be a continuous
Additionally, they need to pass along the amount of light spectrum of values (see Figure 3-2).
they detect to digital circuitry, which has to be close by
if the data is going to stay true and accurate. So where 10
does this circuitry go? What happens to light that lands in
9 Analog Input
between the photosites? Where does color come into play? Digital Output
8
How do light photons become digital values? It’s easy to
Example Digital Output

7
say that the sensor records the scene focused by a lens but
6
there are many details that make such a feat possible.
5
An image sensor translates an optical image, comprised
of many millions of photons, into an electrical signal that 4

is stored and interpreted by a computer. The sensor’s 3

photosites contain a material (often silicon) that produces 2

electric charges when struck by these photons, a behavior 1

called the photoelectric effect. The more photons that reach 0

0 0.2 0.4 0.6 0.8 1
a photosite, the more charge is generated. The greater the Normalized Analog Input Voltage
charge, the brighter the resulting pixel value. This charge is
Figure 3-2 Converting analog light information into digital counts
recorded as voltage: an analog phenomenon. In a perfect requires quantizing, making an infinite amount of variability into a
world, every photon of light is faithfully detected and discrete and finite set of numeric values.

73
 s ecti o n 1: Captu r e

Pixel defects are another unavoidable artifact in record-

ing digital images. Making a sensor that is totally free of
defects is nearly impossible because of the large number of
pixels and the complex manufacturing processes involved
in making sensors. There are hot pixel defects, in which
the pixel has a high dark-current noise, and there are dark
pixel defects, in which the pixel has a lower response than
surrounding pixels. Other than scrapping the sensor and
manufacturing a new one, there’s no easy fix for these
defects. Instead, cameras may incorporate a form of
masking whereby the internal software evaluates its sensor
Figure 3-3 The Bayer color filter pattern sits above the light-sensitive
for defects, notes their locations and borrows from neigh- photosites.
boring photosites to fill in the defective spots. This may be
bypassed when shooting raw files but is often an auto-
matic process in JPEG capture.
Photosites are covered with color filters to record color Bayer Pattern Color Filter Array
information from a scene. When a blue filter is placed over The color filter array (or CFA) most commonly found on
a single photosite, for example, the only light energy that camera sensors is the Bayer Pattern filter, invented in 1976
travels to the light-sensitive material and recorded as a by Bryce Bayer at Kodak. This pattern alternates the filter
charge is the blue light present. If there is no blue light, the colors between red, green and blue. Rather than having an
photosite doesn’t record much of anything. Since every even distribution of each, this pattern uses twice as many
color can be made through a combination of red, green green filters as it does red and blue: the ratio of a Bayer
and blue (see Chapter 13), it follows that filtering an image pattern CFA is 50% green, 25% red and 25% blue (illus-
sensor with a combination of red, green and blue filters trated in Figure 3-3). This results in photosite data for only
gets us closer to a full color recording. Introducing any color one-third of the color information needed to render a full
filters means a little less light sensitivity, as we’re preventing color photograph. The other two-thirds are algorithmically
some of the incident light from traveling to the photosite. In derived through the process of demosaicing. Many dem-
this way, a monochrome sensor needs less light to generate osaicing algorithms exist. The simplest method averages
an exposure equivalent to one made with a color-filtered the surrounding pixels to interpolate the missing data. This
sensor. Today, photographers have the option of buying method is applied to every pixel location, though rarely
cameras with monochrome sensors, sans color filtering. is such a simple approach used today. Most demosaicing
These are niche products, though, and often come with algorithms include additional strategies for identifying and
higher price tags in exchange for specialized functionality. preserving edges, weighting local brightness trends and
Now’s a good time to point out that you often hear pixel groupings. In addition to the CFA, an infrared blocking
photographers call the photosites pixels since we ultimately filter is fitted in front of the sensor due to the natural sensi-
get a photographic image file from a sensor’s capture tivity of silicon to infrared radiation.
comprised of pixels (the fundamental building block of a
digital image). However, it’s best to use photosites when
discussing the sensor, as there is not always one-to-one X-Trans Color Filter Array
relationship between the number of photosites and the Fujifilm developed an alternative approach that forgoes the
number of pixels. We’ll look at this challenge of color Bayer filter array for one of its own devising. First appearing
filtering the photosites and translation to color pixels a little in Fujifilm cameras in 2012, the X-Trans color filter array uses
later. a larger 6x6 pattern of red, green and blue photosite filters.

74
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

Figure 3-4 The traditional Bayer CFA pattern compared to the X-TransTM CFA developed by
FUJIFILM. Illustration courtesy of Jonathan Liles

Figure 3-4 shows that, unlike the Bayer pattern, there are of the CCD as it profoundly transformed consumer elec-
adjacent green photosites. This arrangement promises to tronics and photography. In the early days of digital
minimize moiré patterns (discussed later in this chapter) cameras, CCD sensors could be found in all types of form
and other artifacts inherent to color filtering photosites factors ranging from point-and-shoot to SLRs and medium
coupled with demosaicing algorithms. Real-world testing format backs. You might find CCD sensors in some low-
shows the X-Trans CFA to be extremely capable but not end point-and-shoot models today but their popularity is
necessarily superior to the Bayer CFA.1 The raw mosaicked waning in consumer camera hardware.
image file captured by an X-Trans CFA requires its own The CCD builds up a charge within each photosite that
special algorithm for demosaicing. is proportional to the amount of light that strikes it. The
CCD uses shift register electronics to move the charges from
their respective locations on the sensor array to circuitry
Camera Sensor Technologies along the edge of the chip. The electrical charges are fed,
pixel by pixel, row by row, to the analog to digital (A/D) con-
Imaging sensors are made of silicon and are semiconductor verter as illustrated in Figure 3-5 (overleaf). When the first
devices. Sensors convert photons of light that strike them line is completed, all the lines shift down and the process
into an electrical current. In this process, free electrons starts again. This method is efficient in that it only requires
are released from the material, which in turn produce the a single A/D converter through which all signals are pro-
electrical charge recorded and converted to a digital signal. cessed. On the other hand, the serial, pixel-by-pixel readout
There are several different sensor types that exploit this takes time and leads to an unavoidable delay between
behavior to record digital photographs. a photographic exposure taken, readout and saved and
when a new exposure can be made.
Astronomers were among the early adopters of CCD
Charge-Coupled Device Sensor technology. CCD technologies have a quantum efficiency
One of the earliest image sensor technologies is the of ~90%, meaning that they convert 90% of the light that
charge-coupled device, or CCD, invented in 1969 by Willard strikes them into pixel values.2 CCD photosites are particu-
Boyle and George Smith at AT&T Bell Labs. Boyle and Smith larly sensitive to ultraviolet and infrared radiation, which
received the 2009 Nobel Prize in Physics for the invention happens to be the makeup of many celestial objects’

75
 s ecti o n 1: Captu r e

Figure 3-5 A simplified illustration of a CCD sensor readout. Figure 3-6 A simplified illustration of CMOS sensor readout. Analog to
digital conversion happens at each photosite and that signal is sent to
readout electronics on the chip.

emitted radiation. CCDs produce low amounts of noise, Complementary Metal Oxide Semiconductor
especially when kept at cold operating temperatures using Sensor
special cooling devices.3 Lastly, CCDs have a linear response The complementary metal oxide semiconductor (CMOS)
to light which makes them excellent measurement devices. sensor came along after the CCD and, early in its existence,
exhibited worse image quality performance. The photosites
on a CMOS sensor are called active pixel sensors (APS) and
Sparse Color Filter Array Sensor consist of the photodetector, an amplifier and integrated
Kodak developed a special color filter array for CCDs called circuitry. This is a key distinction from CCD technology:
TrueSense. Also known as a sparse color filter array, the CMOS photosites independently convert their detected
design differs from the traditional Bayer pattern in that it voltages into digital signals and can be read out individ-
includes “white” photosites alongside the red, green and ually, all at once, instead of row by row as illustrated in
blue ones. These panchromatic photosites do not use a Figure 3-6. This allows for more complex behaviors like
color filter and instead collect all of the available light to triggering specific rows or groups of pixels to expose for
record brightness. Removing the color filter allows 66% durations different than that of the other photosites in the
more light to reach to the photosite. By including these array. Sometimes the CMOS sensor as a piece of electron-
color-indifferent collectors, the sensor is able to make use ics is confusingly described as an APS-type sensor, when
of increased luminance data for improved image quality the sensing is actually being done by many millions of
and light sensitivity. This technology is now owned by ON photosites.
Semiconductor.4 Since a CMOS sensor’s photosites use additional cir-
cuitry, gaps or trenches exist between the light-sensitive
areas. It’s also necessary to use barriers to separate each

76
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

Figure 3-8 A photograph of a CMOS sensor taken under a

Figure 3-7 A cut-away view of a CMOS photosite with a microlens.
microscope. The bright rings are reflections of the light source on the
curved surfaces of the photosite microlenses.

photosite to minimize crosstalk where photons land in one from blue light than from green or red; image process-
well but get recorded by a neighboring one (see Figure ing has to accommodate by boosting the blue-filtered
3-7). A notable amount of light is lost if the sensor design photosites’ signal. It also explains why CMOS sensors are
doesn’t account for these trenches. Consider collecting designed with infrared-blocking filters in addition to the
photons like collecting raindrops with buckets. Arranging microlenses on their surface. If left unfiltered, a great deal of
the buckets along the ground as close together as possible IR radiation travels to the photosites and records signal. The
helps collect the rain falling in your yard, yet there’s some optical filter prevents this energy from making it that far via
space between the buckets that fails to catch a percentage interference.
of the rainfall. One solution is to introduce large funnels One potential pitfall with smartphone cameras and sim-
atop the buckets to redirect the maximum amount of ilarly small form factors in regard to this IR-cut filter is the
raindrops to their collection vessels. Similarly, CMOS sen- possibility for color shading, an image quality artifact. Color
sors employ microlenses to help gather light energy that
otherwise strikes light-insensitive edges of the photosite
architecture. Each photosite has its own microlens directly
above it as shown in Figure 3-8. These lenses increase
sensor efficiency as they help capture a higher percentage
of the photons reaching the sensor plane. CMOS sensors
have a quantum efficiency of approximately 50%.
The architecture of a CMOS sensor helps it to maximize
light collection but its overall quantum efficiency is an
average across wavelengths and doesn’t consider the role
of the incident light’s wavelength. Figure 3-9 plots the
spectral sensitivity of unfiltered CMOS photosites across
the visible spectrum. Notice that there is relatively low
sensitivity around 400 nm (the blue region). On the other
end, there is an increasing sensitivity beyond 750 nm,
heading into the infrared region of the spectrum. From
this we understand that photosites record a weaker signal Figure 3-9 The spectral sensitivity of an unfiltered CMOS photosite.6

77
 s ecti o n 1: Captu r e

shading happens when IR light traveling at oblique angles material whose properties dictate that light is absorbed
(due to the short image distances rendered by wide-angle differently depending on its wavelength, making this
lenses in compact spaces) manages to transmit through approach possible. Foveon technology is described as a
the filter. Compact camera systems employ a combina- “direct imaging” sensor since it does not rely on interpola-
tion of IR filtering material and glass absorption filters; the tion. However, the latest incarnation, the X3 Quattro, isn’t
efforts of the two in combination ensures minimal IR trans- quite this simple. It uses a higher density of photosites at
mittance. Without this hybrid approach, these cameras can its blue layer (four times as many as the other two layers),
suffer from color non-uniformity between the center and meaning that additional image processing considers four
edges of their images.5 photosite data points for every two red and green at a
CCD was a superior technology choice in the early given location. Why does blue light get special treatment
years of consumer digital photography for price, low light when we often hear about green being the favored part of
sensitivity and overall image quality. However, advances the visible spectrum? Our recording material is less sensi-
in CMOS designs, manufacturing and their ubiquity in so tive to blue wavelengths and thus needs help in recording
many consumer devices flipped that relationship. It con- enough data to match what gets recorded at other wave-
tinues to be expensive and difficult to produce very large lengths. Sigma’s solution samples four times as many blue
CMOS chips. Still, the majority of imaging devices used by photosites, kind of like giving the scrawny one of your three
photographers today use CMOS sensor designs. kids extra glasses of milk. The Foveon-type sensor offers
advantages in color reproduction and sharpness, though
it’s not a clear-cut race with other options; its exclusivity
Backside Illumination Detectors to Sigma cameras means that it simply isn’t as common to
As discussed, CMOS sensors have circuitry at every pho- come across in the photography world.
tosite in order to independently read detected voltages.
The drawback of this arrangement is that the circuitry
occupies a certain surface area, lessening the area avail- Image Sensor Sizes
able for light-sensing. Microlenses above each photosite The size of the sensor is often dictated by the camera’s
help gather the light toward the light-sensitive well of the form factor and its size. Using a larger sensor requires more
photodiode. We gain additional light-sensitivity and effi- physical space in the camera’s body to house it and to
ciency by moving the electronics underneath the well. This support the necessary optics for forming an image large
is the premise behind backside illuminated (BSI) detectors, enough. Smaller sensors are easier to fit into compact
a type of CMOS technology that is particularly prevalent camera bodies but lack some of the image quality benefits
in smartphone cameras. BSI sensors detect more light of larger ones. Consider again the analogy of collecting
reaching them compared to conventional CMOS photosite raindrops in your yard. Arranging buckets over a small area
construction. of the yard is effective, yet arranging them over a larger
area offers the greatest collection potential.
When it comes to defining image sensor sizes offered
Foveon Sensor by camera manufacturers, think of it like choosing the size
Sigma has its own approach to recording color on a of popcorn at the movie theater (small, medium, large and
CMOS image sensor called Foveon currently found in their bucket). Perhaps the four options started out as standard
mirrorless cameras. The Foveon sensor uses three layers and uniform sizes across the industry. As time went on,
of photosites, one stacked on top of the other. Light of each movie theater decided to use their own unique bag
varying wavelengths travels different depths such that the sizes, making the size categories imprecise and con-
top layer records blue, the middle layer records green and text-specific. A small popcorn at one location could be the
the bottom, red. Incoming light travels through silicon, a size of a medium at another. In the sensor world you can

78
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

still get the equivalent of a bucket of popcorn but there are Sensors smaller than full frame yet designed for SLR
more sizes to choose from and most of us rarely need or bodies and lenses are called crop sensors because they
take advantage of it—a larger image sensor is not a neces- crop the image formed by the lens if all things are kept the
sity in many imaging applications. same. Crop sensors vary in their dimensions as manufac-
Common throughout the sensor size categories is the turers adopted their own standards, though they all use
moniker APS. This term originally started with a camera the shorthand of APS-C to indicate this size category. You
system developed by Kodak and released in 1996 (one of may see a sensor described by the crop factor (example:
the authors worked on the team that developed the color 1.5x) which describes the multiplier necessary to calcu-
balance algorithm for this system). APS stood for Advanced late the effective focal length resulting from its use with a
Photo System to describe the film format used. The film lens designed for full frame sensors. Crop factor is simple
included a magnetic strip along its edge that recorded enough to calculate:
information about each exposure as a form of capture
Diagonal of full frame
metadata. That information, such as shutter speed, aper- Crop Factor = (Eq. 3.1)
Diagonal of smaller sensor
ture, time of the exposure and print aspect ratio, was used
at the time of printing to improve quality. Kodak marketed Going smaller still, the Four-Thirds or 4/3 sensor was devel-
it under the brand name Advantix while Konica, Agfa and oped by Eastman Kodak and Olympus for a generation
Fujifilm marketed it under different names. This was a of smaller, mirrorless cameras with new lens designs to
major step forward in imaging technology. The APS acro- match. These tend to be about 30–40% smaller in area than
nym has since morphed into use to designate an active full frame sensors. There are also many sizes smaller than
pixel sensor, describing the modern silicon photosites that 4/3 used in modern cameras, particularly those found in
comprise the sensor’s light detectors. phones, tablets and laptops. As Figure 3-10 shows, these
Full frame sensors are the dominant choice of profes- sizes are quite small—some fractions of an inch in height
sional photographers as they offer exceptional image and width.
quality and a traditional form factor. The designation Sensor size naming conventions are decidedly confus-
comes from a legacy film standard: 24x36mm, the dimen- ing. There’s 4/3, 2/3, 1/3 and 1”, APS-C, APS-H and the list
sions of a 35mm film frame. This size also produces a 3:2 goes on. So where do these come from? It turns out that
width to height ratio. Though common today, the early these are left over from 1950s television standards. The
years of digital photography saw much smaller sensors due sizes are not defined by the TV diagonal as some assume;
to the cost and complexity of manufacturing sensors of this they refer to the diameter of the glass envelope of the TV
size. Most SLR-style cameras house full frame sensors with tube. The usable area of the imaging plane was approxi-
lenses to match. mately two-thirds of the tube.
Any larger than full frame and we enter medium format As we moved from film to small digital sensors, experi-
sensor territory. This, too, follows in the footsteps of film enced photographers quickly noticed that looking through
photography standards. Medium format cameras are
a larger form factor themselves, primarily in service of
housing such a massive light-recording material. Medium
format film comes in a variety of sizes but is commonly
more square in aspect ratio; image sensors in this cate-
gory follow suit. Medium format sensors offer some of the
highest megapixel resolutions available. Film photography
also has a category of large format sizes but we don’t see
many digital equivalents due to the cost and manufactur- Figure 3-10 A sampling of common image sensor sizes stacked
ing challenges. against one another.

79
 s ecti o n 1: Captu r e

Figure 3-11 The effective field of view is lessened when using a crop sensor relative to a full frame sensor. Here, the full frame system images the
tree in its entirety while the crop sensor system’s view is limited to a portion of it.

a full frame DSLR with a 50mm lens compared to a 50mm

lens on a crop sensor, the field of view looked different (see
Figure 3-11). The “normal” field of view expected with a
50mm lens appeared narrower. A lens produces a circular
image and yet a sensor is square or rectangular (see Figure
3-12). A full frame sensor captures a larger field of view (the
image area) relative to a smaller crop sensor at the same
location. The crop factor provides the effective focal length
of the lens. For example, a 50mm lens used on a camera
with a 1.6x crop factor has an effective focal length of
80mm.
Sensors also differ in their aspect ratio, the rectilinear
shapes made by varying width and height independently.
The traditional aspect ratio of still photography is 1:5:1
which is the same aspect ratio of full frame sensors and
35mm film before them. Medium format sensors are typi-
cally 1:33:1, more square than rectangular.
Finally, sensors are no longer constrained to a flat plane:
Figure 3-12 A lens projects a circular image at the imaging plane.
curved sensors where the edges curve outward toward
A sensor is sized and positioned such that it records image content
from edge to edge; a crop sensor positioned similarly to a full frame the lens are a forthcoming technology.7 This design tweak
sensor records a lesser field of view. means that light travels an equal distance from the lens to
both the center and the corners of the light-sensitive array.
From an image quality standpoint, this lessens light fall-off
and sharpness fall-off at the edges of the frame without
the need for expensive and complex optical corrections
in lenses.8,9 Curving the sensor plane simplifies the optical

80
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

components. It may even offer additional light sensitiv- The larger the photosites, the more light-gathering abil-
ity.10 Since new lenses have to be designed specifically for ity. A 50 megapixel full frame sensor produces an image file
this type of sensor, you won’t be upgrading your SLR to with the same number of pixels as a 50 megapixel compact
a model with a curved sensor to use with an older set of camera sensor but these two are not completely equiva-
lenses. On the other hand, it can make fixed focal length lent. Fitting 50 million photosites onto a compact camera
cameras embedded in a variety of electronics cheaper and sensor means designing similarly tiny photosites. Thus, the
lighter. Unlike the short-lived trend of curved televisions, light-gathering ability and consequently the image quality
the added dimensionality of a curved sensor offers measur- performance of the two sensors is unlikely to be equal. We
able quality improvements and not just a novelty shape. describe the size of photosites as pixel pitch and measure
their dimensions in microns. A sensor with a smaller pixel
pitch uses smaller photosites which in turn records lesser
Camera Sensor Resolution amounts of light at each location. Less light equals less
signal relative to noise, ultimately resulting in lesser image
Image sensor dimensions vary depending on the appli- quality, particularly in low light situations. Smartphone
cation or camera type but it’s not the only aspect that camera sensors have pixel pitches of approximately 1–2
characterizes a photograph. Spatial resolution describes microns whereas full frame SLR sensors have pixel pitches
the number of picture elements (pixels, in this case) that in the 5–8 micron range. The difference in light-capturing
comprise a photographic image. The more photosites on potential between the two is like the odds of you suc-
a sensor, the greater the image resolution recorded. Just cessfully hitting a ping pong ball with a ping pong paddle
like the convenience of describing a plank of wood by its versus a tennis racket. More surface area means more
dimensions (i.e. “2x4”), describing the spatial resolution opportunities for successful contact between ping pong
of a sensor is accomplished by counting the number of ball and racket in the same way that it means more oppor-
photosites in the horizontal and vertical dimensions (i.e. tunities for photons of light to strike the photosite well.
5,632 x 3,750). These numbers don’t roll off the tongue or
offer easy commitment to memory. Instead, multiplying
the horizontal dimensions by the vertical gives us the total Image Bit Depth
number of pixels (21,120,000, for example) which we then Pixels in an image have numeric values. The range of these
shorten to an easy-to-remember value like 21.1. This value values is dictated by bit depth. Since all digital data is built
describes the megapixels or the number of millions of by bits, binary values of either 0 or 1, a greater number of
pixels. bits means a greater number of possible combinations
available to describe a pixel’s value. Bit depth defines the
range of potential values that a pixel describes sort of like
Pixel Pitch knowing that a gallon jug of milk contains some amount of
Spatial resolution tells us the number of detectors but milk anywhere between zero and one gallon. A gallon jug
doesn’t specify that number in the context of a physical of milk doesn’t necessarily have to hold a gallon, it dictates
area. If an airline announces that its popular flight routes the range of milk it could hold. A sensor’s bit depth tells us
have an additional 50 available seats yet they continue how many bits are available to describe any given pixel in
to use the same aircraft, you’d suspect that they lessened an image at the time of capture.
the legroom to cram in additional rows. A bigger number What amount of bit depth is needed or desired when
doesn’t guarantee a better, or even equivalent, experience. making photographs? The leftmost version of the photo-
The same logic applies to sensor resolution: the amount graph in Figure 3-13 (overleaf) shows a bitonal image, that
detectors matters but it should always be considered is, an image constructed from only two possible tones:
against the physical dimensions of the array. black or white. This is a black and white photograph in

81
 s ecti o n 1: Captu r e

Figure 3-13 A 1-bit version of a scene (left) and an 8-bit version (right) that has up to 256 shades of gray to represent the scene.

the most literal sense. A 1-bit representation of tone is This only describes a photographic image in shades
good enough for text and simple graphics and not nearly of gray. Where does the color come into play? A color
enough for a true-to-life photograph. The rightmost ver- image is actually three sets of pixel data: one for red, one
sion of the photograph in Figure 3-13 uses 8-bits, or 256 for green and one for blue. These sets are called channels,
gray levels, to more closely represent continuous tone as and color photographs always have three. In order to have
we see it with our eyes. 256 possible brightness values—the standard bit depth
The more possibilities for how bright or dark a pixel can established for grayscale images—for each color channel,
be, the more exact our recording becomes and the more the total amount of bits (8-bits x 3 channels) is 24. Combi-
gradual and natural the transitions between brightnesses nations of values in each of the three color channels yields
appear. The bit depth of an image describes the amount of a wide range of observable colors: 16.7 million, in fact. This
tones (gray levels) represented as shown in Figure 3-14. number is calculated by raising 2 to the 24 power (224).

Figure 3-15 A full color photograph made from 8-bits of data in each
Figure 3-14 Gray levels in an image per number of bits. of the three color channels, or 24-bits in total.

82
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

Figure 3-15 uses 8-bits per color channel, or 24-bits in total, system. This filter introduces a slight blurring to the image.
to create a full color photograph. This may make you shout “why on Earth?” since it seems
Confusingly, a full color image file is often referred to as to undermine all the hard work our optical components
8-bit even though we understand that, taken literally, this do to resolve a sharp image. We run into some trouble
makes a pretty ugly looking photograph. The confusion is when details in a scene are so small and fine that they hit
abated by specifying if that number describes bits per chan- the resolving power limit of our optics. This is exacerbated
nel (which it does) or bits per pixel (which it does not). The by the demosaicing algorithms that struggle to accurately
images seen regularly on websites have bit depths of 8 bits interpolate detail edges and patterns. In essence, the sub-
per channel. Importantly, photographers have the option ject detail exceeds the detail our camera system records.
of shooting and editing images in formats that support up The visual artifact that results in this circumstance is called
to 16-bits per channel. Recent video formats and playback moiré and looks like wavy-lined patterns not actually
support mean that traditional, 8-bit video will soon be sup- present in the scene (see Figure 3-16). Aliasing is similar in
planted by 10-bit encoded video to extend the dynamic
range of video content.
Research shows that 5-bits per color is typically enough
information to represent continuous tone and natural
color gradations in both display and print output. It is also
unlikely that the average viewer on a given display detects a
difference between, say, a 7-bit image and an 8-bit one. This
will change as displays evolve in their capabilities; for now
we understand that 8-bits is the appropriate volume of bit
depth for most applications of sharing and viewing images.
In the context of the camera sensor, the capture bit depth
is an inherent and static characteristic. It describes the total
range of values that can be detected and recorded when
photons strike the silicon material and produce an electric
signal. Many sensors record in 10- or 14-bits. The bottleneck
isn’t the photosites, it’s the analog to digital converter. A
cheaper A/D might limit the nuance of brightnesses trans-
lated to digital signals even if the photosites capture a larger
range of voltages. This information is stored in a raw file
format that’s adjustable later down the line when process-
ing and exporting photographs on the computer; not all file
formats support a bit depth as great as that captured by the
camera. Higher bit depth information must be downsam-
pled, often with minimal visual degradation, to exist in these
other formats. Chapter 8 offers more on image file formats.

Aliasing and Moiré Patterns

Earlier, we showed that the image sensor is covered with Figure 3-16 Moiré pattern artifacts captured by a camera
lacking a low-pass filter. The fine detail of the suit fabric is
color filters and an IR-blocking filter. It’s also common to not faithfully reproduced after capture and demosaicing.
have an optical low-pass or anti-aliasing filter in the imaging Photograph by Hieu Pham

83
 s ecti o n 1: Captu r e

that it renders straight, continuous lines as stair-stepped

and jagged due to a lack of resolved detail. These artifacts
are unavoidable if left alone but are negated if the image is
slightly blurred right before it reaches the photosites. That’s
the job of the optical low pass filter: a slight bit of blurring
is the tradeoff to avoid unnatural and distracting artifacts.
Low-pass filters were typically installed on every con-
sumer digital camera on the market. Today, photographers
have the option of buying cameras with sensors that Figure 3-17 Aperture shapes can vary with lens design. The left lens
specifically omit the low-pass filter. At the risk of recording is stopped down to a high f-number with the aperture blades forming
a heptagonal opening. The right lens is set to a low f-number with an
moiré patterns, this translates to sharper images right out
opening approximating a circle.
of the gate. The Achilles heel of such cameras is clothing
and fabrics. These subjects often feature fine details, tex-
tures and patterns that threaten to introduce significant Setting the aperture means changing the size of the
moiré artifacts. The primary way to minimize the artifacts, opening. The mechanism that accomplishes this is the
assuming they’re noticed during a shoot, is to change the aperture diaphragm, a set of overlapping blades or petals
subject distance from the camera. Otherwise, it’s diffi- that form a circular opening at their center pictured in
cult to mitigate the artifact without significant, manual Figure 3-17. The range of diameters of an adjustable aper-
retouching. ture diaphragm is lens-dependent and has implications
As sensor designs continue to push more photosites for how that lens may be used (see Chapter 4). Traditional
into small areas, we approach the Nyquist frequency. The camera apertures approximate a circular opening. They
sampling Nyquist theorem states that the minimum rate may also appear as hexagonal or octagonal with flat-edged
for sampling a signal without introducing errors is twice sides. The discrete positions of the aperture’s overlapping
the highest frequency present. With light and images, this blades are called f-stops.
frequency is related to object size. Small objects and fine The aperture shape dictates the shape of a point of light
details in an image are considered high frequency content formed on the image plane; the closer this shape is to a
or signal. The imaging sample rate or pixel size, then, must circle, the closer small resolved points of the image are
be half the size of the smallest object or detail recorded to rendered as circular points. This is not the only way to form
avoid visible artifacts. Moiré and aliasing artifacts are often a an image but it’s the most common and visually pleas-
direct result of having undersampled the scene. ing approach. Creative visual effects are possible using a
uniquely shaped aperture opening. This shape could be a
star, an amorphous blob or any other geometric, two-di-
Apertures mensional form. All form a usable photographic image, yet
their differences become apparent in out-of-focus areas,
The aperture is the opening at the center of a lens through particularly specular highlights, that take on the appear-
which light passes. The concept of an aperture is fun- ance of the aperture shape.
damental to the conception of the camera. The camera The aperture diaphragm mechanism quickly adjusts
obscura, the precursor to photography, is effectively an the size of the opening with automatic, electronic pro-
aperture opening allowing light into a dark room. The gramming. A key behavior of electronically controlled
larger the opening in a lens, the more light passes through apertures is that they often stay at their maximum opening
to the light-sensitive recording medium. A smaller opening until the shutter button is pressed. In the milliseconds that
means less light but potentially greater resolved detail and follow, the camera electronics tell the aperture diaphragm
in-focus scene content. to move into the predetermined position—say, f/8.0, as

84
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

dictated by the user manually or from automatic exposure back over the lens. Cheap and easy but not at all con-
calculations. The shutter opens immediately following this sistent. Those early cameras used film that was relatively
change to create an exposure. This is notable because it insensitive to light and therefore a photograph’s expo-
means that what the photographer sees in the moments sure was measured as a function of minutes. Today we
before pressing the shutter button will likely be different use exposure times on the scale of fractions of seconds.
from the recorded image. Even the quickest-draw lens capper can’t nail a 1/8000th
SLR cameras typically have a depth of field preview exposure time. Instead, our cameras house shutters that
button that, when held, temporarily changes the aperture control when and for how long light reaches the image
diaphragm to the position used during an exposure. This sensor. These shutters operate mechanically, electroni-
darkens the observable image through the viewfinder or cally or a combination of the two for a duration called the
the electronic live preview since it’s stopping down the shutter speed. The physical button on the camera body
aperture from its maximum opening. The main benefit, that signals a shutter action is the shutter button or shutter
as the name implies, is in previewing the depth of field release, though other triggering options include sound,
of the about-to-be-captured exposure. A fully manual timers or even face detection via software algorithms
lens or an analog camera inherent provides this behavior (only capturing when the subject smiles, for example).
since the user directly adjusts the aperture setting when The physical shutter button is most commonly designed
composing. with two stages of travel, meaning that it can be pressed
Smaller cameras including smartphones use fixed aper- down halfway to indicate a forthcoming exposure trigger
tures. These are typically large openings like f/1.4 or f/2.8 to or to lock a focus position or light meter reading. Pressing
keep exposure times short and to maximize light detection it down completely initiates the shutter action to capture
on their relatively small sensors. A few recent smartphones an image.
introduced variable apertures that adjust based on light
level; these mechanical components are challenging to
build in small form factors. Leaf Shutters
Lenses designed for motion video or cinema cameras A leaf shutter is a set of overlapping metal blades (also
use a unique aperture design called transmission stops called leaves), much like the lens aperture. The difference
or t-stops. These are precisely calculated to control light is that the leaf shutter blades close down completely; aper-
transmission in video exposure and allow for continuous ture blades always allow for a small opening at their center.
adjustment from the widest to the narrowest aperture When triggered, a leaf shutter opens and dilates from its
opening. This is notably different from a traditional still center, exposing the image plane to light as it eventually
camera whose f-stops are discrete and abrupt steps in expands beyond the sensor area. Figure 3-18 (overleaf)
aperture blade position. The t-stop design is critical for shows a sequence of leaf shutter positions from closed
motion video work in creating smooth, gradual changes to to completely open. After the length of time set by the
the aperture setting during recording. photographer for shutter speed, the blades close until the
image plane is once again blocked from all light.
The primary advantage of a leaf shutter is in its ability
Shutters to expose the entire sensor at once: if a short-duration
flash is used, say, in a studio setting, we know that the
All photographs require that the image sensor receives whole frame receives that flash as part of the exposure.
exposure for a short period of time. Back in the earliest Put another way, leaf shutters are flash-synchronous at any
days of photography, this was achieved in the simplest of speed. This is put to great use when mixing flash and ambi-
ways: a photographer removed a lens cap from the lens, ent light for on-location shooting. It’s easier to work out
counted out the exposure time and then placed the cap functional exposure times that get the best of both artificial

85
 s ecti o n 1: Captu r e

Figure 3-18 A leaf shutter in stages of opening as seen in a medium format lens. Photographs by Rochester Institute of
Technology photography alumnus Dan Wang

and natural lighting, especially under bright sunlight. Shut- delay of the second curtain that both remain out of the
ter speed is measured as the duration of time that the leaf way for a time, leaving the entire sensor exposed.
shutter remains completely open. The downside to using a moving slit to expose the
Leaf shutters are not common in consumer cameras sensor reveals itself when using an electronic flash. Its
and are typically only found in professional medium and duration is many times shorter than that of the shutter
large format cameras. There are exceptions to everything, curtains and a poorly synchronized exposure could see
of course, and a few select mirrorless cameras feature the flash exposing only a small slit of the sensor while
such shutter designs. One notable shortcoming of the leaf the rest receives only ambient light. This creates partial
shutter design is the speed ceiling: they cannot achieve frames as shown in Figure 3-19. Cameras tend to include
the fast shutter speeds that other methods offer. Another an automatic shutter speed adjustment when they’re
shortcoming is cost. Leaf shutters must be placed in the attached to electronic flash sources to avoid this problem
lens, rather than the camera body, meaning that each lens and to give the photographer one less thing to have to
is more complex and expensive to make and you pay for remember.
it each time you buy a lens rather than once as part of the
camera body cost.

Focal Plane Shutters

A focal plane shutter is a set of curtains that travel in
tandem forming a slit opening in front of the image
sensor, usually from top to bottom. As one curtain moves
downward, the second curtain waits some fraction of
a second before following, leaving a small, horizontal
opening through which light reaches the focal plane. The
curtains then move upwards to reset their position, this
time without leaving a gap between them. The amount of
time set for the shutter speed determines how much time
the sensor is exposed to light before that second curtain Figure 3-19 A partial frame due to an improper focal plane shutter
blocks it again. Longer shutter speeds have a long enough speed used with electronic flash.

86
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

Electronic Shutters
So far we’ve described physical camera components that
block light from reaching the sensor save for a brief instant
when they move out of the way. This is the only way to
control exposure time in analog systems because film
reacts to light the moment the two are introduced to one
another. A digital image sensor isn’t so reactionary. It only
pays attention to light when it’s switched on and recording
photons. If we leave it switched off, it pays no mind to any
incident light. It follows, then, that flipping the electronic
switch on and off for a short time has a similar effect as a
physical, light-blocking shutter would.
CCDs aren’t a great candidate for the electronic shutter
approach as they can’t turn on and off as rapidly as typical
photographic exposures require. CMOS, on the other
hand, can be turned on, record light and then turned off
in extremely short durations. Since CMOS photosites are
self-contained units with their own circuitry, we can go one
step further and command specific photosites or rows of
photosites at will. Thus, a simulation of a focal plane shutter
is created by triggering the topmost line of sensor pho-
tosites to record, then the next line, the next, and so on,
turning each trailing line off in sequence. If we’re really feel-
ing crazy, some sensors and accompanying programming
allow for multiple exposures to be recorded at effectively
Figure 3-20 The visual artifact resulting from a rolling shutter with fast
the same time: every even photosite turns on for one
camera movement (top) and an undistorted view of the same scene
duration and every odd photosite exposes for a different (bottom).
duration, for example. Handling these disparate exposures
becomes a challenge for processing yet it offers a unique
opportunity in simultaneous capture that is not available
with analog methods. lines are recorded. The rolling shutter artifact that results is
The common electronic shutter method used by CMOS described as jelly-like: solid objects like buildings or cars
sensors is called a rolling shutter which records and reads appear to warp, undulate or skew. Figure 3-20 shows one
out exposure line by line rather than with the entire sensor such example of an aerial imaging system that was moving
area all at once. Each row receives the same amount of quickly as the sensor exposed the scene using a rolling
exposure; the row exposures are offset in time. Reading shutter. The larger the sensor (in terms of number of pho-
each line in sequence allows the sensor and accompany- tosites and number of rows), the longer it takes to read out
ing electronics to process the huge amount of light data exposure from top to bottom. Higher-end cameras used for
recorded in video capture and avoids the need for a phys- video recording in particular use CMOS with global shutters.
ical shutter in the small form factor. This can be a problem Global shutters capture all sensor rows at the same time. As
when filming fast-moving subjects. The subject may be at mobile processors become more powerful, we expect to
one location when the top lines of the sensor are recorded see global shutter methods incorporated into smaller and
and at a very different location by the time the bottom cheaper camera systems.

87
 s ecti o n 1: Captu r e

Electronic first curtain shutter is a technique that uses the scene through the viewfinder while rotating a ring on
both mechanical and electronic behaviors. An exposure the lens barrel, trusting our eye to set the camera focus to
begins when each successive line of the CMOS sensor is the sharpest position for the scene. There is no question,
switched on to detect photons and a mechanical curtain though, that automatic focus is convenient and relied upon
moves to cover up the sensor to end exposure. This allows by many to get photographs in sharp focus. The basic
for faster exposure times using a partially electronic shutter, functionality behind automatic focus is described in the
reducing in-camera vibration and shutter noise. following paragraphs. Note that today’s camera systems
increasingly layer strategies with additional algorithms to
infer subject priority or anticipated motion. Face detection,
Automatic Exposure and Metering for example, identifies likely human faces in the frame and
ensures that the face is in focus above all other subjects.
Cameras determine the amount of light required for pho- There are two components required for a system
tographic exposure in one of two ways. They evaluate the offering automatic focusing. The first is mechanical, as
light hitting the sensor (by opening the shutter), effectively the internal lens elements must be electronically control-
taking photographs that aren’t permanently saved. This lable to adjust focus. The second is either a combination
means using the imaging sensor as a light sensor: they’re of hardware and software or exclusively software that
one in the same. The downside to this method is that the determines how the lens must be adjusted to achieve
camera must take initial guesses as to a correct exposure. optimal focus. For traditional, interchangeable lens
If it’s way off, it must continue to guess and check until it systems, mechanical focusing is accomplished with small
zeroes in on the amount of available light. motor drives in the lens housing. There’s no room for
The second approach involves a dedicated light sensor gear-driven motors in smaller camera form factors. Instead,
inside the camera body. This is called through the lens (TTL) these cameras use voice coil motors that rapidly move tiny
metering as it measures light coming into the camera lens elements using electromagnets and applied voltage.
to determine exposure (refer back to Figure 3-1). A basic The physical movement of lens elements and the whirr
TTL meter offers three modes: matrix, spot and center- of the motors introduces both speed and noise consid-
weighted. In more advanced cameras, the TTL meter can erations when evaluating lenses for a photoshoot. Video
be tied to the autofocus mode. The two behaviors work recording specifically demands quiet motor drive systems
hand in hand to create the proper exposure for the area if the audio recording equipment is anywhere near the
of interest in a composed scene. TTL metering takes into lens; sports photography demands the quickest possible
account stray light entering the lens, the angle of view adjustments to achieve focus with fast-moving subjects.
and close-up exposure corrections that handheld meters All camera systems with autofocus capabilities use the
cannot. On the other hand, it’s only capable of reflective strategies of phase detection, contrast detection or a com-
light readings by virtue of its location in the camera and bination therein.
position relative to the subject. This fact can be limiting
with very bright or dark subjects. Through the lens meter-
ing is discussed further in Chapter 4. Phase Detection Autofocus
Phase detection autofocus takes a beam of light entering
the system and splits it into two beams. Each of these new
Automatic Focus beam strikes one of two adjacent photosites. The signals
detected on these photosites are compared. If they are
Manual focus is nearly always an option on professional the same, they are in phase and the image reaching the
cameras and many prefer to trust their hand-eye coordina- sensor is in focus. If they are not the same, or out of phase,
tion to adjust the lens moments before capture. We view the camera focus point is adjusted and the process is

88
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

repeated until the correct focus point is located. This phase in at the highest contrast. This strategy is slower than phase
detection information is used to calculate subject distance detection and does not provide decisive guidance to the
and is fast at getting the camera system focused. The lens focusing motors to know which direction to move.
out-of-phase detection information is explicit in indicating This may lead to focus seeking where the system adjusts
front- or back-focusing, further speeding up the process of focus distance all the way in one direction before doubling
finding correct focus. back and trying in the other direction before settling. Focus
Phase detection often occurs on a separate sensor seeking can easily lead to missed shots, particularly with
within the camera. For every focus point in the viewfinder fast motion events, as it may take entire seconds before
(often indicated by overlayed dots or boxes), there are getting the scene in focus. Contrast detection autofocus
two dedicated photosites that perform phase detection is well-suited for studio still life shooting as it’s unlikely to
focusing. Manufacturers sometimes advertise the focus encounter constantly changing focus needs. It may also
point patterns of these detectors as a means of indicat- be the only option in low light conditions where a phase
ing effectiveness. Phase detection can also occur on the detection system’s signal is too noisy to rely on. Some
main sensor by splitting designated “focusing pixels” in hybrid systems exist that use the best of both strategies:
two using microscopic masks over photosites. This system phase detection quickly hones in on an approximate focus
allows phase detection focusing to be used when the distance, then contrast detection fine-tunes the exact
camera is in “live view” as well as when composing through setting.
the viewfinder. This approach is increasingly common on
small camera form factors including smartphones because
it adds minimal complexity to their design. However, the Active Autofocus
signal must be clean enough to make accurate measure- Phase and contrast detection methods work by measuring
ments and camera systems may struggle to leverage phase the light coming into the system, functioning as passive
detection autofocus in low light conditions. detectors of image focus. Another approach is to actively
emit a signal, either infrared or ultrasonic, that reflects off of
the scene and back to onboard sensors to measure subject
Contrast Detection Autofocus distance. The camera measures the time it takes for the
Contrast detection autofocus is used most commonly in signal to return, calculates it against the speed of the signal
compact cameras, mirrorless cameras and when shoot- and drives the autofocus mechanisms to focus the lens at
ing in live view modes on SLRs. The image sensor itself is the sensed subject distance. This is called active autofocus.
employed to detect contrast and determine optimal focus Active autofocus sensors can work in complete darkness
rather than having a separate, dedicated sensor in the but are more effective at close ranges as the emitted signal
camera body. A major upside to this approach is that the takes longer to return and is more likely to scatter with
full resolution image sensor is leveraged to look at any and farther subject distances. They can also fail if the subject
all areas of the frame, if desired, to find sharp focus. Phase absorbs the emitted energy rather than bouncing it back.
detection sensors are typically lower resolution and only Additionally, infrared active sensors can also be confused
look for focus points in specific, patterned areas. Using by other heat energy sources in a scene.
contrast detection autofocus means having thousands of
spatially sampled points when phase detection might have
dozens. Autofocus Points
Contrast detection evaluates the contrast of the initial Similar to auto exposure techniques, cameras can use
image using image processing algorithms. The highest either a single autofocus point or multiple points to aver-
contrast is achieved when the image is in focus. The lens age or calculate a best-guess estimate for focus distance.
moves back and forth while evaluating contrast and settles Autofocus points light up in the traditional SLR’s viewfinder

89
 s ecti o n 1: Captu r e

photographers enjoy the level of control afforded by

toggling between different focusing behaviors depending
on the shooting context; entry-level cameras may bury
their selection in menus while higher-end cameras include
dedicated hardware buttons or switches.

Image Stabilization

Image stabilization (IS) counteracts camera shake that

otherwise leads to blurry images. Excessive camera shake
Figure 3-21 An illustration of the autofocus point overlay seen means that the image is rapidly changing and moving
through an SLR viewfinder. around. Sometimes the shake comes from the photogra-
pher while other times it’s introduced by the environment
or shooting platform. Fast shutter speeds are somewhat
when the shutter button is pressed halfway (see Figure immune to this movement because they expose for such
3-21). Depending on the manufacturer and model, these a relatively brief moment of time, successfully freezing a
points may appear in different patterns or shapes and sharp image at capture. It becomes harder to keep the
light up to indicate subject detection. They also appear camera steady as focal length increases due to its weight
as a graphic overlay on the preview screen when using and its field of view; the longer the lens, the more suscepti-
live view. SLR models featuring touchscreens allow the ble your images are to camera shake. There is also a direct
photographer to tap a location in the image frame to relationship between the amount of camera shake relative
direct autofocus. This is particularly helpful when there are to the amount of caffeine consumed by the photogra-
objects at multiple distances from the camera. pher. A low-tech image stabilization technique involves
Detecting movement and focus in different zones of controlled breathing, keeping the camera close to your
the image frame allows for subject motion prediction. For body and keeping shutter speeds as fast as possible. This
example, if the system detects a subject at the left edge of is a functional approach until light levels or other expo-
the frame moving toward the right, it may opt to lock focus sure considerations paint you in a problem-solving corner:
for this area rather than at the frame center, which might sometimes there’s simply not enough light to get a good
otherwise put the focus farther back. Using autofocus with exposure at a fast shutter speed. This is where we look to
moving objects takes practice to achieve good results for other solutions for image stabilization.
every exposure.
Finally, focus behavior is further specified by setting the
camera to either single servo focus or continuous servo. In Lens-Based Optical Image Stabilization
the former, the lens focus distance locks when the shut- Optical image stabilization (OIS) allows for slower shutter
ter is pressed halfway. The camera assumes that subject speeds while still getting sharp images. In lens-based OIS,
position remains static until the shutter is fully pressed and a single floating lens element is positioned within the
a capture is recorded. In the latter behavior, pressing the compound lens. This lens shifts to compensate for the
shutter button halfway down causes the autofocus mech- camera (and therefore image) movement. It makes sure
anisms to constantly evaluate and adjust focus position. that the projected image is stable by the time it reaches
This active, responsive behavior is excellent for catching the sensor. Disregarding the technical merits of the anal-
proper focus with quick-moving subjects but it requires a ogy, lens-based stabilization is like really good suspension
constant power draw to run the lens motors. Experienced in a car where your sharp image is the comfy passenger

90
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

on the bumpy road of camera shake. The downside is that eliminated some of the physical components moving
the approach requires additional components built into around immediately preceding exposure. Electronic first
the lens, making it an expensive proposition with a camera curtain shutter, mentioned earlier, is another available
system that encourages swapping lenses. Each lens must setting for minimizing vibration.
bring its own OIS solution to the table. There’s a price pre-
mium for lenses with built-in OIS relative to ones without it.
Electronic Image Stabilization
Lastly, software processing can offer electronic image
Camera-Based Image Stabilization stabilization (EIS). While it cannot rely on physical hardware
If we want to pay for a stabilization technology once to counteract or minimize physical camera movement, it
when buying the camera instead of every time we buy a works with the captured pixels to reduce blur. Still pho-
new lens, camera-based image stabilization is the answer. tographs benefit from merging or fusing multiple, rapid
Here, the sensor moves to compensate for shaky camera captures to achieve sharpness. Recording video at higher
movement instead of a lens in front of it. This way, even resolutions than the final intended capture, a technique
with an interchangeable lens system, every setup used can called overscan, is valuable for realigning consecutive
take advantage of the image stabilization. The degree of frames to correct for temporal camera shake. Advanced
correction or movement compensation may be less than video editing software can additionally offer frame warping
that which is achieved with a lens-based solution, however. to minimize distortion or framing changes despite camera
Both lens- and camera-based image stabilization strat- movement. Electronic image stabilization is ultimately
egies can be damaging to the equipment if left on when limited by the quality of its input. The best strategy is
shooting with a tripod. The system may attempt to correct to eliminate the effects of camera shake as early in the
for movement when there is none and puts wear on the image-making pipeline as possible for best results.
moving components. It may also lead to more camera
shake, ironically, as it introduces vibrations and image
movement artificially as it struggles to operate. Both forms Camera Viewfinders
of image stabilization are also power-hungry and drain
batteries faster than non-image-stabilized setups, and so Composition is a key skill for photographers. While crop-
they should be toggled on only when needed. ping at the post-processing stage is an option, we rely
Aside from photographer-induced shakiness, a small on getting most of the way to a perfect shot using a
amount of vibration is introduced by the mirror-flipping viewfinder. Holding the camera out at arm’s length and
mechanisms in SLR cameras. It may not feel like much but shooting wherever we happen to be looking is simply too
at longer focal lengths and at close working distances, the dependent on sheer luck; instead we prefer to see as the
vibration can very quickly reduce image sharpness. Mir- camera sees and commit to a capture when everything
ror-lock up is an option typically found in camera firmware aligns. The viewfinder is critical for framing, timing, focus,
that lessens the effect of mirror vibration during exposure. exposure determination—pretty much everything that the
In this mode, one press of the shutter flips the reflex mirror photographer needs to see and know. There are several
into position for capture and the sensor curtains remain methods of providing such feedback in camera design,
closed. Pressing the shutter again opens the curtains for either optically or electronically.
the determined shutter speed duration, followed by the
mirror flipping back to their default position. In between,
there is no image redirected to the pentaprism or view- Optical Viewfinders
finder. This means you’re temporarily unable to see what The SLR camera’s design uses an optical viewfinder where
you’re photographing with the benefit being that you’ve we place our eye to see the image as rendered through the

91
 s ecti o n 1: Captu r e

camera optics. The image seen is right-side up in orien- optical mechanisms became internalized in the camera
tation (recall from Chapter 2 that a lens forms an image design for easier focusing.
that is upside-down and backwards). The SLR camera’s
reflex mirror sits at a 45° angle and directs the light coming
through the lens to a ground glass above it. The image Electronic Viewfinders
is formed on this ground glass, passes to the pentaprism Electronic viewfinders (EVFs) take a live feed from the sensor
which re-inverts the image via mirror reflections. The and transmit it to a miniature screen in the camera body,
optical viewfinder goes dark and the photographer’s view viewable only when the user puts an eye up to the open-
is momentarily obstructed during exposure as the light is ing. This is not necessarily a strict, unedited video feed, as
redirected to the sensor. the image can undergo additional processing in real time
A notable limitation of SLR optical viewfinders is their to boost brightness, introduce software filter effects or
reliance on available light. A dark scene viewed through otherwise overlay relevant metadata. Early implementa-
the viewfinder is at least as dark, likely darker, than how it tions were low resolution, laggy and not particularly good
appears to the eye. Light loss through the optics and any replacements for optical viewfinders. This is changing
attached lens filters results in a dim preview. This is further quickly as microdisplay technologies evolve with higher
exacerbated by the aperture setting. A stopped down aper- pixel densities and faster refresh rates. EVFs are advan-
ture significantly cuts down the available light for real-time tageous in bright light, outdoor conditions because the
viewing even if the eventual exposure uses a long shutter space formed between the eye and the display blocks
speed or high ISO to compensate. The solution that camera extraneous light. They also work to keep form factors small,
manufacturers settled on involves leaving the aperture eliminating the need for mirrors and alternate light paths
wide open until the milliseconds before the shutter opens in the camera body. The downside comes in the form of
for exposure. We must explicitly command the camera to power consumption: using the viewfinder means powering
adjust the aperture to the desired diameter when compos- that small display and running the sensor continuously.
ing through the viewfinder. This is typically accomplished Where optical viewfinders are effectively passive and
with a depth of field preview button. Otherwise, the view- require no battery power to help you compose, EVFs need
finder preview does not accurately reflect the depth of field a steady flow of power from the battery pack.
that results in the final capture. Alternatively, the live view continuous video feed from
An honorable mention here is waist-level viewfinders, the sensor can be piped to a relatively large screen on the
most commonly encountered on analog medium format back of the camera. SLR cameras offer this functionality in
cameras. Forgoing the pentaprism (and therefore yielding a addition to their optical viewfinder. Compact point and
backwards image), this unique viewing mechanic projects shoot digital cameras exclusively rely on this approach.
the lens image onto a ground glass along the top facet of Using live view on an SLR necessitates locking the reflex
the camera. Holding the camera at waist height allows the mirror in its upright position to direct all incoming light
photographer to look down at the ground glass, some- directly to the image sensor. This behavior eliminates some
times shrouded in a foldable, pop-up hood, to focus and camera shake and shutter noise when it’s time to take the
compose. still photo capture. However, it also means bypassing the
Another optical approach is found on rangefinder-type autofocus and light meter sensors. The camera instead
cameras. A rangefinder uses two lenses that place the relies on signal detected at the image sensor and a layer of
two images into a single eyepiece using prisms. When the image processing algorithms to evaluate focus and expo-
camera lens is in focus, the two images are in alignment sure (using techniques like contrast-based autofocus and
and the distance or range to the object is determined. edge detection).
Once known, the range is used to set the focusing mech-
anism on the camera. In later rangefinder designs, these

92
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

Electronic Flash illumination output but a brighter flash requires a larger

energy storage capacity. On-camera flash employs small
Sometimes sunlight isn’t available or enough for making capacitors in the camera fueled by the camera’s battery,
photographs. At the same time, artificial lights indoors, limiting voltage and number of discharges. We trade flash
such as lamps and ceiling fixtures, are not necessarily bright output capacity and size for convenience.
enough for the camera sensor and lens combination. Another challenge with on-camera flash is red eye in
Electronic flash offers bright illumination with extremely portrait photos (see Figure 3-22). Why does this happen?
short durations as described in Chapter 1. Flash is typically It’s common to use flash in low light situations where our
triggered by the camera shutter press and lasts just long pupils naturally dilate for scotopic vision (more on this
enough for photographic exposure, appearing only briefly in Chapter 13). On-camera flash is very close to the opti-
to our eyes. We mentioned flash synchronization earlier in cal axis of the lens and sensor. When the flash fires, light
this chapter with regard to shutter types; using electronic bounces off the fundus at the back of the eye and reflects
flash limits the available shutter speeds with focal plane back to the lens. The red color comes from the choroid
shutters found on SLRs. You may even notice that, when layer of the eye which provides the blood supply to the
flash units are connected to your camera, the camera retina. Since your subject’s pupils are likely to be dilated in
programming automatically adjusts shutter speed (often a low light environment, the reflection is large and noticea-
defaulting to 1/125th second). ble. The only way to completely avoid red eye in portraits is
Many SLR cameras include a built-in or pop-up flash to use off-axis flash. Some camera’s offer red-eye reduction
unit or the ability to add an on-camera flash unit via hot modes that introduce a lower-intensity pre-flash before the
shoe connection. There are drawbacks to these smaller full flash output and camera capture. This first burst of light
light sources despite their convenience. The first issue with causes your subject’s pupils to constrict prior to taking the
built-in flash in particular is the small size. Recall from light’s photograph. The smaller pupil size minimizes the appear-
inverse square law that light falls off with the distance ance of the red reflection off of the fundus. It does not
squared. The effectiveness of the built-in flash drops off entirely eliminate red eye.
sharply as the distance to the subject increases and since Using a dedicated flash unit mounted to the camera
it’s always going to be at the location of the camera, its offers many more creative options for light direction and
illumination ability is not effective when the subject is more quality. Aiming the flash head allows for bouncing the
than a few feet away. This can be countered by higher light off of the ceiling or wall to reduce harsh shadows.

Figure 3-22 An on-camera flash often causes red eye. Photograph by Rochester Institute of Technology photography
student Genae Shields

93
 s ecti o n 1: Captu r e

Placing a diffuser on the flash unit provides soft, diffuse Cameras of many shapes and sizes theoretically offer the
illumination. ability to swap lenses, though the designs of some, like
Off-camera flash units range in size, portability and light smartphones, are self-contained. Any change to focal
output. A sync cable is the cheapest method of synchroniz- length must be accomplished by adding supplemental
ing the shutter and flash discharge, though some camera optics in front of the existing ones.
models require additional peripherals to support these
cables. Off-camera flash equipment can also be triggered
via radio frequency communication or optical sensors. Single Lens Reflex Cameras
Studio strobes are large electronic flash units that either Mentioned many times already in this chapter, the classic
use built-in power packs or require tethering to large SLR form factor features an interchangeable lens mount,
capacitors and controllers. Studio strobes typically include mirror system to support an optical viewfinder and a 35mm
a combination of bulbs to provide the extremely bright, film equivalent (24x36mm) sensor size. Crop sensor variants
instantaneous flash as well as continuous tungsten illumi- eschew the more expensive full frame sensor for smaller
nation for previewing light quality and generally lighting sensor sizes that sit at the same focal plane in the camera
the scene between exposures. body. The mirror system allows for additional components
We defer to the lighting experts when it comes to like light and focus sensors. The SLR form factor was a
modifying and controlling electronic flash with lighting staple of analog photography, so it was the obvious choice
accessories; for more on the topic look to Light Science and for early, professional digital cameras. Replacing the film
Magic: An Introduction to Photographic Lighting by Hunter, plane with a digital image sensor was the first step, then
Biver and Fuqua published in 2015 by Focal Press. came replacing mechanical light sensors, lens controls and
other components with ones that all feed to an onboard
computer for automatic control and processing. The digital
Camera Form Factors and SLR (or DSLR) is most common today, though for our pur-
Characteristics poses we use the terms SLR and DSLR interchangeably in
this text.
Camera form factors are many and their features varied. The ergonomics of the SLR and its rectangular sensor
Few of us are shooting with equipment identical to that require rotating the entire camera to change framing ori-
used in the early decades of photography’s birth yet they entations between portrait and landscape compositions.
all share the principles of recording light. The single lens Portrait photographers may opt for an additional body
reflex camera was a staple of the field-shooting profes- attachment that adds shutter buttons and other controls
sional photographer for many decades. Medium format in more comfortable locations when holding the camera
was the studio photographer’s go-to. SLR and medium sideways is a frequent requirement. SLR models designed
format camera types are used by professionals today but for a variety of applications beyond studio photogra-
there exists considerably more variety in form factor, sensor phy include hot shoe mounts atop the pentaprism for
design, viewfinder style and price points that open up connecting portable flash units. Additional features can
availability to prosumers (semi-professional/consumer/ include Global Positioning System (GPS) modules, motion
hobbyist types) and beyond. These cameras offer a host and orientation sensors, wireless transceivers and articulat-
of buttons, dials, displays and touchscreen interfaces for ing displays.
control and review of camera captures.
Above most other features discussed in this chapter,
the primary delineation of camera form factors comes Medium Format Cameras
down to those with interchangeable lens capabilities and When 35mm film wasn’t large enough for commercial
those without (sometimes called dedicated lens cameras).11 applications and reproduction, photographers turned to

94
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

medium format cameras with film sizes like 6x4.5cm (also the viewfinder and the recorded image. This also means
known colloquially as “645”), 6x6cm and 6x7cm. Note that rangefinders are not a good choice when telephoto
that some of these aspect ratios adhere to a more square focal lengths are required, as the viewfinder field of view
relationship between frame height and width compared is static. Wide focal lengths fare much better. This form
to 35mm film. Analog medium format cameras featured factor forgoes any mirror system, allowing the camera
hot-swappable film backs for alternating between film body and lens-to-sensor distance to be slimmer than that
types and quick reloading. Borrowing a trick from inter- of an SLR.
changeable lens mounts, these designs allow for removing Rangefinders get their name from their unique focusing
the back portion of the camera that houses a roll of film like mechanic: to accurately focus at different distances, the
a cartridge. Analog medium format cameras also feature viewfinder uses a split-image view. A centered subset of
reflex mirrors but forgo the pentaprism viewfinder for the frame shows two split reflections of the scene that
ground glass to focus and compose. align perfectly when optimal focus is set. When it’s out
Digital medium format cameras do not necessarily offer of focus, this subset area shows two misaligned or mis-
sensors equal in physical size to their film counterparts (it matched images. The viewfinder is positioned such that
is perpetually expensive to make such large sensors), nor the photographer keeps one eye on the scene, unob-
do they adhere to the square or square-ish aspect ratios. structed by the camera, while the other sees through the
They do use sensors larger than 35mm; in comparison to camera. This provides additional context and situational
a full frame 24x36mm sensor, a medium format camera awareness.
may house one as large as 40x53mm. Larger sensors can Rangefinders are not common in the digital realm—the
translate to higher pixel counts without compromising main true-to-form models are made by Leica Camera AG—
light sensitivity or dynamic range. Medium format cameras though their split-image focusing mechanic exists as a
offer some of the highest megapixel counts available. In digital emulation in electronic viewfinder interfaces.
addition to fully digital models, medium format shooting is
possible by means of digital backs that attach to traditional
film bodies using the unique, detachable camera back View Cameras
mount. View cameras are the largest of the traditional camera
This form factor is decidedly heavier, not as well-suited formats with design roots reaching back to the earliest
for field shooting and instead sees its primary application in cameras. Also known as large format and technical cameras,
photography studios where super high-resolution images they house the largest image plane and therefore accom-
are needed for retouching, compositing and general com- modate large sheets or plates of photographic material
mercial use. Their high price points make them available (typically film). View cameras consist of two components
primarily to working professionals that need the high-end called standards: one houses the lens assembly (includ-
performance and are not as commonly seen in the hands ing a leaf shutter) and the other the ground glass which
of prosumers or hobbyist photographers. functions as a viewfinder and the location at which film is
loaded at the time of capture. The image projected onto
the ground glass is upside-down and backwards, helping
Rangefinders you to appreciate the thankless work done by the penta-
Rangefinders are handheld, field shooting cameras that prism in our SLRs. Large format cameras are simple in that
use a double-lens system. The optical viewfinder is inde- the area between the optics and the image plane is devoid
pendent from the optics used to record an image on the of mechanical components, mirrors or additional optics.
film or image sensor. Since the two components are a The bellows are accordion-like folded black material that
few inches apart in the camera body, it results in a slight serve to adjust the image distance; they make the black
mismatch between what the photographer frames up in box longer or shorter to adjust focus and magnification.

95
 s ecti o n 1: Captu r e

Portable variants are called field cameras featuring foldable since, the compact camera has seen evolutions in materials,
designs for easier transportation; studio versions are called image quality and transitioned from analog to digital imag-
monorail or studio view cameras. Standard film sizes are 4x5 ing technologies. Even in the earliest days, the allure of the
inches and 8x10 inches. Changing from portrait to land- compact camera was the simple controls and consumer
scape orientation is simply a matter of rotating the back convenience. This idea spawned ideas like the one-time-
standard’s film holder. use, disposable plastic camera that consumers used to
expose a roll of film, mail or drop off at a camera store, and
received a set of prints. These disposable cameras used
Mirrorless Cameras fixed aperture, shutter speeds and focus distances. They
Mirrorless cameras offer many of the features of SLRs includ- are alternatively named point-and-shoot cameras as a result
ing full frame sensors but are able to fit it all in smaller form of their ease of use and are almost always dedicated lens
factor bodies. By removing the need for the mirror reflex (non-interchangeable) form factors. Unconventional com-
system and optical viewfinder, the camera housing is much pact cameras include instant-film types like the Polaroid
more compact with a shorter distance needed between and Fujifilm Instax that continue in popularity due to their
the back of the lens and the sensor plane. uniquely analog output.
Mirrorless cameras grew in popularity in the mid-2010s, Digital compact cameras came about in the early 2000s
spearheaded by Sony and Fujifilm. Canon and Nikon, the with affordable pricing and put the power back in the
industry leaders in SLRs, continued to iterate on their full user’s hands to manage, edit and print their image files.
frame models. It is only recently as of this writing that the Focal lengths tend to be normal (around 50mm full frame
two companies have brought professional mirrorless, equivalent) or zoom/telephoto. They also feature built-in
SLR-replacement models to market.12 Taking advantage of flashes that are notorious for causing red eye in portraits.
the smaller body sizes, these newer models have their own They sometimes employ physical shutters though the avail-
lens systems that are also more compact and lightweight. able shutter speeds are lesser in range than what an SLR
Nearly all mirrorless cameras have optional lens adaptors can achieve. Compact cameras stick with fixed apertures
available that bring compatibility for SLR lenses. for the sake of mechanical convenience and because the
The principle of mirrorless camera form factors is average user takes similar-style images in bright light while
building a camera’s internals and technological capabilities expecting everything to be in focus.
from the ground up with the latest imaging capabilities. The compact camera category is shrinking due to
In many ways, the DSLR is an analog film camera design mobile device cameras. Offering the same convenience
with its internals converted for a digital sensor. The evolu- and user-friendliness, they’ve eclipsed the need for sep-
tion was logical and functional. Mirrorless designs promise arate, single-function hardware for casual photography.
an exciting future for camera form factors as they build Some limited compact camera models exist that target
around digital-exclusive functionality. The balance they prosumers looking for raw capabilities and a degree of
strike between small hardware with minimal moving parts advanced controls, featuring nearly pocketable designs and
and high-quality sensors makes them an excellent choice the latest sensor technologies.
for aerial imaging with remotely piloted drone platforms
(unmanned aerial vehicles).
Mobile Phone Cameras
It’s impossible to understate the sea change in democra-
Compact Cameras tized image-making brought about by cellphones in the
Compact or small format cameras are a category that started last decade. The early 2000s saw the earliest integrations
with Kodak in the late 1880s when the only previously of camera systems into cellphones as full color screens and
available option was large format. In the nearly 150 years data networks became available. The birth of smartphones

96
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

in the middle of that decade then brought high-resolu- remains to be seen how imaging systems may be inte-
tion screens, touch interfaces, mobile operating systems, grated into wearable computing like watches and smart
capable processors and internet connections to devices glasses.
whose previous, primary function was voice calling. The
embedded camera offerings matured quickly after that.
Today, camera systems are a staple hardware feature on Action Cameras
every consumer smartphone and eschew traditional image A niche but notable category, action cameras are a class
processing workflows for mobile apps and instant upload of ultraportable all-in-one imaging systems designed
to social media platforms. primarily for video capture with fixed focal lengths and
The ubiquitous form factor of smartphones is that of remote tethering. These cameras feature a similar class
a thin metal and glass slab with a few physical buttons of sensors used in smartphones and often use super
on either side. This is radically different from any tradi- wide-angle lenses or multiple sensor and lens combina-
tional camera design and presents serious challenges to tions to extend the field of view to 180° or 360°. Action
designing an imaging system within it. Phone cameras cameras come with a slew of waterproofing, stabilizing
are afforded extremely small footprints for compound and mounting accessories in order to get cameras into
lenses and image distances to the sensor. They feature very places they were previously unable to access. While they
small sensor sizes, fixed apertures, electronic shutters and are largely advertised for high-definition video record-
fixed focal lengths. The sensors can rival professional SLR ing, still photos (sometimes offering raw capture) can be
megapixel counts but do so with smaller pixel pitches and of high enough resolution for professional output and
therefore greatly diminished low light performance. LED reproduction. Action cameras are frequently mounted or
flash units are often positioned near the primary imaging built into drones and other remotely operated motion
lens which often leads to red eye in portraits. Optical image platforms.
stabilization is available on some; others opt for electronic We’ve listed the common form factor categories, a
image stabilization to minimize the impact of handshake selection of which are shown in Figure 3-23 (overleaf). It’s
(which is ever-present given the ergonomics). Phase growing increasingly difficult to sort cameras into these
detection and contrast detection may be used for autofo- discrete categories, however. For example, the Fujifilm GFX
cus. Since smartphones have computing demands beyond 50R houses a medium-format-sized sensor, uses a mirror-
running the camera, processors and therefore image less imaging pathway and uses a compact body design
processing capabilities of phone cameras are very capable. reminiscent of a rangefinder.13 Some camera manufactur-
Fortunately so, as some image quality shortcomings from ers are leveraging the familiarity of smartphone interfaces
these tiny imaging systems can be mitigated with image and software, running mobile operating systems directly
processing corrections. on dedicated camera hardware. The industry continues to
Smartphones are permanent mobile computing evolve with a combination of technological advances and
fixtures in our modern lives. It’s likely that the imaging user preferences. Camera form factors are increasingly bol-
capabilities of our devices will continue to grow in qual- stered and reliant upon complex software with important
ity, complexity and features. Developing trends in this integrations to our networks, remote workstations, cloud
form factor include multi-camera arrangements (adding storage and social web platforms.
supplemental focal lengths, depth detection or otherwise
specialized modules), miniaturized variable apertures and
more. Truly, exciting developments in the mobile camera
space are pushing the medium forward as computational
photography leverages software to create previously
unimaginable image assets with mobile hardware. It

97
 s ecti o n 1: Captu r e

Figure 3-23 Camera technology comes in many shapes and sizes. Pictured along the top row from left to right: a compact camera with
interchangeable lens capability, a rangefinder, an instant film camera and an action camera. Pictured along the bottom row is a 4x5 view camera, a
full frame DSLR, a full frame medium format camera and a 360° stereoscopic camera. Photographs by Rochester Institute of Technology alumnus
and Lecturer Dan Hughes

Camera Storage Media The two common camera storage flash memory formats
are CompactFlash (CF) and Secure Digital (SD). Many pro-
Removable, solid state flash memory is the go-to in-camera prietary, manufacturer-specific memory card formats have
capture storage solution. Every exposure must be stored, at come and gone while SD and CF have withstood waves of
least temporarily, into local memory before it is transferred technology evolutions and ever-increasing storage needs.
to other storage media. Flash memory is ideal because it Though market competition has whittled the options
offers small form factors, quickly reads and writes digital down to these two contenders, there’s no shortage of
information, has no moving parts and can be swapped variants, tiers and new generations of each. We recom-
out as needed. This is particularly welcome if you are old mend checking online for the latest numbers and product
enough to recall the awkward growing phases of digital developments, buying the best you can afford for the work
cameras that used floppy disks or compact discs as image you’ll be shooting in the near-term. It’s a dizzying blend of
file storage solutions built into the camera body. Some spec bumps and marketing speak. A bit of research before
cameras today offer built-in flash memory storage for purchasing ensures an educated purchase.
emergency use, though the capacities are rarely enough CompactFlash was developed by SanDisk in the mid-
to get by for long without an additional memory card. A 1990s along with the establishment of the CompactFlash
well-prepared photographer keeps a set of cards on-hand Association. Versions of CF cards offering increasing read/
when shooting, protected from dust and debris in plastic write speeds add nomenclature like Ultra and Extreme to
cases or card wallets. the product names. CF slots exist on a number of camera

98
 c h a p t e r 3 : D i g i ta l C a m e r a T e c h n o lo gy

models available today and capacities top out around the data or otherwise interrupt its processing. As men-
512GB. Their read/write speeds are marketed in terms of an tioned earlier, any memory cards that reside outside of the
Ultra Direct Memory Access or UDMA rating; the higher the camera should be stored in cases or sleeves to protect the
UDMA, the faster the card. The highest UDMA available is metal contacts. Haphazardly handling the cards or throw-
7.14 The XQD format, also created by SanDisk, was proposed ing them in your lint-filled pocket alongside car keys is an
in 2011 as a faster, non-backwards-compatible successor to excellent way to render the card unusable by your camera
CF with a smaller form factor.15 Its use is supported in some and card reader.
Sony and Nikon camera systems. Writing camera captures to a memory card is short-
Secure Digital cards are smaller than CompactFlash but term solution and truly the only convenient one when
offer similar storage sizes, read/write speeds and ubiquity shooting in the field. Transferring them to desktop storage
in camera system support. The SD format was developed is accomplished using an external memory card reader or
in the late 1990s by the SD Association, an organization by connecting the camera to a computer via USB cable.
formed by SanDisk, Toshiba and Panasonic.16 The SD Higher-end cameras offer ethernet connections to tether
memory card format expanded over the years to include directly to a computer or network storage solution (ideal
variations like SD High Capacity (SDHC), SD Extended for studio shooting or a fully equipped field production)
Capacity (SDXC) and most recently, SD Ultra Capacity and offer the fastest transfer speeds. Cameras across all
(SDUC). The largest capacity of an SDXC card is 2TB and categories are increasingly including WiFi or Bluetooth
SDUC expands that up to 128TB.17 Iterative versions of connectivity to support file transfers to phones, tablets or
SD cards offer faster read/write speeds for large raw files, laptops. These functions eliminate the steps of plugging
increased rapid-capture frames per second and video a cable from the camera to a computer, or removing the
recording. SD cards are also marketed with Ultra High memory card and inserting it into a wired card reader.
Speed (UHS) ratings and classes that indicate minimum However, your particular use-case should consider the data
and maximum read/write speeds.18 They use a video class transfer rates of each to best serve your workflow: don’t
(“V” class) label on cards specifically designed for handling get caught waiting for huge raw files to download one at
high-resolution video recording. If you work with large raw a time over a slow WiFi network, for example. We review
files or rely on continuous shooting for fast-action events, storage options in Chapter 8.
research the latest versions of memory cards that are
compatible with your camera as the advertised speeds may
not be achievable with older camera models. The microSD Notes
card is an even smaller variant used for expandable storage 1 Liles, Jonathan Moore. “X-Trans vs Bayer sensors: Fantastic
for electronics like action cameras. The microSD form factor claims and how to test them.” PetaPixel, Mar. 3, 2017,
comes in the same flavors as its larger, standard SD form petapixel.com/2017/03/03/x-trans-vs-bayer-sensors-
factor sibling. fantastic-claims-test/.
2 Spring, Kenneth R., and Michael W. Davidson. “Concepts in
Some professional cameras include multiple memory
digital imaging technology quantum efficiency.” Hamamatsu
card slots, either of matching design (dual SD, for example)
Learning Center: Avalanche Photodiodes, hamamatsu.magnet.
or mixing different card types (SD and CF). Camera firm-
fsu.edu/articles/quantumefficiency.html.
ware offers different configurations of the memory slots to
3 “What is a CCD?” What is a CCD? Charge Coupled Device,
further expand the total number of captures or to have one Special Instruments Inc., www.specinst.com/What_Is_A_
function as a mirrored backup of the other. CCD.html.
Reading and writing activity on a removable memory 4 “ON Semiconductor to acquire Truesense Imaging, Inc.”
card is commonly indicated with an LED on the camera LM317T: 3-Terminal 1.5 A Positive Adjustable Voltage Regulator,
body. Never take a card out of the camera when it is Apr. 2, 2014, www.onsemi.com/PowerSolutions/newsItem.
actively reading or recording an image as it can corrupt do?article=3076.

99
 s ecti o n 1: Captu r e

5 “More information about color shading.” DXO Labs, 2003, 16 “SD Association celebrates 10 years of innovation at CES
web.archive.org/web/20180503071520/https://www.dxo. 2010.” Hughsnews.ca, Jan. 5, 2010, www.hughsnews.ca/
com/us/more-information-about-color-shading. sd-association-celebrates-10-years-00970.
6 DeVriere, Frederic. “Quantum efficiency advances 17 Gaz, Bryan. “SD express SDUC memory cards
sensors for night vision.” Photonics Media, Photonics. will give you 985MB/s transfer speeds and
com, Jan. 10, 2014, www.photonics.com/Articles/ 128TB of storage.” DPReview, DPReview, June 29,
Quantum_Efficiency_Advances_Sensors_for_Night/ 2018, www.dpreview.com/news/7847794311/
a51008#Comments. sd-express-sduc-memory-cards-association.
7 Sanyal, Rishi. “UPDATED: Sony’s curved sensors may allow 18 “Difference between speed class, UHS speed class,
for simpler lenses and better images.” DPReview, DPReview, speed ratings (performance) and video speed class
July 7, 2014, www.dpreview.com/articles/6229436014/ for SD/SDHC/SDXC cards.” SanDisk, Apr. 11, 2017,
sony-s-curved-sensors-may-allow-for-simpler-lenses-and- kb.sandisk.com/app/answers/detail/a_id/1996/~/
better-images. difference-between-speed-class%2C-uhs-speed-class%2C-
8 Courtland, Rachel. “Sony creates curved CMOS sensors that speed-ratings-%28performance%29.
mimic the eye.” IEEE Spectrum: Technology, Engineering,
and Science News, IEEE Spectrum, June 12, 2014,
spectrum.ieee.org/tech-talk/semiconductors/devices/
sony-creates-curved-cmos-sensors-that-mimic-the-eye.
9 Guenter, Brian, et al. “Highly curved image sensors: A
practical approach for improved optical performance.”
Optics Express, vol. 25, no. 12, 2017, p. 13010., doi:10.1364/
oe.25.013010.
10 Reshidko, Dmitry and José Sasian. “Current trends in
miniature camera lens technology.” SPIE Newsroom, 2016,
doi:10.1117/2.1201602.006327.
11 Modrak, Rebekah and Bill Anthes. Reframing Photography:
Theory and Practice. Routledge, 2011, p. 64.
12 “Nikon introduces the new Nikon Z mount system, and
releases two full-frame mirrorless cameras: The Nikon Z 7
and Nikon Z 6.” Nikonusa.com, Aug. 23, 2018, www.nikonusa.
com/en/about-nikon/press-room/press-release/jktu9ets/
Nikon-Introduces-the-New-Nikon-Z-Mount-System%2C-
and-Releases-Two-Full-Frame-Mirrorless-Cameras%3A-the-
Nikon-Z-7-and-Nikon-Z-6-.html.
13 “FUJIFILM GFX 50R.” Fujifilm Value from Innovation, www.
fujifilm.com/products/digital_cameras/gfx/fujifilm_gfx_50r/.
14 “SanDisk CompactFlash read/write speeds and UDMA
information.” SanDisk, Aug. 21, 2014, kb.sandisk.com/
app/answers/detail/a_id/31/~/sandisk-compactflash-
read%2Fwrite-speeds-and-udma-informatio.
15 “The CompactFlash Association announces the recently
adopted XQDTM specification as a new memory card
format.” Compact Flash Association, Dec. 7, 2011, www.
compactflash.org/assets/docs/cfapress/xqd_111207.pdf.

100
 4 Photographic
Exposure

Photograph by Ruben Vargas

Photographers with a clear understanding of light, optics and camera components are
primed to fully explore and master photographic exposure. We spend a considerable
amount of time with cameras in hand chasing good light in the field and in the studio. We
must be equally good at recording it as we are at finding it. This chapter fully explores what
photographic exposure entails. This includes how exposure is calculated and how to work
with the camera controls of shutter speed, aperture and ISO. It also requires a complete
understanding of scene dynamic range, image histograms and light metering methods. It
is both a technical topic and a constant creative puzzle that photographers are challenged
to solve with every captured frame. The last section of this chapter offers a shooting exer-
cise to test your understanding of exposure variables, metering and image quality tradeoffs
by shooting bracketed exposures of a challenging nighttime scene.

101
 s ecti o n 1: Captu r e

Exposure Determination Determining the correct exposure is done through light

metering. The goal of metering light is to determine the
First and foremost: exposure is not an acceptable form of average scene brightness to then determine the correct
payment from a potential client! More seriously, exposure is camera exposure. This is the basis of spot metering and is
a combination of: the basic camera metering approach. However, complex
metering systems and software algorithms mean that there
• Light intensity from the scene are many different ways to reach an opinion on optimal
• Light volume permitted through the lens aperture exposure. No matter the method of determination, proper
opening exposure is just a planning step—configuring the camera
• Light duration on the sensor dictated by the shutter and executing is what gets us recorded photographs.
opening Before going further, let’s define the universal photo-
• The sensitivity of light recording via the sensor itself graphic exposure descriptor: the stop. A stop is one full step
change in exposure. A one stop change equates to a halv-
A studio environment allows for complete control of ing or doubling of light relative to the original conditions.
a scene’s light intensity while natural or uncontrolled Stops are used to indicate changes in aperture, shutter
environments leave the other three variables available for speed, ISO and even flash output. The contributing factors
dictating a photographic exposure. might be a mix of any of those exposure parameters as
We’ve seen numerous attempts to draw diagrams long as the result is a full doubling or halving of light in the
explaining photographic exposure (search for “exposure captured image. Camera exposure indicators in viewfinders
triangle” online) that are more confusing than useful. So, typically show a scale or bracket with a pointer or needle.
we’re keeping it simple. The amount of light, the time The center of this scale indicates the optimal photographic
it’s allowed to hit the sensor and the electronic configu- exposure as determined by the internal light meter. Any
ration of that sensor must be assigned in harmony with time that the pointer moves to the left or right of this
one another based on light measured from the scene to center point is an indication that the exposure resulting
achieve a proper exposure. Each of these variables presents from your current settings will yield an underexposed or
its own trade space (in other words, pro’s and con’s) to overexposed frame, respectively. This scale is often marked
weigh based on circumstance, aesthetics and sometimes with reference points to indicate whole stop increments of
simple personal preference. underexposure and overexposure.
There are two types of exposures to review. The first
is photographic exposure (H ) defined as the quantity of
light per unit area reaching the sensor and is calculated Dictating Exposure Time with
by multiplying the illuminance (E ) and the exposure time Shutter Speed
(t). Photographic exposure is expressed as H = E x t. The
second type is camera exposure, referring to the combi- Shutter speed is a measure of the exposure time or the
nation of shutter speed, f-number and ISO speed used to amount of time that light is collected by the sensor. DSLRs
properly expose an image. offer shutter speeds in increments as small as ⅓ stop and
Photographers are primarily concerned with achieving with ranges spanning from as quick as 1/8000th second
the proper camera exposure (avoiding an overly dark or to 30 seconds. Figure 4-1 lists the common whole stop
bright image). Proper camera exposure makes the most of shutter speeds, though many cameras allow for increment-
the camera sensor’s capabilities and renders the scene with ing exposure in half, third and quarter stops between these
visually pleasing tonality. We explore all of the components settings. Exposure times longer than 30 seconds rely on
needed to achieve an optimal exposure over the next the bulb mode where the photographer actively opens and
several sections. closes the shutter and uses a handheld timer. Don’t get too

102
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

hung up on the term as it’s a remnant of analog camera

designs where the shutter was triggered by a shutter
release cable. Early cables used a rubber bulb that used air
pressure to physically depress the shutter button. The term
persists even though the mechanics of initiating exposures
longer than 30 seconds are different today.
Shutter speed selection dramatically impacts the final
image if there are moving objects in the scene or when
there is camera motion. Let’s use a merry-go-round as an
example. To get a picture of a child on the ride, we select
a short shutter speed to stop the motion and have a sharp
image of the entire merry-go-round and the child. To
instead highlight the motion of the merry-go-round, we
select a much slower shutter speed to introduce motion
blur.
A second consideration when selecting shutter speed is
equipment: are you hand holding the shot or is the camera
mounted to a tripod? Additionally, what lens are you using? Figure 4-1 A reference scale of whole stop ISO, aperture and shutter
speeds. Note that the three columns are not directly tied to one
Hand holding a camera at too slow of a shutter speed often
another.
introduces blur in the image, as it is difficult to remain
perfectly still during exposure. Here we turn to the hand-
held shutter speed rule of thumb. It states that the minimum estimate the slowest shutter speed to safely use without
shutter speed to avoid camera shake and blur is found with risking handshake or blur. That said, lenses with built-in
the equation: minimum handheld shutter speed = 1/(effective optical image stabilization help to lower the minimum
lens focal length). We specify effective focal length as the handheld shutter speed. Refer back to the Image Stabiliza-
lens focal length may need to consider a crop sensor’s crop tion section in Chapter 3 for more on this topic.
factor. As an example, this rule dictates that the minimum A section on shutter speed would not be complete
shutter speed for shooting handheld with a 500mm lens without providing a list of common whole stop incre-
on a full frame camera is 1/500th of a second. Notice the ments of shutter speed: 1/1000s, 1/500s, 1/250s, 1/125s,
inverse relationship between shutter speed and focal 1/60s, 1/30s, 1/15s, 1/8s, 1/4s, 1/2s, 1s. Since one full
length. The minimal acceptable shutter speed increases stop change halves or doubles the exposure, additional
as focal length increases. Using a crop sensor requires shutter speeds in either direction are easily determined.
changing the equation to: minimum handheld shutter speed Figure 4-2 shows a spinning black and white disc photo-
= 1/(focal length x crop factor ). This rule of thumb helps to graphed with a sequence of shutter speeds, going from

Figure 4-2 A spinning black and white disc photographed with decreasing shutter speeds.

103
 s ecti o n 1: Captu r e

fast to slow, as other exposure parameters are adjusted to

compensate for the change in brightness. The portions
of black to white are accurately depicted at 1/1000th of
a second, as the short shutter speed captures the disc in
motion.
Not all cameras have physical shutters. The exposure
behavior whereby the sensor detects and records a light
signal is described as the integration time. The photosites
are controlled electronically to “turn on” and “turn off” such
that they record incoming light exclusively during the time
in between. Integration time is a function of time lapsed
and can adhere to the same traditional set of speeds like
1/30s, 1/60s and so on. We can say that a smartphone
camera exposure uses an integration time of 1/30s to Figure 4-3 Determine the effective aperture by placing a point source
of light one focal length behind the lens and measure the diameter of
accurately describe capture behavior of the shutterless the beam of light that emerges from the front of the lens.
system. This integration exposure process happens in rapid
succession, over and over again when watching the live
video preview in our camera app. the time they enter the lens and when they pass through
the diaphragm opening.
F-numbers are calculated by dividing the lens focal
Dictating the Amount of Light length (f) by the effective aperture (D), or f-number = f/D.
with F-Number The effective aperture is rarely the same size as the aper-
ture (the physical opening in the diaphragm); the diameter
Shutter speeds are easy to keep track of because they are of the entering beam of light that just fills the diaphragm
described as a function of time. When setting the shutter opening must be measured. One method to determine
speed to 2 seconds and then again to 1/30th second, you this diameter is to place a point source of light one focal
have a general, intuitive sense for how this change trans- length behind the lens so the diaphragm opening restricts
lates to a mechanical action. Aperture is not as intuitive the beam of light emerging from the front of the lens. This
but using the standard unit of exposure measurement, beam is the same size as an entering beam that just fills
the stop, you can quickly become fluent in anticipating its the diaphragm opening (see Figure 4-3). The diameter of a
influence on exposure. beam of light is measurable with a ruler in front of the lens.
Relative aperture or aperture are alternative names for This would be a quick calculation if all lenses featured
f-number. The value of the relative aperture or f-number fixed apertures. Instead, most lenses have a range of aper-
depends upon two things: the focal length of the lens and tures to pick from when selecting photographic exposure.
the effective aperture. The effective aperture is defined as It’s useful to know the series of f-numbers that represent
the diameter of the entering beam of light that just fills the whole stops changes: f/0.7, 1.0, 1.4, 2.0, 2.8, 4, 5.6, 8, 11, 16,
opening in the diaphragm of a camera lens or other optical 22, 32, 45, 64. The f-number series is extended in either
system. The diameter of the opening in the diaphragm is direction by understanding that the factor for adjacent
known as the aperture. When the diaphragm is located in numbers is the square root of 2, or approximately 1.4. Also
front of a lens, the effective aperture is found by measuring note that alternate f-numbers vary by a factor of 2, with a
the diaphragm directly. Rarely is the diaphragm located in small adjustment between 5.6 and 11, and between 22 and
front of a photographic lens, so it’s necessary to take into 45, to compensate for the cumulative effect of fractional
account how the light rays may change direction between units.

104
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

Each time that the diaphragm is stopped down one

stop (i.e. from f/8 to f/11), the amount of light reaching the
sensor is cut in half (divided by 2) and the exposure time
required to obtain the same photographic exposure is
multiplied by 2. The aperture effectively dictates the light
transmitting ability of an imaging system.
Recall that we explored depth of field and bokeh in
Chapter 2. Selecting an f-number when determining
camera exposure impacts the visual characteristics of the
resulting capture. As the aperture gets smaller (increasing
f-number), depth of field increases. Increasing the aperture
size (decreasing f-number) decreases depth of field. Figure
4-4 illustrates the effect of shooting the same scene at
either end of a lens’ available f-number settings.
The amount of in-focus scene content (that which falls
within the depth of field) is also codependent on sensor
size. Smaller sensors like those found in point-and-shoot
and smartphone cameras have some of the smallest sen-
sors used for photography. Photographs made with these
sensors have particularly large depth of field, rendering
most, if not all, of the scene in sharp focus. This is great
for avoiding out-of-focus portraits and not so great when
you’re going for that classic cinematic look with extremely Figure 4-4 The depth of field at f/1.4 (top) compared to f/22 (bottom)
with focus position kept constant. Photographs by Rochester Institute
shallow depth of field. Since this characteristic is inherent
of Technology photography alumni Jonathan Damian and Nathan
to small sensors due to geometric optics principles, clever Pallace
software workarounds exist to simulate the effect. Artificial
blurring of the background is possible either through edge
detection (blurring anything that the algorithm thinks is widths depending on the variable focal length used. Slow
surrounding environment) or advanced computer vision. lenses may be limited, for example, to a maximum open-
This approach is called synthetic depth of field or synthetic ing of f/4.0 or f/5.6. This requires slower shutter speeds
aperture effects. (multiple stops-worth) to achieve proper exposure. A
lens does not need to be fast to be a valuable asset in the
photographer’s bag, it simply offers additional flexibility for
Lens Speed adjusting exposure settings.
Lastly, a lens’ range of available aperture settings is com-
monly used to describe it with shorthand terminology.
Some lenses are described as “fast” and others, “slow.” Dictating Effective Sensitivity with
A lens capable of opening wide to an f-number such as Gain and ISO
f/1.4 or f/2.0 is a fast lens because the wider the opening,
the more light available for exposure and the shorter the The third parameter for setting exposure is ISO, also known
possible shutter speed. Not all lens designs offer aperture as the sensor speed. ISO is an acronym for the International
settings this wide, especially as the focal length increases. Standards Organization, though that’s not particularly
Zoom lenses often have different maximum aperture descriptive in this context. The term speed comes from

105
 s ecti o n 1: Captu r e

the characteristic we mentioned describing lens apertures:

the “faster” the speed of a sensor, the less time it takes to
record a proper photographic exposure. ISO is a standard-
ized industry scale for measuring sensitivity or response to
light. This definition might lead to a misunderstanding of
camera ISO. Every camera has a base or native ISO, some-
where between ISO 50 and ISO 200. Thus, ISO is a measure
of the sensor’s sensitivity to light. So why do cameras
have many ISO settings, ranging from 50 to 102,400 and
beyond?
Image noise must be better understood before answer-
ing this question. There are several types of noise that
occur in a camera system. The visual result of all noise is the
same: random brightness or color variance in the image.
Noise is an image artifact and degrades image quality. The
best quality image is one captured when shooting at the
camera’s base ISO. Chapter 11 unpacks the surprisingly
varied forms of image noise. The important takeaway here
is that excessive noise should be avoided when setting
photographic exposure.
Doubling the ISO or speed increases exposure by
one stop. Cameras offer many ISO settings to pick from,
ranging from the base ISO (100 or 200) to over 102,400
depending on the model (see Figure 4-5). A common mis-
conception is that changing the ISO alters the sensitivity
of the camera to light. This is not the case. Changing the
ISO changes the gain applied to the electrical signal. The
amount of gain dictates the degree of amplification of the
original recorded voltages read off of the sensor’s photodi- Figure 4-5 A sequence of equivalent exposures of gray and black
odes. This amplification translates to brighter pixel values targets at increasing ISO settings. Image noise increases with higher
ISOs, most often becoming visually apparent in dark shadows.
overall, however, it amplifies any random, non-image
forming signal at the same time. Keep in mind that noise is
present in all images even when using the sensor’s native
ISO. Setting a higher ISO to create a desired photographic Avoid these settings, as skillful post-processing of an
exposure means adding additional gain to the incoming underexposed or overexposed raw file offers superior
photon signal. An image may be unusable after a certain results.
degree of signal amplification due to the distracting visual When possible, use the lowest ISO available. However,
quality of noise. ISO is one of the three settings available when setting
Some cameras come with “low” or “high” settings that exposure and should be leveraged when other settings are
allow for going above or below the native ISO range fixed for one reason or another.
on the camera, suspiciously forgoing the use of an ISO
number. These settings are software simulated speed
changes and are likely to result in poor image quality.

106
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

Exposure and Dynamic Range Dynamic range is described as a ratio such as 1,000:1.
Photographers find it useful to measure it in terms of
The term dynamic range is used for both capture and exposure stops since it relates directly to camera settings.
output in photography. In this section, we’ll stick within The human eye perceives a range of brightnesses upwards
the context of photographic exposure. The sensor dynamic of 24 stops. A sensor typically records about 12 stops. It’s
range is the maximum recordable difference between the not a fair fight when you recognize the eye’s ability to
brightest and darkest areas of detail. It can be stated as a rapidly modulate incoming light via the pupil and that the
ratio of light contrast or as the number of stops. retina uses two mechanisms for seeing (one for bright, one
We derive a begrudging appreciation for the range of for dark scenes) while the camera sensor only has one type
brightnesses in our world any time we’ve woken up in of photosite. Sensor technology continues to evolve and
the middle of the night to a dark bedroom only to shuffle perhaps it will someday match the biological evolution of
into the bathroom and flip on the lights. Even though our visual system. For now, a single capture with an image
the darkness isn’t anywhere near pitch black, nor is the sensor represents a more limited dynamic range recording
bathroom lighting as extremely bright as the sun, our eyes capability.
can’t handle the wide range of light values to see in both We measure scene dynamic range by taking lux
environments in quick succession. The natural world offers readings in the brightest highlight area and the dark-
a massive range of luminance from one moment or envi- est shadows. The difference between the readings, in
ronment to the next. This is a challenge for photographic stops, defines the contrast ratio. We always want our
exposure because the scene dynamic range is potentially camera’s dynamic range to be at least as large as our
greater or wider than the sensor’s ability to record in a scene dynamic range so as not to lose detail in either the
single instant. While we don’t always photograph scenes highlights or shadows. When the scene’s dynamic range
with dramatically varying brightnesses (also called the is larger, we make adjustments such as adding a reflector,
scene luminance ratio), we need a sensor that keeps up for fill lighting or using high dynamic range imaging tech-
the times that we do. Sometimes a simple scene leads to niques. Alternatively, we choose an exposure that strikes
lost detail at capture due to bright reflections or dark, light a reasonable balance or prioritizes a subject or scene
absorbing fabrics. Areas with detail in extreme highlight element over others. Figure 4-6 (overleaf) shows a typical
or extreme shadow are captured as white or black respec- outdoor scene where foreground elements in shade
tively when the scene dynamic range is greater than the represent different light levels compared to areas of deep
sensor’s dynamic range. shadow and the bright sunny sky. Our eyes concurrently
A sensor photosite’s bit depth indicates the range of saw clouds with detail and foliage in the dark bushes, yet
brightnesses it can record. Consider a photosite like a our camera could not capture the full dynamic range in a
drinking glass and the photons like drops of water filling single exposed photograph. Capturing a proper exposure
it. Once filled, there’s no room to hold additional water. A for the foreground tree proved the most important given
photosite’s maximum photon-holding capacity defines the composition and intent.
the brightest incoming exposure it records before losing When the camera’s dynamic range exceeds that of
detail. This tells us something about the upper limit of the the scene, we’re afforded some exposure latitude. Expo-
sensor’s dynamic range. The lower limit is the black level or sure latitude is the difference between camera dynamic
the darkest possible detail recorded sans noise. Although range and scene dynamic range expressed in stops and
the sensor photosite’s bit depth can indicate the senor’s it allows for wiggle room in getting capture exposure
dynamic range it is not always completely accurate. The bit just right. It is also defined as the range over which the
depth does not take into consideration other factors that exposure can be increased and decreased from the
may further lower a camera’s dynamic range such as scene correct or optimal exposure and still produce acceptable
reflections and lens flare. results.

107
 s ecti o n 1: Captu r e

Figure 4-6 Ratios of dark to light in different environments can be extreme, though our eyes are capable of helping us
see well in most of them. The camera is more limited in its ability to record such brightness ratios.

As exposure is decreased from the optimum level, called Blooming

underexposure, the darker areas or shadows of the scene Recall that photosites are wells that catch photons and
first lose contrast and then detail. If we continue to reduce convert them into voltages. Using the analogy of a photo-
the exposure, detail is lost in progressively lighter areas of site being akin to a glass filling with water, what happens
the scene and eventually there is no image detail even in when the glass fills up before the faucet is turned off?
the lightest areas or highlights of the scene. Similar to liquid overflowing, overexposure causes photons
Taking the process in the opposite direction increas- to spill out from the photosite well and into neighboring
ing the exposure and overexposing the images, the first photosites. This is called blooming and it occurs when
noticeable change is commonly an increase in shadow a photosite cannot hold all of the electrical energy it’s
contrast (and therefore an increase in overall contrast). receiving. Any photons reaching the photosite past this
Overexposure also produces a decrease in image definition saturation point spills or leaks over. Not only does this
and as we continue to increase exposure, image contrast mean we’re missing out on light information (a fully sat-
decreases, first in the highlight areas and then toward pro- urated photosite means clipped highlights), but its effect
gressively darker areas of the scene. cascades. When one photosite overflows, the excess spills
into its neighbors, who then also overflow more quickly.
Blooming occurs most commonly when a particularly
bright part of an image (like the sun in the sky) is causing a
subset of photosites to overfill. Blooming appears as bright
halos on object edges or linear streaks of brightness across

108
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

“um,” you add a tick mark in that column. At the end of the
lecture, you plot the distracting nervous crutches with the
x-axis containing all the different filler words and the y-axis
indicating how often they were used. This is an example
of a frequency histogram. It’s also a great way to distract
yourself from actually learning something from the lecture
(which we do not recommend).
Knowing the frequency of pixel brightnesses in an
image is helpful to visualize proper exposure and therefore
making a frequency histogram a logical choice for visual-
izing captured image data. Most digital cameras provide
an on-camera histogram so that you don’t have to dust
Figure 4-7 Blooming results when photosites receive more photons off your spreadsheet software. The on-camera histogram
than they can handle. is valuable for checking proper exposure immediately
after capture and before it’s too late to try again. Taking
advantage of the histogram plot to perfectly nail exposure
the image. Figure 4-7 is an example of blooming mixed requires us to quickly interpret its insights.
with flare from having the sun in the frame. An image histogram plots the frequency of tones in an
image based on scene luminosity. In other words, you’re likely
to find a luminosity histogram unless it’s specifically made
The Utility of Image Histograms clear that it’s a color channel histogram (typically visualized as
overlapping red, green and blue channel data). Checking the
A frequency histogram is a graph that helps visualize the histogram helps to check that the scene is properly exposed
frequency with which some countable quality or event and includes a logical range of brightnesses.
occurs. When listening to a boring lecture, you may idly An image histogram is a breakdown of the image’s
mark the number of times the speaker says “um,” “uh” or tonal scale and the frequency of pixels at a given
“like” in the margins of a notebook. Each time there’s an tone value (see Figure 4-8). While no two captured

Figure 4-8 A black and white image (left) simplified to a much smaller collection of pixels (middle). Each pixel describes a brightness value.
Reorganizing these pixels by brightness and stacking all same-toned pixels yields a basic frequency histogram (right). It’s impossible to interpret
from the histogram alone that it’s an image of a building and trees.

109
 s ecti o n 1: Captu r e

Taken as a whole, the histogram reveals an image’s tonal

range. Many photographers characterize the shape of an
image’s histogram as reminiscent of a mountain range
in silhouette (blame those unimaginative landscape
photographers).
We can alternatively generate a histogram to show
the pixel colors at each brightness level. Figure 4-10
shows an example of such a histogram: the image pixels
are rearranged by brightness while retaining their color.
Figure 4-9 A frequency histogram of image pixel values ranging
from 0 to 255. We can see how a majority of the green grass pixels are
midtones and the blue sky ranges from midtone to bright
highlights. Most pixels in the scene are green, second
photographs are likely to ever be identical, it is possible to up is blue, followed by some grays and browns in the
have two images with the same histogram. Think of it like bench and walking path. Unlike a traditional luminos-
assembling a popular children’s toy of interlocking plastic ity histogram, this histogram contains all of the pixels
bricks: you can make the model shown on the box with in the original image file. The only difference is their
the included pieces or make something entirely differ- organization.
ent from that same set. Both constructions have, say, 20
red bricks, but they’re not going to be used in the same
location or in the same arrangement. The same is true for Histogram Interpretation
a photograph: it’s tempting to think of a histogram like an A histogram represents a scene’s tonal range ranging
image’s fingerprint, however, its lack of spatial informa- from black (digital count: 0) to white (digital count: 255)
tion means it’s not a complete encapsulation of the final for an 8-bit image. A quick glance at an in-camera histo-
product. gram easily reveals overexposure or underexposure when
Let’s review the image histogram shown in Figure 4-9. we understand that the leftmost edge of the histogram
The x-axis plots the brightness levels. This is sometimes records absolute black, the right side absolute white. Data
labeled pixel value, as pixel values are the brightness levels recorded at these extreme ends has no opportunity for
in the image. This particular histogram’s x-axis spans variation or differences considered to be recorded detail.
from 0 to 255, the range of brightness levels available in When a histogram shows pixels stacked up against these
an 8-bit JPEG image. All image histograms go from the edges, that suggests its shape (representing the scene’s
darkest to the brightest possible pixel values; higher bit tonal range) would have continued had it not been cut
depth files see the maximum value increase. The y-axis off at value 0 or 255—highlight or shadow details are lost.
of a histogram is frequency. This is a count of how many This characteristic is called clipping, where detail is clipped
pixels exist at each brightness level. It’s helpful to consider or abruptly cut to the point of being irreversibly lost.
the histogram in three pieces: the left third represents Additionally, the term crushed specifically refers to under-
the shadow information from solid black to a dark gray. exposure and clipped shadow detail. Camera previews
The center third represents the midtone information, all include a clipping warning that overlays a blinking color
middle shades of gray. The right third of the histogram or alternating black and white pattern to indicate clipped
represents highlight information from light gray to pure or crushed areas of the scene. Desktop software offers
white. There is no absolute boundary between each of similar visual feedback to easily identify when image
these, as one person’s bright shadow tone is another per- edits have introduced clipped highlights or shadows for
son’s dark midtone. Keeping these three general areas in which some data existed in the original exposure. A small
mind makes it easier to interpret photographic exposure. amount of clipping is common and not an automatic

110
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

Figure 4-10 A image histogram can be arranged by luminance frequency and plotted using the pixel colors at
each of those brightness values. This histogram was generated using a web-based tool called Pixel Chart.1

indication of poor camera exposure. A large amount can 0 to 255 indicates a wide dynamic range: a scene contain-
suggest that a different exposure would capture more ing tones of dark black, bright white and everything in
scene information. between. A low dynamic range scene appears as a narrow
In addition to overexposure and underexposure, the histogram that tapers off well before reaching either end
tonal range plot gives a sense of scene contrast and of the plot. While these characteristics may be inherent to
dynamic range. Figure 4-11 (overleaf) showcases varied the scene, it’s useful to identify, for example, if your studio
examples. We expect that, in an average scene, the lighting is overly flat and failing to generate intensely bright
histogram clumps toward the center (midtone) region highlights.
of the plot. Unless it’s a high key scene of a white rabbit It’s valuable to point out that the y-axis of a histogram,
in the snow, the presence of bright highlights and dark representing pixel frequency, is nearly always scaled to the
shadows is generally outweighed by middle-gray subject data. This means that two images can show high frequency
reflectance. A histogram with clumps or peaks at either peaks around the same pixel value—approaching the top
end of the histogram and a valley in the center is a high edge of the plot—yet have a different number of pixels at
contrast scene because it has a relatively large amount of that value. The y-axis is scaled depending on the data for
very dark and very bright pixels. A low contrast scene looks readability.
like a single pile or peak near the center that tapers off on
either end. This indicates that most of its pixels have similar
luminances rendered as midtones. A wide histogram with
some amount of pixel frequency across the whole range of

111
 s ecti o n 1: Captu r e

Figure 4-11 Image histograms reflect the range of pixel values that comprise the scene and their shape is dependent on the image content.
These images have different histograms that indicate variations in contrast, the range of tones and even the dominant brightness value.

112
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

In-Camera versus Desktop Software Exposure value is a useful metric for describing exposure
Histograms changes that are easily translated to stops. EV is often used
An image histogram can look different when viewed in exposure bracketing notation, discussing flash compen-
in-camera compared to the histogram generated by sation and measuring lighting ratios. Many light and flash
Adobe Photoshop or Capture One Pro even when shoot- meters can measure in EV.
ing raw. Why does this discrepancy exist? A raw image
must be processed through a demosaicing algorithm to
produce a full color image. We address demosaicing in Equivalent Exposure
Chapter 9; for now, let’s accept that there are different
algorithms and methods for doing so. The resulting histo- So far, we’ve laid out the adjustable parameters for setting
grams are different due to the different image processing photographic exposure and how to evaluate histograms
decisions made. for optimal image capture. Shutter speeds, apertures and
The histogram viewed on the camera is a histogram rep- ISO are used to gather the necessary amount of light to
resenting a processed JPEG interpretation of the raw image form a well-exposed photograph. We’ve elaborated on
content. Again, each camera manufacturer uses their own each of these parameters because there’s a great deal
algorithm for this step which is most likely different than of subjective and interpretive control involved in setting
the desktop software companies, therefore a difference them: there’s more than one solution to satisfy the goal of
may exist. The only foolproof way of knowing what your optimal photographic exposure. The concept of equiva-
image data histogram consists of requires loading the files lent exposure means recording the same amount of light
into desktop editing software. using different combinations of photographic exposure
Image histograms play a pivotal role in post-processing controls.
techniques. We revisit them in Chapter 9 to ensure that the Table 4-1 demonstrates how six unique captures of
hard work of properly exposing at capture is not under- the same scene are made using equivalent exposures.
mined by haphazard editing decisions. We determined Image A’s exposure using a handheld
light meter and confirmed optimal exposure by checking
the resulting image histogram. This baseline exposure
Exposure Value (referred to in shorthand as simply “EV”) is 1/60s at f/5.6
and ISO 800. Hypothetically, we’re interested in increas-
Similar to the concept of stops, exposure value, or EV, is ing our shutter speed to better freeze motion in the
a measure of exposure that’s used as a shorthand when scene. Changing the shutter speed to 1/125s accom-
adjusting exposure settings. EV is defined by Equation plishes this, however, exclusively changing the shutter
4.1 where N is the f-number and t is the shutter speed in
seconds.
Table 4-1 Examples of equivalent exposures using different
N2 combinations of camera controls.
EV = Log2 (Eq. 4.1)
t
Image ISO f-number Shutter Speed
A camera exposure setting combination of f/11 at 1/15th A (EV) 800 5.6 1/60
second yields an EV of 11. The same EV is achieved using B 800 4 1/125
f/8 at 1/30th second. Altering the first combination to f/11
C 800 8 1/30
at 1/30th second changes the EV to 12. Changing shutter
D 1600 8 1/60
speed from 1/15th to 1/30th second is a one stop change
in photographic exposure. The exposure value calculation E 1600 16 1/15

shows this same change equating to a change in EV of 1. F 1600 5.6 1/125

113
 s ecti o n 1: Captu r e

speed parameter means underexposing the scene by The Image Capture/Exposure

one full stop (EV-1). This change is compensated or bal- Sequence
anced by opening up the aperture one full stop to f/4.0.
The resulting exposure of Image B is equivalent to Image In order to highlight features like autofocus and auto expo-
A in brightness because we’ve maintained equivalent sure in this chapter, it’s helpful to take a bird’s eye view of
photographic exposure. Equivalent exposures are all the image processing pipeline, the path and processes our
about balancing the equation of adding and subtract- photograph undergoes before, during and after pressing
ing exposure and is easily managed using stops as the the capture button. A generic pipeline looks like this:
metric.
Looking again at Table 4-1, Images B through F yield 1. Clear voltages from photosites
photographic exposures with the same overall brightness 2. Open shutter or turn on photosites
as our baseline photo, though they differ in other ways. 3. Read out voltages
Image E, for example, becomes noisier than Image A as a 4. Amplify voltages
result of a higher ISO; it shows considerably more depth 5. Analog to digital conversion from voltages to bits
of field and may require the use of a tripod due to its long 6. Store bits in temporary random access memory buffer
shutter speed. Alternatively, Image B is a good option if 7. Transfer and store bits to storage medium.
the subject is on a single focal plane and the background
is distracting. It also doesn’t require having a tripod Note that all of the these events take place in the camera.
handy. Photographers use the term processing to describe making
This technique is further extended for considering changes to image data, though they’re usually concerned
overexposure or underexposure bracketed frames. Using with post-processing that happens off-camera (discussed
Image A again as the baseline exposure, imagine that in Sections 2 and 3). Here we’re talking about the digital
you’re looking to capture one frame that’s very overex- churnings of onboard components to formalize a raw digi-
posed (EV+3) and another that’s equally underexposed tal image stored to memory.
(EV-3) to record highlight and shadow information A camera’s image processing pipeline may include addi-
beyond what the sensor’s dynamic range can capture in tional steps between the analog to digital conversion and
the baseline exposure. Here, the goal isn’t to formulate an storage that introduce post-processing changes. Setting a
equivalent exposure but one that is specifically three full camera to shoot in JPEG, for example, implicitly appends
stops over or under relative to Image A. For the overex- the image processing pipeline with post-processing adjust-
posed frame, we can increase ISO three stops or perhaps ments to the raw image to make it aesthetically pleasing.
leave ISO the same; opening the aperture one stop and This is because a JPEG is assumed to be a finished, fully
slowing the shutter speed by two stops nets an exposure cooked image product with enhancements to tone, color
of EV+3. The key is to consider the starting point and and sharpness and other characteristics. Surprisingly, set-
how changes to the three exposure parameters impact ting a camera to shoot in the raw format does not exempt
exposure relative to it. It’s best to alter shutter speed over us from seeing automatic post-processing decisions to our
the other two variables, when possible, as it is less likely to image; the preview on the camera’s playback screen is an
introduce visual differences (noise, differences in depth of enhanced JPEG derived from the raw data. It’s easier for our
field) between the bracketed frames. cameras to do a little hard work up front in interpreting and
enhancing a version of the complex raw data and showing
us the fruits of its labor any time thereafter when reviewing
what we’ve shot. This means that we only see a version of
what the raw image could look like, subject to the camera’s
onboard automatic post-processing decisions.

114
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

When a camera uses live preview or digital viewfinder,

a version of an image processing pipeline happens con-
tinuously on the device. This is the case when taking
photographs on a smartphone; the display shows a near-in-
stantaneous, processed feed of images as they’re recorded
(though not permanently saved) by the camera module.
Expediting the image processing pipeline to allow for this
incredibly fast translation from light detection to processed
image is made possible by a system-on-chip (SoC) image
processor. The SoC lives in the camera or phone hardware
and is dedicated to the task of real-time image processing
from the image sensor. Feedback about exposure, focus
Figure 4-12 Shooting a gray card in the scene as a test frame
and color balance is generated as the image data works its helps to evaluate exposure at capture. It should not appear clipped
way through algorithms immediately following their analog (too bright) or approach black (too dark) if it represents the average
scene luminance. Photograph by Rochester Institute of Technology
to digital conversion. This information is piped back to the photography student Daniel Lemaster
sensor controller to make adjustments for the next expo-
sure, over and over until a photograph is taken by the user.

Using a Neutral Reference Target In-Camera Exposure Metering

An 18% gray card is a uniform target of midtone gray, or We showed how a camera internally takes light meter
a perfect neutral midpoint between black and white. Its readings in the previous chapter. We described the
visible light reflectance is 18%. Achieving a proper exposure hardware components that make it possible but not the
is easy using an exposure meter and a normal scene with software logic behind its operation. Since the camera is
an average midtone gray tonality, or a reflectance of 18%. capable of taking multiple readings, continuous readings,
The exposure meter is calibrated to this same reflectance or readings from different locations in the frame, it opens
and identifies the exposure settings required to record it up possibilities for many different methods for metering.
on the sensor. When a scene’s average reflectance is darker Through the lens (TTL) metering uses a light meter some-
or lighter than this, following the exposure meter reading where in the pentaprism or below the mirror assembly
leads to overexposure or underexposure, respectively. A in the camera to instantly read the light level entering
logical strategy, then, is to include a reference target con- the camera. This approach is advantageous because it’s
sistent with the exposure meter’s calibrated expectations. metering the light after it passed through the camera
A neutral target with an 18% reflectance is placed in the lens and any filters or optical elements introduced in the
scene and metered to avoid exposure errors as demon- imaging path. The result is an accurate measurement of
strated in Figure 4-12. the light near the point where it is detected by the sensor.
The ColorChecker used by photographers includes an In short, it provides a meter reading that fully considers
18% gray patch: the fourth patch on the last row. Using the impact of everything between the subject and the
a ColorChecker to determine exposure is possible with sensor. TTL metering is what makes all the fully automatic
spot metering and is additionally useful as reference for or partial automatic shooting modes function properly.
adjusting white balance in editing software. The target’s There are many manufacturer specific variations of TTL
neutrality makes it valuable for identifying color casts in metering available, though they are generally one of three
addition to its light reflectance properties. basic types: matrix, center-weighted and spot. Figure 4-13

115
 s ecti o n 1: Captu r e

Figure 4-13 Generic in-camera metering patterns: matrix, center-weighted and spot.

illustrates generic in-camera metering patterns. The small Spot Metering

boxes represent light measured zones of the image frame The spot metering mode takes a reading from one small
and the gray overlay indicates the weight given to those point in the frame, with the default position (until defined
zones. differently by the photographer) at the center. The size of
the spot ranges from 1° to as high as 10°. This conveniently
simulates the use of a dedicated handheld spot meter
Matrix or Evaluative Metering whereby the photographer targets or aims the metering
Matrix or evaluative metering is most often the default zone at the subject. Spot metering is so narrow in its scope
metering mode on SLR cameras. The frame is split into that it is best used to meter and recompose the shot
multiple zones. Each zone is then evaluated individually through the viewfinder.
for highlights and shadows. An average of these zones is
then determined placing extra weight on the zone where
the scene is focused. The exact design, arrangement and Exposure Compensation
prioritization of these zones is unique to the camera model The settings for a good exposure are found using one of
and its programmed firmware. the aforementioned strategies. They help us record an
image whose histogram avoids clipping at either end
whenever possible. There are some scenarios where we
Center-Weighted Metering want to take the camera’s metering results and go one step
Center-weighted metering measures the entire frame, further. A good exposure is not always the right exposure.
however, considerably more weight is given to the light Here, exposure compensation settings on a camera allow
requirements of the center of the frame. Some manufactur- us to inform the camera to expose exactly one half stop,
ers ignore the corners or the outer edges of the frame. The or maybe two full stops, over its normal recommendation.
success of this metering mode is closely tied with the pho- This is useful when a particularly reflective subject may fool
tographer’s compositional strategies and should only be the metering into suggesting an overly dark exposure. Or
used when the subject is in the center of the frame. Unlike perhaps the photographer has an aesthetic motivation for
focus that can be set by depressing the shutter button making a series of portrait exposures brighter than normal.
halfway and then recomposing the images, the exposure is
set when the image is taken.
Automatic Exposure Behaviors
There’s no shame in leaning on automatic exposure in a
variety of contexts where the photographer’s focus and

116
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

energy might be better spent elsewhere. Its predictable Spot metering is tricky for many photographers. Recall
behavior can be piggybacked with intuition or creative that a meter is calibrated to an 18% reflectance. It takes
intent to achieve the best exposure for the shot. However, experience to properly identify an object that is an 18%
automatic exposure is not a perfect tool and it can be reflector. We recommend taking a spot meter reading from
fooled. There are four common automatic exposure modes a calibrated gray card placed in the scene.
or behaviors on SLR cameras: Full Auto, Aperture Priority,
Shutter Priority and Program.
A Full Auto mode selects all exposure settings: shut- Alternative Metering Methods
ter speed, aperture and ISO. It may prioritize keeping ISO
low or it may use other algorithmic decisions (including Smartphone apps offer an interesting alternative to in-cam-
motion detection) to find a combination that yields a era and dedicated handheld meter equipment. After all,
proper exposure. Aside from exposure compensation, the our phones have capable camera apps themselves and
photographer does not have control over photographic are able to take well-exposed photographs with much
exposure beyond pressing the shutter button to cap- less hardware than what is found in a photographer’s bag.
ture. The automatic exposure mode is typically coupled From this, we highlight the humble light meter app: using
with automatic white balance, flash and other settings to the smartphone’s imaging sensor and some clever soft-
completely minimize the photographer’s need for camera ware, a poor man’s incident light meter is born (see Figure
setting decision-making. 4-14). Our students have tried several of the available apps
Aperture and Shutter Priority exposure modes are worth- and found that they perform relatively well under normal
while when you want more control of image characteristics shooting circumstances. This strategy can be further aided
without manually juggling all three variables. Aperture Prior- by using half of a ping pong ball over the phone camera as
ity allows user-set f-number and ISO to prioritize control of an improvised dome diffuser substitute.
depth of field; the camera automatically sets shutter speed
to achieve a proper exposure. In Shutter Priority mode, the
user sets shutter speed and ISO to prioritize freezing or
blurring motion; the camera sets the appropriate aperture.
Program mode is similar to these, sometimes prioritizing
ISO, but additionally allowing for user intervention to adjust
any of the three variables after initially setting suggested
values.

Handheld Light Metering

Take care to avoid metering off of specular highlights on

reflective surfaces when spot metering. These areas are
best left overexposed and if measured for overall exposure
will likely leave the scene underexposed. Specular high-
lights are significantly brighter than the light reflected off
of the rest of a surface, which scatters a percentage of the Figure 4-14 A dedicated light meter (left) may offer better accuracy
and additional features over a smartphone metering app (right).
incident light. Tiny specular highlights like those seen in
the eyes of a portrait’s model appear most natural when
they are rendered as bright and overexposed.

117
 s ecti o n 1: Captu r e

Synchronizing Flash and signal on the sensor. This is limiting, of course, because you
Camera Exposure can no longer see what you’re shooting as you stand in the
dark.
Photographic exposure can be accomplished with the Rear curtain synchronization is an exposure mode that
assistance of artificial or otherwise controlled lighting in a allows for longer shutter speeds to mix ambient light and
scene. Adding your own lights means greater control of flash. Using rear curtain synchronization and setting the
scene dynamic range, overall scene brightness and light shutter speed to 1/4th second, for example, means that
characteristics (low contrast, high contrast, soft, harsh, etc.). the exposure is long enough to build signal from low levels
Working in the studio with strobe equipment or shooting of environmental light that otherwise wouldn’t at the
in the field with handheld flash units requires communi- traditional 1/125th sync speed. Just before the rear curtain
cation between the camera shutter and flash. The flash begins its movement to block the sensor from further
trigger is communicated through a camera’s hot shoe con- exposure, the flash is triggered to introduce additional,
nector or through a dedicated PC sync port when a shutter artificial light to the scene.
press initiates an exposure. Some cameras offer built-in
radio or infrared transmitters to communicate wirelessly
with flash units. Exposing to the Right
Focal plane shutters necessitate precise flash synchroni-
zation in addition to a limited maximum shutter speed to Unlike the human eye, camera sensors have a linear
avoid partial frames (illustrated in Chapter 3). The shut- response to light. This means that the more energy gath-
ter speed must be set such that both the front and rear ered and output as a voltage by the photosites, the more
curtains are out of the sensor’s way. This typically happens signal there is to create a properly exposed photograph.
around 1/125th second and is therefore a functional sync A bright area in an image always exhibits less noise than
speed for flash exposure. Some cameras offer faster sync a shadow area; it simply had more light from the scene
speeds of 1/250th second. from which to output an electrical signal. When the
Flash units can be adjusted in terms of stops, EV or guide signal greatly outweighs the noise, the result appears
numbers. Guide numbers are calculated as the subject noise free. When the reverse is true, the noise distracts
distance multiplied by the f-number used at ISO 100. This and obscures detail. It is this behavior that leads us to
system is not perfectly accurate and assumes that the flash the expose to the right (ETTR) method of photographic
is a point source that behaves according to the inverse exposure.
square law. Using guide numbers is relatively out of fashion Exposing to the right means taking the brightest pos-
given our reliance on digital camera hardware to estimate sible exposure that falls just short of clipping highlights. If
subject distance. the traditional method of exposing for the midtones yields
In a pinch, there is one workaround that allows for flash a histogram largely clumped at its center, ETTR yields a
exposure without precise synchronization. Keep this idea in histogram pushed to the right of the plot. This strategy
your back pocket for the day when your strobe transmitter aims to collect as much photon signal as the scene lighting
batteries die or you forget a sync cable: turn out the lights. allows to take advantage of the sensor’s high bit depth
If you can find a completely light-controlled environment, recording. The image signal is cleaner and more detailed
turn out the lights and set the camera shutter speed to the more light we’re able to feed to the sensor (accom-
bulb or some multi-second duration. Press the shutter plished by maximizing camera exposure).
button to start exposure, fire the flash manually and press Many cameras have automatic exposure algorithms
the shutter button again to end exposure. The lack of syn- that function in accordance with the ETTR rule. With
chronization is not a problem as long as the open shutter manual exposure, however, we review the in-camera his-
time before and after the flash firing does not build up any togram and push it as far to the right as possible without

118
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

significant clipping. The resulting image may appear in Figure 4-15 (overleaf). Combining information from
slightly overexposed according to the camera’s preview these exposures is handled by image processing software
screen, however, adjusting the file in post-processing that looks to minimize clipping and map a wide dynamic
yields an image with good separation in the shadows with range of tones into a range supported by our output
minimal noise. media. We can also manually mask and composite multi-
This approach only works with scenes containing pri- exposure brackets for a fully hands-on approach to HDR
marily dark or middle tones with limited tonal ranges, since imaging.
any other scene would quickly clip in the highlights. ETTR A related exposure strategy is shooting a sequence of
is more likely to yield better image quality results when frames while keeping all exposure settings the same. This
accomplished with aperture and shutter speed rather than means there’s no exposure bracketing going on, however,
ISO. Recall that boosting the gain increases noise, which is every one of the sequence’s frames will have its unique
what ETTR works to minimize. variation in random noise. A set of six exposures taken one
after another, for example, can be leveraged by stacking
the image data from all six. This is an option in Adobe
Bracketing Exposure Photoshop whereby image layer stacking or blending
can function as an extremely effective noise reduction
Bracketing means taking two or more photographs of the process. The random noise from a single frame is replaced
same subject, altering the camera settings between them by true image pixel data from a different frame. The more
to capture different exposures of the same scene. This frames the algorithm can look at, the more confident the
exposure technique is primarily used to ensure proper final output becomes where a larger percentage of pixels
exposure and is also used for high dynamic range (HDR) are true brightness and color information from the scene.
imaging. Exposure bracketing is not new to digital imaging, This is a great approach when shooting static scenes on a
as even the earliest wet plate analog photographers took tripod.
multiple frames and combined them in the darkroom. Digi-
tal cameras today have automatic exposure bracketing (AEB)
modes that take a baseline exposure followed by two more Setting White Balance
images at a variable stop increment over and under the ini-
tial exposure. Bracketing a scene is often successful when The visible light characteristic of color temperature was
altering shutter speed, as altering the f-number changes introduced in Chapter 1. Every light source appears warm,
the depth of field and altering ISO potentially changes the cool or neutral to our eyes. Evaluating the color tempera-
amount of noise. AEB typically defaults to bracketing shut- ture of the light in a scene was critically important in the
ter speed as a result. analog film days because a particular color positive film
HDR bracketing and post-processing are employed was explicitly formulated for a single, primary light source
when photographing a scene with a dynamic range like daylight or tungsten. A photographer loaded their
that is larger than what a camera can capture in a single camera with film designed to shoot under the light source
exposure. Capturing the bright sky and clouds in a present in the scene to achieve accurate, balanced color.
landscape scene is likely to underexpose the foreground. If the two were mismatched, the resulting images would
Exposing for the foreground overexposes the sky and exhibit a significant color cast where all colors exhibit
clips details in the clouds. Bracketing the scene means an obvious tint or skew toward a single color. The other
taking each of these exposures in turn. This might be as option involved using color correction filters over the lens
few as two frames or as many as five. This ensures that the to compensate. Color negative film lacked this degree of
scene’s highlights, midtones and shadows are properly control at capture, opting for a daylight balance and relying
exposed in at least one frame from the sequence as seen on corrections in color print reproduction.

119
 s ecti o n 1: Captu r e

Figure 4-15 Bracketed exposures and the combined HDR composite. Photographs by Rochester Institute of Technology alumnus and Lecturer
Dan Hughes

120
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

The color temperature of a scene’s light is considered white balance to use? It’s a subjective decision, but it’s
in order to find a desirable white balance whereby white or common to see the dominant light source made neutral.
neutral objects are rendered as such. Camera sensors are It’s difficult to get all areas to be neutral with a global
designed to record light regardless of color temperatures, correction, but correcting for the dominant light source
so in this sense, it’s not necessary to set a white balance is a notable improvement. Other times, you may prefer
in the camera before taking a picture. The concept still or desire a considerably warmer or cooler cast to the
matters because we want the resulting image to show the dominant light source to convey a mood or exaggerated
colors and color relationships as they appeared in person. interpretation of the scene.
However, we can set the interpretation of color neutrality Generally speaking, an improper white balance appears
in image editing software long after the image is recorded warmer (more yellow-orange), cooler (more blue) or, less
with no ill effect to quality (assuming raw capture). The commonly, greenish (due to fluorescent light sources).
color of digital images is interpretable in the raw exposure Cameras offer a set of predefined color temperature
data whereas film chemistry was formulated to record options that cover most of the bases: incandescent
accurate color specific to a tungsten, daylight or fluores- (tungsten), fluorescent, daylight, shade/overcast daylight,
cent color temperature. flash. Selecting one of these generic options often gets
There are many instances where a scene contains a the image preview in the ballpark of where the white
variety of lights sources, making it impossible to render balance needs to be in order to produce a neutral white.
each one or each area illuminated by those sources By virtue of being predefined color temperature categories
perfectly neutral. This is a simple reality, neither good in a world of infinitely varying light quality, it won’t hit the
nor bad—we’re accustomed to seeing this, since our mark every time. There is no single-color temperature for
vision similarly calibrates for a single white point. Figure daylight on an overcast day, for example.
4-16 shows an indoor scene containing multiple light Fully automatic white balance functionality is always
sources with different color temperatures. You can quickly offered in addition to the generic light source presets. Algo-
determine which white balance setting was used by rithms work to make the best guess as to the dominant light
the photographer by identifying the light most closely source by analyzing image pixel data and color distribution.
approximating a neutral white tone. What is the correct Automatic white balance is a helpful default when shooting

Figure 4-16 An environment with mixed illuminants using the camera’s automatic white balance (left) and after a user-defined white
balance for fluorescent lights (right).

121
 s ecti o n 1: Captu r e

Figure 4-17 An automatic white balance result (left) and its user-corrected rendering (right).

under dynamic shooting conditions when it’s often enough Proof of Concept:
work juggling exposure settings, focus and composition. Optimizing Photographic Exposure
Image editing software additionally offers its own strategies with Low Light Shooting
for determining white balance automatically when review-
ing and editing raw files. Instances that are likely to trip up Night photography offers a number of challenges, par-
automatic white balancing algorithms include: ticularly in regard to lighting and exposure, that aren’t
present during the day. Nighttime also makes the world a
• extremely skewed dominant light sources (i.e. sodium very different looking place compared to when the sun is
vapor street lamps); shining. Photographic exposure is versatile with the right
• when the image frame is filled with a single color (i.e. a tools, allowing you to make images under incredibly var-
pile of red autumn leaves) or the scene does not contain ying conditions. It allows us to create documents of time
any neutral objects; that extend beyond the immediate, experienced present
• mixed illuminant environments. moment. This is particularly intriguing given our poor low
light vision. Use a tripod and your knowledge of camera
Figure 4-17 is an example of a composition dominated exposure (shutter speed, aperture and ISO) to explore the
by a single color where automatic white balance failed. world at night.
The lack of a neutral reference and a dominant set of
warm hues in the pile of leaves caused the algorithm to 1. Set the camera to collect raw images. Turn off any
overcorrect toward something much cooler. As algorithms in-camera noise reduction.
become more robust and better implementations trickle 2. Find and compose an image in a low light environment
down to even the cheapest of cameras, such misfires with the camera mounted to a tripod.
become less frequent and less egregious. 3. Set the camera to its lowest marked ISO. Use the
White balance is a non-destructive parameter that can in-camera meter and perhaps some trial and error to
be changed after the photograph is taken with no nega- get a good exposure (consider what constitutes a good
tive consequence to image quality or color reproduction. exposure and use the histogram). Be careful: the meter
We tend to keep our cameras on auto white balance is easily fooled by the challenging lighting conditions of
save for special scenarios and adjust as necessary during nighttime scenes. If you do not have a cable release, set
post-processing. the camera’s self-timer to allow for a delay between the

122
 c h a p t e r 4 : P h oto g r a p h i c E x p o s u r e

Table 4-2 Exposure settings for low ISO and high ISO captures.

Image ISO Shutter Speed Aperture

Low ISO Proper Exposure (EV) 100 1s f/11

Low ISO Under Exposure (EV-2) 100 1/4s f/11

High ISO Proper Exposure (EV) 6400 1/60s f/11

High ISO Under Exposure (EV-2) 6400 1/250s f/11

shutter button press and the shutter opening. Take an the underexposure versions? Is this relationship consist-
image. ent across the ISO settings?
4. Take a second image that is two full stops underex- • When shooting at the highest available ISO, does your
posed (EV-2) at the same ISO setting. camera produce an acceptable image?
5. Change ISO to the highest that your camera offers. • What is the downside, if any, to using the lowest availa-
Adjust shutter speed to compensate for this increase in ble ISO setting?
sensitivity to create an exposure as close as possible to
the original, properly exposed image.
6. Take a second image that is two stops underexposed
at the same high ISO setting. Table 4-2 documents the
exposures used for our example nighttime scene.
7. Back at your workstation, transfer the images and
import them into your image editing software of choice.
Figure 4-18 shows one pair of our proper exposure
and underexposure frames. Correct the underexposed
images using the exposure slider. The exposure slider
in Adobe Photoshop Lightroom and Adobe Camera
Raw corresponds to changes in EV. Since our images
are underexposed by two stops, increasing the expo-
sure slider by +2.0 produces an equivalent brightness
correction.

Look at the two ISO exposure pairs side by side and

zoom in close to areas of shadow. Consider the following
questions:

• What is observable in each ISO pair? Are they

comparable in image quality after the underexposed
versions are edited to equivalent exposures in
post-processing?
• Is an equivalent exposure at capture better or worse
than a different exposure at capture made equivalent in Figure 4-18 A low light, nighttime scene photographed with a proper
exposure (top) and underexposed by two stops (bottom). Photographs
post-processing? by Rochester Institute of Technology photography student Matthew
• Is there more visible noise in the correct exposures or Sluka

123
 s ecti o n 1: Captu r e

Figure 4-19 Before and after thoughtful noise reduction is applied to our nighttime photograph. Photographs by
Rochester Institute of Technology photography student Matthew Sluka

Examine all photographs zoomed to fit the screen as well Note

as at 100%. Look closely at areas of detail and darker tones. 1 https://anvaka.github.io/pixchart/
Start to get familiar with digital noise: its texture (luminance
variability or grain) and its color characteristics. We mitigate
the negative visual impact of noise using noise reduction
tools in software. Work with the noise reduction options
in your image editing software to see how much the high
ISO, underexposed frame can be improved without intro-
ducing too much smoothing or blurring. Figure 4-19 shows
the before and after results of our careful noise reduction
choices of a high ISO capture.

124
 5 Lens Filters

Photograph by Rochester Institute of Technology photography student Eakin Howard

Photographers rely on lenses to form sharply resolved images. Their precision, engineering
and materials often necessitate expensive purchases to add lenses to the photographer’s
toolkit. Filters are accessories that modify or adapt the optics of a camera system and, thank-
fully, augment our image-making capabilities for cheap. All lens filters change aspects of the
light entering the camera. Software processing offers simulations of a subset of lens filters
that can be introduced after capture though many filter behaviors cannot be emulated. Our
interest in controlling light at the capture stage means considering lens filters as carefully
as the camera body, lens focal length and exposure settings. In other words, lens filters are

125
 s ecti o n 1: Captu r e

an irreplaceable part of pre-planning and pre-visualization.

This chapter explains how light is filtered, how it impacts
exposure and the different ways in which filters are used
in making photographs. At its end, we take a creative look
at reproducing color from monochrome captures with the
help of lens filters.

Transmittance, Opacity and Density

All photographic filters reduce the amount of light entering

the lens. There are three measurements for quantifying this Figure 5-1 The relationship between the sample thickness and
transmittance, opacity and density.
behavior: transmittance, opacity and density. These meas-
urements apply to light traveling through all materials; here
we’re interested in light filtration materials in the camera ratio of incident light to transmitted light or the reciprocal
system. of the transmittance. It is calculated using the following
Transmittance describes a material’s ability to pass light. formula:
It’s calculated as the ratio of the light energy falling on a
Incident Light 1
filter to that which is transmitted through it (see Equation Opacity (O) = or (Eq. 5.2)
Transmitted Light Transmittance
5.1). Put another way, transmittance describes the relation-
ship of how much light we start with and how much we Take another look at Figure 5-1. As sample thickness and
see after it passes through a medium (the filter material). light-stopping ability increase, opacity also increases—a
more straightforward relationship. Notice that sample
Transmitted Light thickness increases in equal amounts yet the opacity dif-
Transmittance (T) = (Eq. 5.1)
Incident Light
ferences become progressively greater. The opacities form
Consider the scenario illustrated in Figure 5-1 where a geometric (or ratio) progression that becomes incon-
100 lux of incident light travels through a sample material. veniently large as the light-stopping ability increases. If we
Only 50 lux is transmitted from the 100 lux of incident light. continued the experiment, ten layers of the material would
Equation 5.1 tells us that the transmittance of the mate- yield an opacity of 1,024. Large numbers can become dif-
rial is 50/100, which equates to 0.50 or 50% transmission. ficult to deal with, so this might not be the perfect way to
Next, we layer two pieces of that same material, measure keep track of a material’s light-stopping ability either.
the amount of transmitted light from the same 100 lux This brings us to yet another method of measuring
source and find that 25 lux passes through. The light-stop- light transmission. Density is a measure of the extent that
ping ability increases while the transmittance decreases a material or medium transmits light. Density is a valuable
as layers of material are added between the light source metric for describing how light travels through a material
and the transmission measurement. Transmittance values because it is linear and additive. Density is calculated as the
get increasingly small as more layers of the sample mate- logarithm of opacity:
rial are stacked. Even when using percentage to report
1
transmittance, the very small values make it challenging to Density (D) = log (Opacity) or log Transmittance
(Eq. 5.3)
appreciate the magnitude of change as the light-stopping
ability of a material grows. Review Figure 5-1 one last time. Density increases as
Opacity is a measure of the lack of transparency and is we layer additional material. However, its values increase
inversely related to transmittance. Opacity is defined as the in equal amounts in a linear fashion, a feature that

126
 c h a p t e r 5 : L e n s F i lt e r s

Figure 5-3 Light falling on a surface is (a) reflected, (b) absorbed,

(c) transmitted or (d) scattered before reaching the surface.

surface. Common mnemonics for remembering these

behaviors are ARTS (absorbed, reflected, transmitted,
scattered) or RATS. Scattered light is the behavior of least
concern to photographers other than the fact that it can
introduce flare or an inefficiency in our light sources if, say,
we have a photoshoot in a foggy graveyard. An object,
such as this book, absorbs some of the light striking it while
reflecting the rest. A stained glass window, on the other
hand, absorbs, transmits and reflects varying amounts
of incident light. We don’t need to know or observe the
exact degree to which these different events occur with
the objects we photograph. We do need know the degree
Figure 5-2 Plotting (a) transmittance, (b) opacity and (c) density to which they occur in materials placed over the camera
measurements from layering sample material in front of a light source.
lens (read: filters). Lens filters help control the types of light
passing through our imaging system by leveraging the
differentiates it from opacity measurements. To summa- reflective, transmissive and absorptive characteristics of
rize: the less light transmitted by a material, the lower its optical materials.
transmittance, the higher its density and opacity. Figure The perceived color of any object depends on the
5-2 illustrates these relationships by plotting the measured spectral quality of the reflected or transmitted light by that
values. The density plot shows that density measurements object. An opaque white object illuminated with white
are linear. light appears white because the surface reflects a high
portion of the incident light at all wavelengths. A black
object appears black because it reflects only a small por-
How Light is Filtered tion of the incident light at all wavelengths. A red object
appears red when illuminated with white light because
Light approaching a filter may scatter before ever reach- it selectively absorbs most of the blue and green wave-
ing its surface. Consider shining a flashlight across a foggy lengths striking it while simultaneously reflecting most of
graveyard: some of that light never reaches the rustling the red wavelengths. Here we’re grouping wavelengths
bushes you point it toward and instead scatters in the thick, into large categories, but there are hundreds of possible
moisture-rich air. If the light does travel to a surface without wavelengths in the visible spectrum that can be reflected
scattering first, it is then reflected, absorbed or transmitted or transmitted.
(or a combination therein) by or through the foliage. Figure Photographic filters work by removing part of the
5-3 illustrates the possible light paths as it approaches a incident radiation by absorption or reflection such that

127
 s ecti o n 1: Captu r e

acting as darkening filters. The primary examples are

neutral density filters, discussed in a section to follow.
3. Selective by angle of polarization. These transmit light
oriented in a specific direction. Polarizing filters are the
primary example and are described in a later section.

The Filter Factor

Figure 5-4 A green filter appears green because the filter absorbs
the red and blue components of white light and transmits the green.
Any filter placed in front of a lens absorbs light. A notable
reduction in light requires a change in camera exposure to
compensate else you risk underexposure. The filter factor
specifies the change in exposure necessary to obtain the
same recorded brightness in a neutral subject area with
the filter as without. Filter manufacturers often include filter
factors on their product packaging to save us the guess-
work. Table 5-1 lists common filter factors and the exposure
difference rendered by each in stops.

Figure 5-5 Blue light has no red component and therefore using a Table 5-1 Filter factors and corresponding number of stops of
red filter to transmit only red light yields no visible result. reduced exposure.

Filter Factor Number of Stops

1 0
the transmitted radiation is of the desired spectral quality.
Filters cannot add to the incident radiation, only take away 1.3 1/3

from it. A green filter passes green light and absorbs red 1.4 1/2

and blue as seen in Figure 5-4. A red filter cannot transmit 1.5 2/3
red light if there is no red light in the incident light. If a 2 1
green filter is used when only blue light is present, no light
3 1 2/3
passes through to the sensor or eye (see Figure 5-5).
4 2
There are various ways to classify filters. We consider
three basic categories defined by their behavior in trans-
mitting incident light. A filter can be: The relationship between filter factor and the required
number of stops by which the exposure must be adjusted
1. Selective by wavelength. These block specific wave- is defined in Equation 5.4. For example, we solve the equa-
lengths or ranges of wavelengths (such as all shades of tion for a filter with a filter factor of 16 to determine how
red). This category is easily represented by color filters many additional stops of exposure are needed:
that simply transmit a single color; think rose-colored
glasses that make everything seem rosy (in appearance, 2number of stops = filter factor (Eq. 5.4)
not emotional optimism). It also includes filters that 2number of stops = 16
selectively transmit other forms of radiation like infrared.
2. Non-selective by wavelength. These reduce the amount of We then take the natural log (ln) of each side of the
transmitted light equally across wavelengths, effectively equation:

128
 c h a p t e r 5 : L e n s F i lt e r s

number of stops x ln(2) = ln(16) (5.4) without the filter. The difference in exposure readings is
ln(16) compared against the published filter factor and adjusted
number of stops =
ln(2) as necessary.
number of stops = 4

An easier method is to divide the logarithm of the filter Color Filters

factor by 0.3 (see Equation 5.5). Recall that 0.3 in neutral
density is equivalent to a one stop change in exposure. Color filters are commonly identified by the perceived color
Continuing from the previous example with a filter factor of the transmitted light when illuminated by white incident
of 16: light. In Figure 5-4, the green filter transmits green light and
absorbs red and blue light. Color filters differ with respect
log(filter factor)
number of stops = (Eq. 5.5) to a) which part of the visible spectrum is transmitted freely
0.3
(identified by color hue or peak transmittance wavelength);
log(16)
number of stops = =4 b) the width of the region transmitted and the sharpness
0.3
of the cutoff (identified by a general description such
If the filter factor adjustment is applied to the exposure as narrow, wide or specific bandpass); c) the degree of
time without a filter, the new exposure time is found by absorption of the unwanted colors (identified by general
multiplying the original exposure by the filter factor. For terms such as light yellow and dark yellow, or specific trans-
example, with an initial exposure time of 1/500 second mittance and density data).
and a filter factor of 16, the new exposure time is 1/500 * Saturated color filters that almost completely absorb
16 = 1/30. If the filter factor is instead applied to the aper- the unwanted colors of light are used in black and white
ture setting and the initial f-stop is f/8, the adjusted f-stop photography to lighten or darken subject colors. Photogra-
for maintaining equivalent exposure when using the filter phers need to predict the effect that such filters have on
is f/2. the tone reproduction of a subject. One method is to look
The published filter factor may not always produce at the subject through the filter and observe which colors
optimal results. Manufacturers commonly indicate that are lightened and which are darkened.
published filter factors should be modified by the pho- The Maxwell triangle is a diagram containing the three
tographer as necessary. Two additional aspects of the additive primary colors red, green and blue and the three
photographic process to take into account are the color additive secondary colors cyan, magenta and yellow (see
quality of the illumination and the color sensitivity of the Figure 5-6). The triangle helps predict the effect of color
digital sensor. Red filters, for example, have larger filter filters on tone assuming that the image is rendered in black
factors with daylight illumination than with tungsten illumi-
nation because they absorb a larger proportion of the bluer
daylight illumination. Conversely, blue filters have smaller
factors with daylight.
Manufacturers claim that using the through-the-lens
light meter inherently compensates for the decrease in
the amount of light that reaches the sensor when a filter
is added. These claims are valid to the extent that the
spectral response of the meter corresponds to the spectral
sensitivity of the sensor. Conducting a simple experiment
to determine the accuracy of the in-camera meter is Figure 5-6 The Maxwell triangle is useful for predicting the effect of
accomplished by metering a neutral test target with and color filters on subjects in black and white photographs.

129
 s ecti o n 1: Captu r e

and white. We interpret the effect of color filters by under- adjacent colors, blue and green. Figure 5-7 shows a scene
standing the following rules: rendered in full color, converted to black and white with an
automatic software filter and using a color filter at capture.
1. A filter lightens objects that are the same color as the The differences in tones and tone contrast can be dramatic
filter. It also lightens colors adjacent to it on the triangle when color filters are properly considered. Understand-
to a lesser degree. ing the color relationships of the Maxwell triangle proves
2. A filter darkens objects that are on the opposite side valuable beyond lens filters: we show how they help us to
of the triangle to the filter’s color. It also darkens colors balance color and tone in post-processing in Chapter 10.
adjacent to that opposite color to a lesser degree.

Thus a red filter lightens red, magenta and yellow sub- Neutral Density Filters
ject colors while darkening the opposite color, cyan, and its
Neutral density (ND) filters reduce the transmitted illumi-
nance by a known factor. ND filters offer an alternative
method of controlling exposure when the conventional
aperture and shutter controls are inadequate. With highly
sensitive photographic materials, a combination of the
smallest diaphragm opening and the highest shutter speed
may still result in overexposure at high illumination levels.
There are also situations in which the diaphragm opening
is selected to produce a certain depth of field and the shut-
ter speed is selected for a certain action stopping capability
rather than for exposure considerations. The intent is to
have the same transmittance for all wavelengths of radia-
tion within specified limits. ND filters are usually calibrated
in terms of white light density. A filter with a neutral density
of 0.3 has an opacity and filter factor of 2 (the antilog of 0.3)
and a transmittance of 0.5 (the reciprocal of the opacity)

Figure 5-7 Image B is a black and white version of image A rendered Figure 5-8 The spectral transmittance of 0.3 neutral density filter.
with an automatic conversion filter in image editing software. Image C
was photographed using a yellow filter and a monochrome sensor.

130
 c h a p t e r 5 : L e n s F i lt e r s

as plotted in Figure 5-8. Some ND filters are calibrated polarization and the plane of polarization of the light are
in terms of stops, where each 0.3 in density corresponds the same, the maximum amount of light is transmitted.
to one stop of exposure change. Variable neutral density If they are at right angles to each other, no light passes
filters use a rotatable sandwich of filter elements to allow as shown in Figure 5-9. Circular polarizers never align in
for fine-tuning the degree of light reduction. such a way that completely extinguishes light transmis-
Graduated neutral density filters are a special subcategory sion. Instead, varying amounts of light are transmitted at
of particular interest to outdoor and landscape photogra- different rotation angles between the two sandwiched
phers. These filters start out completely transparent on one filter elements. In this way, polarizing filters can be thought
edge and gradually increase in density until they reach a of as variable neutral density filters. Figure 5-10 demon-
maximum density (usually 0.3, 0.6 or 0.9) at the opposite strates that rotating a circular polarizing filter alters the total
edge. This is useful when the brightness range between amount of energy transmitted. Note how the filter does not
the sky and the foreground is significant. Often times the discriminate by wavelength.
dynamic range of an outdoor scene is so large that a single If polarizing filters simply reduced the amount of light
exposure leaves the sky overexposed or the ground under- transmitted, we wouldn’t be highlighting them here.
exposed. We can lessen the difference by introducing a Figure 5-10 doesn’t tell the whole story of their photo-
graduated neutral density filter and aligning its orientation graphic impact. The unique characteristic of polarizing
with the scene horizon. This way, a single shutter speed,
aperture and ISO combination captures detail in both the
foreground and sky—the area of greatest density on the
filter blocks some of that light from the bright sky. The
gradual transition makes its presence potentially unde-
tectable to the observer. Since its alignment in front of the
camera is dictated by the photographer’s composition,
square graduated ND filters with filter holders that mount
to the lens are an available solution. This way, you have the
ability to raise or lower the higher density portion of the Figure 5-9 When filters are parallel as in (a) polarized light is passed,
when they are at right angles (b), no light is passed.
filter in relation to your composition.

Polarizing Filters

The phenomenon of polarized light provides photogra-

phers with an opportunity to further control light (see
Chapter 1 for a refresher). Placing a polarizing filter over the
lens limits the light recorded by the camera to that which
oscillates exclusively in one orientation. This is possible
because the filters are manufactured with microscopic
crystals aligned a uniform, parallel orientation. There are
both circular and linear polarizing filters, the former more
commonly used than the latter as photographic lens
attachments. Linear polarizers can cause exposure or focus
issues due to their interaction with the camera mirror Figure 5-10 Spectral transmittance of a polarizing filter measured at
components. With linear polarizers, if the filter’s plane of 0° and 90° rotations.

131
 s ecti o n 1: Captu r e

Figure 5-11 shows a landscape with a blue sky photo-

graphed using a polarizing filter. The effect is dramatic: the
sky darkens and often becomes a more saturated shade of
blue. You’ve likely experienced a similar result from wearing
polarized sunglasses in the summertime. The surfaces of
tree leaves reflect light in many directions, resulting in
scattering and causing a hazy appearance from a distance.
A polarizing filter over the lens reduces this non-image
forming light (flare). We know how colorful and vibrant the
changing autumn leaves are, yet a standard photograph
fails to capture it in part because of this natural haze. A
polarizing filter cuts out a lot of the scattered, randomly
oriented light waves, leaving an image with more inher-
ent saturation and contrast (Figure 5-12). It comes as no
Figure 5-11 An outdoor scene photographed without (left) and with surprise, then, that many outdoor and landscape pho-
(right) a polarizing filter mounted to the camera lens. Photographs by
tographers carry polarizing filters. Clever software tricks
Stephen J. Diehl
may emulate some of the benefits of polarizing filters
(like darkening blue skies) but there is no replacement for
physically filtering the light before it reaches the sensor—it
fundamentally changes the transmitted and therefore
recorded light.

Ultraviolet and Infrared Filters

Color filters are defined by their ability to selectively limit

the visible light wavelengths transmitted. Ultraviolet and
infrared filters behave the same except that they limit the
transmission of electromagnetic radiation outside of the
visible spectrum.
Ultraviolet (UV) absorption filters play an important role
in color film photography. The blue-sensitive emulsion
Figure 5-12 An outdoor scene photographed without (left) and with
(right) a polarizing filter mounted to the camera lens. The colors in the
layer is also sensitive to ultraviolet radiation (recall that on
scene are more saturated thanks to the filter’s ability to remove glare. the electromagnetic spectrum, UV wavelengths precede
Photographs by Stephen J. Diehl violet visible light). Placing a UV filter over the lens prevents
captured images from having a blue color cast. The need
filters is in how they manipulate an aspect of light that we for UV filters with digital sensors is an area of debate, as
don’t perceive. Light vibrates in all directions but we don’t CMOS chips are fairly insensitive to the UV portion of the
notice any vibration orientation differences. Polarizing spectrum.
filters do, and limiting which light rays get through to the Ultraviolet radiation can have a deleterious effect on
sensor offers the ability to cut down on reflections and photographs of some outdoor scenes because short
haze because their orientations are much more likely to be wavelength radiation is scattered much more by the
random and nonuniform. atmosphere than longer-wavelength radiation, creating

132
 c h a p t e r 5 : L e n s F i lt e r s

the appearance of haze. Haze lowers contrast and obscures Spectrophotometric Absorption
detail in distant subjects, as discussed in the previous sec- Curves
tion on polarizing filters. This Rayleigh scattering is inversely
proportional to the fourth power of the wavelength of the The types of filters discussed thus far are usually sufficient
radiation. A UV filter reduces the appearance of haze due to meet the needs of photographers. When more specific
to this atmospheric light scattering. From this, we might information is required for technical and scientific appli-
conclude that attaching a UV filter is a prerequisite for out- cations, spectrophotometric curves showing the absorption
door shooting. However, many manufacturers incorporate characteristics of filters are useful. The spectral transmit-
a UV filtering material on the sensor itself, rendering the tance of a filter is measured using a spectrophotometer.
practice redundant. Still, many feel that using a UV filter is A spectrophotometer disperses white light to form a
worthwhile if only as additional protection. Camera lenses spectrum and measures the spectral response, using
are expensive; UV filters are not. Photographers like to use either transmittance or reflectance, over small wavelength
UV filters to protect against dust and scratches on the front bands. A spectrophotometer typically measures from short
lens surfaces. After all, we’ve all used our shirt or sleeve to wavelength ultraviolet radiation through the visible region
wipe off a lens and it is cheaper to replace a scratched filter to long wavelength infrared radiation. Spectrophotometric
than a scratched lens. The only caveat to this practice: if the curves represent wavelengths on the horizontal axis and
UV material is cheap, it can lessen the quality of the image transmittance or density values on the vertical axis, with
that reaches the sensor. Any time that additional material the baseline representing 0% transmittance (or a density
is added to the optical path, you risk degrading image of 0).
quality. Curves for blue, green and red color filters are shown
Ultraviolet transmission filters are a different story. Instead in Figure 5-13. The maximum transmittance (or minimum
of blocking UV energy from reaching the sensor, they density) is in the wavelength region that identifies the hue
make sure that nothing except for UV gets through. These of the filter. Notice how the curves do not divide the visible
filters are used by photographers looking to take images spectrum neatly into thirds. It’s impossible to obtain col-
exclusively using UV, blocking any visible light present in orants that have perfectly sharp-cutting characteristics at
the scene. Exposures are often much longer than in normal the desired wavelengths. Additionally, most photographic
photographs or require significant boosts in ISO because
CMOS sensors are not very sensitive in this region.
Infrared (IR) energy, in contrast, is inherently detected
by camera sensors. Manufacturers place an IR cut-off filter
in the sensor to block IR radiation from getting recorded.
Such a filter makes it challenging or impractical to shoot
IR-exclusive images. IR cut filters are not designed to be
removed. However, a quick internet search can direct any
enterprising photographer to many sites demonstrating
techniques for removing them. Every so often a camera
comes to market offering IR photography capabilities by
specifically excluding this filter from its design. If you find
yourself in possession of such a camera, you’ll need an IR
transmission filter that, similar to the UV version described
earlier, only allows IR energy through.

Figure 5-13 Spectrophotometric curves for blue, green and red filters.

133
 s ecti o n 1: Captu r e

filters transmit infrared radiation freely; infrared radiation 1. Set the camera to display a three-color histogram when
transmittance is of little consequence because sensor reviewing images, as it helps to indicate proper expo-
designs include infrared blocking filters. sure of each color separation image.
2. Take a traditional, non-filtered full color exposure first.
3. Determine proper exposure for the three sequential
Proof of Concept: Constructing Color frames. Adjust shutter speed or aperture to obtain
Images Using Color Separations equivalent exposures across all three. Varying shutter
speed is recommended because varying aperture results
We can use color filters to make color photographs from in depth of field mismatches.
three separate images, each exclusively exposing for 4. With the red filter covering the lens, take a properly
one additive primary (red, green and blue). This is called exposed image.
shooting color separations. James Clerk Maxwell first 5. With the blue filter covering the lens, take a properly
demonstrated the technique in 1861. It’s assumed that exposed image. Consider waiting 30 seconds or a few
a color photograph taken today is captured in a single minutes to allow for movements or changes in the
instant, but it doesn’t have to be. Shooting a scene that scene.
includes moving elements becomes aesthetically interest- 6. With the green filter covering the lens, take a properly
ing when the frames are combined; the “errors” involved exposed image. Three example images following these
in having a subject move during a sequence of exposures steps are shown in Figure 5-14.
reveals the nature of the process. An outdoor scene with
clouds in the sky, trees blowing in the wind or people Open the three R, G, B images and the traditional full
walking by are all good candidates for color separation color image in Adobe Photoshop. Paste the R, G, B images
shooting. into the full color image document as layers. This is accom-
Use a tripod in order to make the series of photographs. plished by dragging-and-dropping them or by copying
Take three gel filters: red, green and blue to hold in front and pasting. You should have a document with the full
of the camera lens in turn. Cheap cellophane material color image as the bottom layer and the three R, G or B
designed for placement over studio lights is one way to images as layers above it.
make such filters. Since they’re not specifically calibrated or Add a solid black color layer between the RGB separa-
designed to function as lens filters, we calculated their filter tions and the full color version. This is done with a color fill
factors by taking density measurements and converting to layer: Layer > New Fill Layer > Solid Color. This results in the
stops. window shown in Figure 5-15. Click OK. Set the R, G and B
values to 0 in the color picker to make the Fill Layer black.

Figure 5-14 Three color separation images photographed sequentially through gel filters. Photographs by Rochester Institute of Technology
photography alumni Carissa Hurdstrom and Nicole LeClair

134
 c h a p t e r 5 : L e n s F i lt e r s

Now you are ready to put the additive color process to

work! With the red image layer selected, change the Blend
Mode to “Linear Dodge (Add)” as shown in Figure 5-16.
Make this same Blend Mode change for the blue and
green layers. If done correctly, the result is a full color image
created by three independent exposures, each of which
only recorded one of the three additive primary colors at Figure 5-15 The New Fill Layer dialog box.
capture. Use the Move tool to line up the three layers if
there is poor registration between them.
Any differences in the scene between the three expo-
sures appear as colorful ghosting because moving subject
matter was recorded with R, G or B information exclusively
at a given location (see Figure 5-17, overleaf). A significant
colorcast in the composite is indicative of improper expo-
sure in one or more of the three filtered images and we
recommend trying again with more due diligence paid to
exposure compensation.
Toggle visibility of the R, G, B layers to compare the
composite version to the traditional full color image made
in a single capture.

Figure 5-16 The layers palette with all image and color
fill layers in place. Set the Blend Mode for red, green and
blue separation image layers to Linear Dodge.

135
 s ecti o n 1: Captu r e

Figure 5-17 A full color image made in a single capture (top) and full color image created using three color separation
captures. Photographs by Rochester Institute of Technology photography alumni Carissa Hurdstrom and Nicole LeClair

136
 6 Motion Video
Fundamentals

Photograph by Rochester Institute of Technology photography student Justin Scalera

Still photography and motion video share some fundamental mechanics and are often used
hand in hand as visual media assets in commercial and personal spheres. The last decade
saw a hardware convergence where professional still cameras gained video recording capa-
bilities and professional video cameras gained high resolution still photography capabilities.
Each medium continues to retain unique applications and artistic possibilities while the
dividing line between photographer and videographer becomes less rigid. The concepts

137
 s ecti o n 1: Captu r e

discussed in this chapter are for still photographers that Ixt=k (Eq. 6.1)
find themselves working with video; to fully detail the
materials and technologies of motion picture science Additionally, the apparent motion effect is observed
requires a book of its own. We’ll use our familiarity with whereby a static stimulus is shown in a sequence of
photography to inform our discussion of motion video fun- changing locations at 100 milliseconds or shorter intervals.1
damentals: motion perception, video capture parameters, This presentation is perceived as realistic motion rather
formats, codecs, data rates and video quality artifacts. than discrete changes in stimuli position. Motion pictures
began with film projected in a dark room using a rotat-
ing shutter in the projector. Each frame was positioned
Motion Perception and Memory between the light source and the projection lens while the
shutter blocked all light from projection in the time it took
Some of the light now reaching Earth from distant stars to advance the frames. The projection went dark between
originated millions and even billions of years ago. Celestial frames. The persistent image behavior and apparent
events observed through powerful telescopes occurred motion response of our visual system combined with its
long ago. The delay between those events and our percep- flicker fusion prevented viewers from seeing these dark
tion of them is extreme. At smaller distances here on Earth, intervals that would otherwise appear as flickering bright-
the delay between light falling on the retina and its per- ness changes and stilted motion. Early motion pictures
ception is extremely short and negligible as it relates to our were sometimes referred to as flicks because they were
visual sense. Perceptions that occur after light stops falling shown at 16 frames per second and flicker was apparent.
on the retina, however, are of considerable importance. Computer and television displays create the illusion of
Such post-stimulus perceptions include persistent images, a constantly changing, smooth motion video by taking
afterimages and visual memory. These human visual advantage of the persistent image effect so that the viewer
system responses play an important role in our experience sees a complete picture. Duration of the persistent image
of motion video media. can vary depending upon several factors including the
image luminance level and whether removal of the stimu-
lus is followed by darkness or by another image.
Persistent Images It’s reasonable to speculate that increasing image
Laboratory studies indicate that persistent images last an luminance produces a stronger effect on the visual system
average of one-quarter of a second after the stimulus is which then increases the duration of the persistent image.
removed. Consider an experiment where the letter “A” is The effect is actually the reverse. The increase in persistent
projected onto a screen in a darkened room and the pres- image duration with decreasing luminance is explained by
entation time is controlled with a shutter. The presentation comparing the visual effect with capturing an image at low
time cannot be made so short that the image cannot be light levels. If the sensor does not receive sufficient expo-
seen, provided that luminance is increased proportionally sure at a selected shutter speed and aperture setting, the
such that the same total amount of light is available to the exposure time must be increased in order to obtain a sat-
viewer. Equation 6.1 defines Bloch’s law which predicts that isfactory image. The visual system effectively compensates
the visual effect is the same for different combinations of for the lower image luminance by sacrificing temporal res-
luminance (I) and time (t) as long as the product (k) remains olution and increasing the duration of the persistent image,
the same. Bloch’s law is not valid for very bright and very which is the equivalent of increasing the exposure time.
low light values, however. Perception occurs with short This effect is illustrated with motion picture film pro-
exposure times because the persistent image keeps the jected onto a screen. If flickering of the intermittent image
image available to the viewer for somewhat longer than is just noticeable at a certain luminance level, reducing the
the actual presentation time. luminance increases the duration of the persistent image

138
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

to include the time when the screen is dark and the flicker with a stimulus having lower luminance, such as a white X
disappears. The decrease in luminance can be accom- on a black background, by staring at it for a long time and
plished by using a smaller projector bulb, increasing the then looking at a uniform surface (see Figure 6-1).
projector-to-screen distance or by viewing the projection Afterimages can be negative or positive. Negative after-
through a neutral density filter. images are attributed to local bleaching of the pigments
in the retina’s photoreceptors. Positive afterimages are due
to the nerve cells continuing to fire signals to the brain.
Afterimages Negative afterimages of colored stimuli tend to be approx-
The human visual response produces afterimages that are imately complementary in color: a bright yellow light tends
recognized by the viewer as visual anomalies. Afterimages to produce a bluish afterimage, for example.
are most evident when the eye is exposed to an intense
flash of light. Looking directly at a camera flash usually
causes the flashed individual to see a vivid afterimage Visual Memory
bright spot for some time. An afterimage can be formed Visual memory shares many characteristics with other
types of memory but it includes an ability to retain or
recall a visual image that is distinct from persistent images
and afterimages. Visual memories can be short term or
long term. They are exploited heavily by filmmakers to
consciously and subconsciously build connections and
structure with imagery. Just as a phone number can be
remembered long enough to dial it by repeating it and
keeping it in the conscious mind, so too can a visual
image be remembered for a short time by keeping it in
visual memory. Experiments indicate that most people
cannot retain more than five to nine independent pieces
of information in short-term memory (phone numbers are
pushing it, even when breaking them up with cadence).
Furthermore, a viewer should not expect to remem-
ber, even for a short time, all of the details in a complex
picture. One tends to remember the details that attract
the most attention or hold the most interest. Short-term
visual memory is used even when scanning a picture and
fixating on different parts of the composition. This enables
us to create a composite perception of the total picture.
Once attention is allowed to go on to other things, long-
term memory is required to revive an earlier visual image.
Except for simple stimuli such as a circle or a square, visual
memory images are generally not facsimile representations
of the original scene; they usually contain just the most
important details.
Figure 6-1 An afterimage illusion. Stare at the The human visual system is designed to detect
isolated white X for about 1 minute. Shift your
gaze to the middle square in the grid below it
motion over the entire area of the retina so that moving
and win the game of tic-tac-toe. objects in the periphery of our field of view are detected.

139
 s ecti o n 1: Captu r e

A semi-reflexive eye movement enables the viewer to posi- Motion video takes the photograph and multiplies it: it’s
tion the moving object image on the fovea of the retina a sequence of image frames shown in rapid succession in
where visual acuity is highest. Once the object is fixated, a way that our human visual system perceives as similar
the visual system is quite good in tracking its movement to our experience of the living, moving world (see Figure
unless the speed is very high or the direction of movement 6-2). This perception of motion from sequences of discrete,
is erratic. It is almost impossible to pan the eyes smoothly still pictures is known as the phi phenomenon. How many
except when tracking a moving object. Instead, the eyes frames are used every second to perceive smooth, natu-
tend to move in quick jumps called saccades that are ral and real-time motion? Full motion video is generally
apparent in high speed video recordings. Skill in tracking considered to be at least 15 frames per second (fps). Video
rapidly moving objects is important for photographers with this frame rate, the number of frames displayed in one
looking to catch the optimal moment, be it an athlete’s second of playback, appears stilted even though it begins
action or the perfect timing of everyday activity. Our visual to convey a series of instances in sequence. This stilted look
memory weighs these tracked objects heavily and may is a video quality artifact called motion jitter. The number of
miss static or surrounding detail as events unfold. frames recorded per second is a form of temporal sampling
similar to how sensor resolution is a form of spatial sam-
pling. The greater the sampling rate, the more information
Defining Motion Video and its recorded over the capture duration and the smoother
Parameters motion appears.
Early silent film endeavors featured 16–18 fps record-
In a world of digital publications, Twitter, messaging apps ings. This satisfied the fluid motion component but was not
and streaming television, how do we draw the boundaries enough temporal sampling for natural-sounding audio. A
between visual media formats? A photograph tradition- higher frame rate of 24 fps supplanted 18 to accommodate
ally meant a single instance captured in an image frame, an audio track which settled into a long-standing standard
viewed on a physical output medium like print. A video for motion pictures. Our many decades of watching and
traditionally meant motion pictures capturing longer enjoying films captured at 24 frames per second at a spe-
durations and viewed on a display or projection screen. cific shutter angle (described later in this chapter) means
Photographs were viewed on a wall or on the page while that we have a particular familiarity and penchant for its
video was watched in a movie theater or on television. motion characteristics.
Today’s use of photography and videography is decidedly Standard capture and playback frame rates for motion
more blended and their output is frequently through video are 24, 29.97, 48 and 59.94 frames per second.
mobile and personal computing devices. Still, a baseline Cinema primarily uses 24 fps; consumer recording devices
set of parameters delineate motion video (sometimes col- tend to record at 29.97 fps. Professional still cameras
loquially called “film” or simply “video”) from photographs. including SLRs offer multiple standard recording frame rate
Let’s expand on the characteristics that make motion options. Anything higher than 30 fps is considered high
video a unique medium while highlighting what the frame rate (HFR). Lower frame rates mean fewer total frames
photographer must consider when working with video for a given video recording duration and therefore might
parameters. be considered if storage space is an issue. High frame rate
capture and playback is particularly beneficial with fast
moving subjects and camera movement like panning
Frame Rate shots. Traditional panning shots must be below a certain
A photograph is a single frame. It captures an ephemeral speed for the viewer to keep track of the scene content at
instance of time even with long exposures since they 24 frames per second. High frame rates reduce motion blur
ultimately compress temporal changes into a static visual. and panning smoothness.2

140
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

Figure 6-2 A sequence of video frames captured at 30 frames per second.

Capture frame rate and playback frame rate are not slows down real-time motion to 1/400th of the original
always matched. Time-lapse imagery can show a flower speed. Expanding these incredibly fast events to play
bud transitioning into full bloom in a matter of seconds out over a time scale that our visual system is attuned to
despite recording the event over many hours. On the reveals a temporal experience of our world that is other-
other end of the scale, recording at frame rates above 48 wise inaccessible. Figure 6-3 (overleaf) shows sequential
fps is common practice when the intent is to play those frames from a high speed video recording of a fast event.
frames back at a standard 24 fps. This is the basis for slow The motion details and microsecond events that unfold
motion video. By sampling many more frames in real time become easily observable in a way that is impossible in
at capture and playing them back at a lesser rate, time is real time.
expanded to show brief moments of action over longer Note that reducing frame rate is a viable form of data
durations. Capturing at hundreds or thousands of frames compression as it lessens the amount of stored data
per second is what makes filming extreme slow motion relative to a higher frame rate original. However, the visual
events like explosions or guns firing possible. Recording impact of lowering a video’s frame rate is often not worth
at 10,000 fps and playing it back at 24 fps, for example, the storage gains. This strategy is also hampered by the

141
 s ecti o n 1: Captu r e

Figure 6-3 A sequence of frames from a high frame rate capture. Playing the frames back at 24 fps creates a slow motion effect that dilates time.
Photographs by Rochester Institute of Technology Assistant Professor Ted Kinsman

math of turning 30 fps into 24 fps, as one example, since it times is vast compared to that of a videographer. Pho-
involves blending, interpolating or dropping select frames tographic exposure can integrate many moments of
in a specific pattern. It’s more common to decrease frame time into one combined image, like in the case of a long
size or data rate to lessen the amount of video data. exposure of highway traffic at night. The changing posi-
tions of car headlights and taillights are never still and the
resulting photograph shows long light streaks along the
Shutter Angle road. Motion video exposure cannot extend or integrate
Photography and videography both start with the act multiple seconds into one frame in quite this way if the
of recording light. However, a single instance frozen in a goal is recording light over a period of time. Instead, frame
photograph makes up just a fraction of a second of screen exposure is limited in duration based on the frame rate.
time in a video. This is an area of friction if your photogra- Analog motion picture cameras use continuous rolls
phy exposure knowledge is ingrained and you’re new to of photographic film. A gate or rotary shutter moves
video; the photographer’s available range of exposure in front of the film for a fraction of a second to create

142
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

distinct, separate exposures. Where a still camera’s focal second at 30 fps, for example, might be great for pulling
plane shutter is a pair of synchronized curtains, a motion a still frame but in motion looks choppy and becomes
picture camera shutter is a circular disc with an angled difficult to watch.
opening whose size dictates exposure duration. Digital Ultimately, video considers shutter speed or shutter
image sensors reproduce the effect of rotary shutters by angle in the context of the desired frame rate as it limits
electronically controlling exposure time and frame read- the upper end of how long a single frame can be exposed.
out. As we learned in Chapter 4, longer shutter speeds Faster frame rates are more light-hungry in the sense that
introduce motion blur as scene motion can occur over a the available exposure durations are more limited. Both
single frame’s exposure duration. Thus, one of the biggest frame rate and shutter speed are time-based characteris-
departures from still photography exposure variables tics, but frame rate on its own does not define the duration
is the shutter angle relative to the frame rate and it is of one frame’s exposure.
described in degrees. Though digital cameras lack a physi- All of this is important because exposure in video is
cal rotary shutter like their analog ancestors, the concept much more dependent on light manipulation and control
here remains the same and the term persists. A 360° to achieve good motion characteristics, desired depth of
shutter angle is the largest possible setting and means field (via the aperture), standard frame rates and overall
that a single frame is exposed for the entire duration brightness. Better sensor sensitivity helps, though video
possible while recording at a given frame rate. Calculating production is strongly aided by bright, continuous light
video shutter speed is accomplished using the formula in sources on set and neutral density filters when natural light
Equation 6.2: is not easily controlled.

1
Shutter Speed (sec) = (Eq. 6.2)
Frame Rate (fps) x 360 (degrees)
Shutter Angle (degrees) Resolution
Digital video resolution, the number of pixels that make up
A 360° shutter angle when recording at 30 fps equates the two-dimensional frame, is strictly standardized. There’s
to a single frame exposure time of 1/30th second. It is no technical reason why a video can’t be 3000 by 250
impossible to record 30 unique frames in 1 second if a pixels but it doesn’t make for an enjoyable viewing experi-
single frame is exposed for longer than 1/30th. A shutter ence when our displays cannot adapt to unusual shapes or
angle of 180° renders an exposure time of 1/60th second dimensions. Viewers appreciate the consistency of standard
at the same frame rate, as it represents a rotary shut- video resolutions and that content makes full use of their
ter that exposes for exactly half of its rotation. Smaller expensive television’s screen real estate.
shutter angles translate to shorter exposure times and A single frame of video, until recently, was lower reso-
less motion blur. A shutter angle of 180° is traditional for lution than any digital still image; the difference is that the
motion picture capture and creates the degree of blur brain has plenty to process when presented with dozens
we’re accustomed to seeing in movies. It also dictates how of images per second. Photographers sometimes describe
quickly filmmakers can pan the camera and expect viewers an image file in terms of megapixels as a shorthand for the
to clearly follow and focus the action. A smaller shutter number of pixels in height and width. However, aside from
angle is imperative if there is an expectation of pulling still bragging, this is not a particularly useful descriptor for still
photographs from high-resolution video: it increases the images. Additionally, photographs can be every possible
likelihood that individual frames are sharp with minimal combination of pixel widths and heights. Video has distinct
motion blur. Small shutter angles mean fast shutter speeds resolutions that tell us a key parameter of the viewing
which translates to crisp, frozen motion. Going too fast in experience. Table 6-1 (overleaf) lists the industry standard
shutter speed is problematic because it leaves temporal video resolutions in pixel dimensions, aspect ratio (the ratio
gaps at standard frame rates. A shutter speed of 1/500th of frame width to height) and the general name for each

143
 s ecti o n 1: Captu r e

Table 6-1 Standard video resolutions and their aspect ratios.

Video Resolution Pixel Dimensions (Width x Height) Aspect Ratio (Width:Height)

Standard Definition (SD), “480p” 640x480 4:3 or 1.33:1

High Definition (HD), “720p” 1280x720 16:9 or 1.78:1

High Definition (Full HD), “1080p” 1920x1080 16:9 or 1.78:1

Ultra High Definition (UHD), “4K” 3840x2160 16:9 or 1.78:1

Digital Cinema 4K 4096x2160 1.85:1; 2.35:1

Ultra High Definition (UHD), “8K” 7680x4320 16:9 or 1.78:1

video frame—but it offers the flexibility to crop in from a

wider view or digitally pan the composition. This strategy
is known as overscan. The additional pixels also help for
electronic image stabilization processing that employs a
form of digital panning to keep subjects at a static posi-
tion in the video frame. Just like rolling out the dough for
covering a pie, it helps to have a little more than what’s
needed rather than leaving zero margin for error. This is
particularly helpful for photographers whose cameras allow
Figure 6-4 Relative frame sizes of standard video resolutions. for ultra-high-definition video recording resolutions.

tier. Figure 6-4 illustrates the relative frame sizes of the Aspect Ratio
different video resolutions. Note that the higher resolution Similar to resolution, video requires working with a limited
formats are large enough that single frames are serviceable set of aspect ratios to ensure supported and standardized
as reproducible photographs. output across displays. The aspect ratio is the relationship
It’s easy to see the progression: standard definition (SD) between the width and height dimensions of a frame or
came first and was superseded by high definition (HD). video without a dependency on physical dimensions or
There are two versions of high definition with the larger pixel counts. A square frame has an aspect ratio of 1:1 as
(and more common) designated as Full HD. From there, its height and width dimensions are equal. The larger the
video started being captured and displayed at ultra high difference between the two numbers, the more rectangu-
definitions or UHD (4K and 8K are both members of this lar the shape.
club). We’re skimming past the distinction between inter- The cinema dictated the early decades of film formats
laced and progressive scan simply because you’re unlikely and commonly used aspect ratios went through different
to encounter interlaced video in today’s media landscape. eras. The Academy aspect ratio was 1.375:1 until the 1950s
However, note that the “p” in “1080p” indicates a video as while some short-lived formats went as wide as 2.35:1
progressive scan. (CinemaScope).3 Today, the cinema standard is 1.85:1.
It’s common for sensors to record at resolutions greater High-definition television and computer displays frequently
than the intended output, i.e. recording at 2K to produce use 1.77:1, also known as 16:9. This was likely picked
a 1280x720 video. The additional captured pixels likely because it sits comfortably between the 4:3 standard for
won’t be used—in the end, 1280x720 pixels are in the final television that came before it and the 1.85:1 standard for

144
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

movies and film. An aspect ratio that is wider than it is tall is Video Formats and Encoding
inherent to motion video due to our vision’s larger horizon- Strategies
tal field of view. This is different from photographs, which
have always allowed portrait and landscape frame orienta- Digital videos are stored in a variety of file formats. In con-
tions in output reproduction. trast to image file formats, video file formats are containers
The only time we accept aspect ratios that differ from and do not dictate everything about their contents in the
our televisions or phone screens is when padding called way that a JPEG or TIFF does. A JPEG file describes the
letterboxing (or its sideways equivalent, pillarboxing) is used; format—JPEG has a particular structure to storing meta-
black bars are added above and below or on either side of data and image pixel information—and it also dictates the
the active video area. Even then, we understand that this use of the JPEG compression algorithm. Video coding sep-
strategy is not making full use of the total pixels available arates the two elements; both the container format (also
on the displays. called the wrapper) and the compression method must be
Consumers buy televisions or hold their phones side- defined and knowing one does not necessarily tell us what
ways and expect that the majority of our display is put to the other will be. Video files consist of the video stream, the
good use in playing video content. When there’s a mis- audio if applicable, metadata and the encoding informa-
match between the aspect ratio of a video and that of the tion encapsulated in a container. Common video container
display, playback software has three options: formats include MOV, AVI, MP4, MKV and MPEG with file
extensions that match.
1. Crop the video in the height or width dimension to fill If the video container is like the carry-on luggage, the
the screen codec is how you pack your clothes to travel. A codec is
2. Add letterboxing (black bars on the top and bottom) the hardware or software that handles writing and reading
3. Add pillarboxing (black bars on the left and right) video data for storage and playback. The term comes from
a concatenation of compression-decompression and its role
An aspect ratio of 4:3 was the norm in the days of stand- is handling the encoding and decoding of video data. A
ard definition and is a ratio that photographers find familiar. video must first be encoded for the data to be organized
High-definition formats, including the broadcast television and stored in a container file. The container likely also
standard, use a 16:9 aspect ratio. This is more of an elon- contains audio data which requires its own type of encod-
gated rectangle compared to 4:3 as shown in Figure 6-5. ing. The act of encoding involves applying compression
algorithms so that the video is efficiently packaged. This
encoded data can be stored to memory or transferred
over a network and when it gets to its destination, it is
decoded—unpacked, translated—before it can be played
back to the viewer. Codecs are the heart of a video file
when we want to know how to play it and what to expect
in terms of compression and compatibility. At risk of over-
using the analogy, video containers are easy to convert
between in much the same way that clothes can be moved
from one carry-on to another. This does not change the
contents or the weird way you fold your socks, only the
Figure 6-5 The width and height of SD and HD frames, independent storage mechanism.
of resolution, shows the difference in aspect ratio. High definition often
Photographers take for granted that JPEG images are
uses a wider 16:9 while standard definition uses the squarer 4:3.
universally and instantaneously decoded by image view-
ers and photo editing software. Video must be similarly

145
 s ecti o n 1: Captu r e

decoded, though it has to happen on a per-frame basis. If of these facts help codecs approach compression strate-
it can’t be done quickly, we don’t get to see seamless full gies to great effect.
motion video. A piece of software or hardware works in A single video frame is subdivided into square groups
real-time to decode and unpack the video (and audio if of pixels called macroblocks that are treated as single units
present) fast enough to put a frame on screen at just the for compression and motion compensation. A common
right moment. size for macroblocks is 16x16 pixels. Intraframe compression
looks at sections of a frame, most often grouped into mac-
roblocks, and independently compresses those sections.
Intraframe and Interframe Encoding The encoder’s compression considers only the pixels in
Frames are the building block of a motion video. Recall that frame similar to JPEG compression on a single photo-
that a standard 30 fps video has 30 unique images shown graph. A video player decodes the data one macroblock at
over a 1-second duration. The process of encoding video a time.
starts with a basic question: how do these frames relate Interframe compression considers the frame information
to one another in sequence? It’s very likely that, over the that comes before and after the frame being compressed.
course of one single second, some of the content in those This is where the video encoding process takes advantage
30 frames is shared (Figure 6-6 shows one such example). of the redundancy across a proximate set of frames called
Put another way, the content remains mostly the same temporal redundancy. If the information exists in frame
across fractions of seconds of elapsed time. Even within a 1 and stays the same for frames 2, 3 and 4, frames 2–4
frame, it’s likely that local groups of pixels are similar. Both don’t need to store that same information. The only pixel

Figure 6-6 Some videos feature relatively static pixel content from one frame to the next with a subset that changes over their duration.
Above, the third image shows a frame-differenced view of the areas that see changing content.

146
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

information that must be stored in those subsequent are the most efficient frame time as a result since they
frames is that which changes in brightness or color. Thus, take advantage of redundant content in either direction.
video encoded using interframe compression specifies A B-frame only contains pixel content that doesn’t exist
coded picture frame types to minimize redundancies and in the previous or following frame which can translate
data storage while maintaining image content over its to a small amount of new data. Encoding and decoding
duration. a video with B-frames is more computationally intensive
There are three frame types in interframe encoding. and is sometimes avoided to ensure compatibility with
The first is the intra-coded or I-frame (also known as the lower-end output devices.
key frame). I-frames are completely new, wholly inde- Both B-frames and P-frames use motion estimation
pendent pixel content. The macroblocks of the I-frame vectors to identify pixel groups that appear to stay the
can be compressed but there is no dependency on the same while moving in the frame between two I-frames.
frame that precedes it or the frame that follows. It helps If the camera is static, motion vectors are effective at
to associate the “I” in I-frame with “independent.” Every so storing only the position changes of pixels. A moving
often (the exact interval is controlled by the encoder), an camera and a moving subject make this technique less
encoded video needs an I-frame as a sort of data land- efficient, as much of the frame is rapidly changing from
mark. These can be one second apart or 20 seconds apart, one moment to the next. Motion vectors can cause visual
for example. The easiest video encoding for a decoder warping or other quality artifacts if there’s any data loss
to process is a video comprised entirely of I-frames as or errors when encoding I-frames. Figure 6-7 (overleaf)
they are effectively sequential JPEG photographs. The shows a video frame with a motion vector overlay. Notice
downside to encoding purely I-frames is that the file size how the static surround does not show any but the falling
becomes gigantic. The increased computational com- log flume is identified as a subset of image pixels with a
plexity of introducing frame types beyond I-frames is downward movement that will continue into succeeding
an acceptable compromise in an effort to keep file sizes frames.
manageable. A few seconds of video may be encoded to consist of
Between I-frames are predicted or P-frames when a all three of these frame types (I, P and B) in an arrange-
video is encoded with varying coded picture frame types. ment called a group of pictures (GOP) illustrated in Figure
P-frames rely on information from a preceding I-frame 6-8 (overleaf). Video decoding software often requires
to decode, as they only contain information related to that playback start at the beginning of a GOP because it
changes that happen in the time elapsed from that last needs to establish the first I-frame that is decoded without
I-frame. They borrow everything else from the refer- dependency on other frame data. You may find that video
enced frame. They are also referred to as delta frames to players take a moment or two to start playing back video
reflect this behavior. P-frames are smaller in size and are content, especially when seeking through a timeline: this is
compressed more aggressively to save on space. Since a related to the software locating the start of a GOP. You may
P-frame is only seen for a fraction of a second, you’re not also notice frame quality “breathe” in and out in heavily
as likely to notice as long as high-quality I-frames appear compressed video: the P-frames and B-frames degrade
at a regular interval. P-frames can be very small relative to quickly but the I-frames show the least compression and
I-frames assuming that little motion occurs from frame to appear with the least amount of quality artifacts on a recur-
frame. ring basis. For the most part, the perceptual phenomenon
Bidirectional or B-frames are sometimes used in of visual memory makes interframe compression extremely
encoded video between I-frames. The bidirectional effective; we remember details from previous frames and
characteristic means that they reference pixel informa- primarily follow pieces in motion.
tion from the preceding frames and the ones that follow.
They can be compressed even more than P-frames and

147
 s ecti o n 1: Captu r e

Chroma Subsampling
Our outsized sensitivity to luminance means that, par-
ticularly when it comes time to keep file sizes small and
convenient, color data is simplified or otherwise made a
lesser priority. That doesn’t necessarily translate to observ-
able quality degradation but it is important to appreciate
how effective compression is accomplished. The story is
the same for digital motion video: it’s costly to store all of
the bits to describe every pixel luminance and color and
every location when there are tens of frames to store for
every second of play time. The forms of video encoding
encountered by photographers employ chroma subsam-
pling to slim down the amount of information needed to
render a full color video. Chroma subsampling is a type of
lossy compression. Skipping chroma subsampling is pre-
ferred only when the post-processing workflow needs the
maximum amount of editing headroom.
The video data is divided into a luma (tone) component
and two chrominance (color) components. Video uses
a sampling ratio shorthand to describe exactly how the
chrominance information is subsampled that we describe
further in Chapter 12.

Video Codecs
Figure 6-7 A single frame from a video clip with a motion vector
overlay. The motion estimation data is used to efficiently encode
P-frames and B-frames. Video codecs are responsible for systematically packaging
the video data and for its equal and opposite unpacking.
They are an outside force that the data relies upon for
reading and writing it. The encoder part of the codec packs
a video into a bitstream of compressed data for storage and
transmission. The decoder unpacks and decompresses the
bitstream, often on-the-fly during playback, for viewing.4
An encoded video takes up less space on disk but requires
a decoder to be a playable asset. The goal is always to
minimize the data stored and transmitted while maintaining
Figure 6-8 A group of pictures structure for sequential video frames
using interframe encoding. The GOP is measured from an I-frame video quality. Codecs are used for videos, videoconferenc-
up until (but not including) the next I-frame. This illustration shows a ing and live streaming. The most commonly encountered
GOP size of 12.
video codecs are described here. Any time that an encoded
video file is re-encoded with different parameters (for exam-
ple, a lower frame rate or frame size) it’s called transcoding.
Transcoding can compound or exacerbate compression
artifacts and must be minimized in a video workflow.

148
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

Motion JPEG it across the same bandwidth previously used for an AVC
The easiest video codec to remember is Motion JPEG (or video. The 50% efficiency number is touted in market-
MJPEG) because its name indicates what you’re getting. ing despite real-world results yielding closer to 35–40%
This codec does not use any interframe encoding and improvements at comparable quality.5 This is considera-
instead treats each frame exactly how stand-alone image ble, though it’s always a good idea to test and evaluate
files are compressed. It is a sequence of JPEGs with each compression strategies with your specific content. A 2017
frame compressed independent of one another (intraframe survey of video developers showed that 28% were working
encoding). MJPEG is like having a video that’s entirely with HEVC in their workflows.6 HEVC is supported by MOV
I-frames (like a favorite childhood breakfast cereal com- and MP4 containers.
posed exclusively of the marshmallow bits). Motion JPEG is HEVC uses smaller subdivisions than macroblocks for
not practical because of its large file sizes, yet we mention compression (transform blocks, further subdivided into pre-
it because it’s such an easy logic jump from working with dictive blocks) and forgoes the straightforward macroblock
JPEG-compressed photographs. encoding approach for a coding tree algorithm. It also
offers motion prediction vectors more advanced than AVC
encoding. These help to compress more effectively but add
MPEG-4 Advanced Video Coding (H.264) encoding and decoding complexity. Early use was limited
The most commonly used codec today is MPEG-4 Advanced to desktop computer platforms that guaranteed a level of
Video Coding (AVC) also known as H.264. Its ubiquity is due computing power to decode for smooth playback. Mobile
to its flexibility and performance; AVC is a family of profiles processors have caught up, though, and are now up to the
that fit a variety of video needs like minimal disk storage or task. HEVC is not the only candidate available for high-qual-
low bandwidth streaming. The codec compresses frames ity, high bitrate motion video, though it has increasing
at the macroblock level and uses motion prediction to support in streaming media applications. AVC is capable of
efficiently store parts of the image that change while encoding large frame resolutions like 4K and 8K, however,
keeping unchanging parts the same. AVC compression is you’ll find that these data-heavy formats increasingly opt
widely supported by browsers, televisions, mobile devices for HEVC encoding instead. This is to minimize the band-
and anything else you find yourself watching video media width needs as our hunger for larger video grows. HEVC’s
on. Its compatibility makes it a great choice when sharing official successor is already in the works and is tentatively
through platforms where the content creator doesn’t have called Versatile Video Coding (VVC).7
precise control over the viewer’s decoding setup. In con-
stant use for well over a decade, AVC is an excellent default
choice for video encoding. The codec is supported by the VP9
common video container formats MOV and MP4. Concurrent to the development of HEVC, Google created
its own video codec for the modern web and beyond
called VP9. There are eight versions that precede it, many of
High Efficiency Video Coding (H.265) them developed by On2 Technologies (which was bought
The successor to the ubiquitous H.264 codec is H.265, by Google in 2010). It’s an open-source standard in contrast
formalized in 2013. It has a secondary name, too: High to AVC and HEVC, which require royalty payments from
Efficiency Video Coding (HEVC). This codec is capable of companies that use them to serve up thousands of videos
compressing video to half the size that AVC is able to on our phones, televisions and websites. Given that Google
encode. This means that the same video file requires half owns YouTube, it is not surprising that they have joined the
of the bandwidth or storage when encoded with HEVC fight to optimize video encoding while avoiding royalties
over AVC. Alternatively, it means that we can make a video or dependencies on outside parties.
twice as large (say, through higher resolution) and stream

149
 s ecti o n 1: Captu r e

AOMedia Video 1 (AV1) recorded at different bit depths (8, 12, 16, 24) that dictate
Very recently finalized, as of this writing, AOMedia Video 1 the maximum number of bits used to define each sample,
or AV1 was developed by the Alliance for Open Media con- similar to image sensors and photographs.
sortium. The consortium designed AV1 to replace VP9 as a A video file container holds an encoded video bitstream
royalty-free, next generation video codec in direct competi- alongside an encoded audio bitstream. This audio stream
tion for widespread adoption against HEVC. Its specification is compressed and packaged with its own codec that is
requires a low computational footprint that is optimized similarly decoded upon playback; audio also inherently
for mobile hardware to produce consistent, high-quality contains redundancy that can be simplified with com-
real-time video delivery that’s scalable to any bandwidth. pression algorithms. The Free Lossless Audio Codec (FLAC)
AV1’s development team expects an approximate 30% is an open source, lossless option. Not all audio playback
improvement over VP9 and HEVC, though current usability hardware or software is capable of decoding FLAC data
is limited due to very long encoding times. and just like TIFF files in the photo world, FLAC audio can
be cumbersome overkill for most end-user applications.
MP3 is a well-known lossy compression codec and file
Audio in Video container that stems from the same encoding protocols
as AVC. Advanced Audio Coding or AAC is MP3’s successor.
Sound is digitally recorded monoscopically (one micro- It’s a lossy codec that produces higher audio quality with
phone, one direction), stereoscopically (two microphones comparable file sizes to MP3. AAC is the default format for
for two-channel playback) or spatially (surround sound or some operating systems and web platforms for delivering
binaural). Stereo audio begins to emulate our two audio audio data.
sensors: our ears. However, our sense of space in three
dimensions is due in part to our incredibly nuanced ability
to sense small differences in volume and direction. The Bitrate and File Size
more we can sample audio in all directions and play it back
as if it’s originating from those many directions, the more Video encoding is often tuned for data transmission. After
immersive and realistic the experience. Professional pro- all, people spend hours watching video content streamed
ductions do not rely on built-in microphones because of over the internet, cellular data connections or cable televi-
their relatively poor sound recording quality and because sion. Playing video files from a local storage device is not the
they are fixed to the camera body itself. If the sounds of primary means of video playback unless we’re the content
interest are not near the camera, the recording ability creator. The rate at which data is transmitted is critical: if
becomes severely limited. Many cameras offer plug-in frames can’t be transmitted and decoded quickly enough
microphone jack options, at the very least, though serious during real-time, on-demand playback, you’re left watch-
productions leave the audio recording to completely inde- ing a buffer animation spin at regular intervals. The only
pendent hardware including digital audio recorders. Audio alternative is to transmit a video file in its entirety before
tracks are synchronized and combined with the visuals in beginning playback. Reflecting on this, the bitrate is a critical
post-production. number when working with video media and it is measured
Similar to video, audio is an analog phenomenon sam- in megabits or gigabits per second. The more data that
pled, recorded and converted to a digital signal that must must be transmitted by satellite, ethernet or USB cable, the
be encoded and decoded. The higher the sampling rate, more content producers weigh the amount of compression
measured as a function of samples per second, the higher and the resulting file size of motion video media. The lower
the audio fidelity. The higher the sampling rate, the more the bitrate, the more video quality artifacts are encountered.
data generated and stored. Common audio sampling rates The higher the bitrate, the greater the potential for excellent
are 44.1, 48, 96 and 192 kHz. The digital audio signal can be video quality with few compromises.

150
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

Table 6-2 YouTube’s recommended bitrates for given resolutions and frame rates.

Video Size Bitrate for 24/25/30 fps Bitrate for 48/50/60 fps8

2160p (4K) 35–45 Mb/s 53–68 Mb/s

1440p (2K) 16 Mb/s 24 Mb/s

1080p 8 Mb/s 12 Mb/s

720p 5 Mb/s 7.5 Mb/s

Bitrate must be considered in context: a high-quality subdivided into two groups: temporal and spatial. Artifacts
720p clip encoded at a bitrate of 10 Mb/s does not mean are any noticeable aberrations or defects in the image. A
that a 1080p or 4K clip encoded with the same bitrate is spatial artifact is observable when the video is paused or
comparable in quality. Bitrate dictates the amount of data in motion. Spatial artifacts include ringing, color bleeding,
per second, and larger frame sizes or higher frame rates blurring, blocking or repeating patterns.9 Temporal artifacts
equate to more data generated. Always consider bitrate span a duration of time in playback and include jerkiness or
relative to the data being compressed. Bitrate should judder and mosquito noise.10 You may be more acquainted
increase as video data grows if there’s an expectation of with video artifacts than you think thanks to years of online
consistent quality. Encoding a 4K clip at 10 Mb/s requires video streaming binges. Many artifacts specific to video are
compressing the data much more aggressively than a the result of quantization errors in the encoding-decoding
720p version. Table 6-2 shows YouTube’s encoding bitrate process inherent to the medium. Let’s describe these video
recommendations for common video resolutions as of quality artifacts in greater detail.
September 2018.
Strategies get more elaborate when considering video
delivery over an inconsistent or unreliable variable band- Blocking
width network. Video streaming services use strategies Blocking is a spatial artifact that also goes by the names
like adaptive bitrate streaming that switch out versions of checkerboard, tiling, macroblocking or pixelating.11 Blocking
the content encoded at tiered quality levels so that you is the result of compression and is usually the first image
never stare at a buffering screen even when your internet defect that viewers notice in compressed video. We find
connection speed fluctuates. This topic extends beyond blocking in JPEG images at lower-quality settings but
that which a photographer is likely responsible to consider their changing appearance from one frame to the next in
and we mention it only to garner some appreciation for the motion video is more distracting. Recall that a video frame
technology that makes our modern video consumption is divided into groups of pixels, macroblocks, and each
habits possible. macroblock is independently quantized. Some detail loss
is expected relative to the video feed prior to encoding.
We’re not likely to stare at a single video frame and note
Video Quality Artifacts loss of detail, necessarily, though we are quick to notice
how macroblock boundaries show abrupt changes or dif-
Motion video shares a host of potential video quality prob- ferences. Blocking is particularly apparent in high frequency
lems with photographs like flare, exposure clipping and content areas because the independently coded blocks
loss of sharpness. All of the technical challenges of using show artificial, grid-like boundaries where there should
optical components, sensors, electronics and post-pro- be organic, unstructured detail (see Figure 6-9, overleaf).
cessing software exist in video as they do in photography. Some video decoders use active deblocking filters that
Video also has a unique set of quality artifacts that are introduce blurring in areas where these compressed block

151
 s ecti o n 1: Captu r e

new data arrives and gets decoded. Entire frames failing to

record or decode on playback are considered frame drops.
Frame drops are distracting because they introduce a tem-
poral discontinuity.

Ringing
Ringing spatial artifacts appear as halos or ghost-like rings
along subject edges, not unlike oversharpening halos seen
in photographs. The ringing halos are relatively static from
one frame to the next. These are a consequence of lossy
encoding and decoding as the algorithms can overshoot
Figure 6-9 Common AVC-encoded blocking artifacts are
particularly apparent when using low bitrates, when the video shows or undershoot edge reconstruction; even high-quality, high
fast-moving objects or both. bitrate videos can have ringing artifacts.

boundaries do not align with content edges. If such a filter Mosquito Noise
is too aggressive or if the blocking itself is strong, viewers Mosquito noise is a temporal artifact seen around high
describe the resulting video as blurry or fuzzy. Interframe frequency details and appears to move and fluctuate.
compression means that I-frames are the least compressed The differing compression characteristics along distinct
in a group of pictures. If a video is encoded with aggressive boundaries causes this artifact to flicker from frame to
settings to keep the file size low, the I-frames can suffer frame, looking like busyness and noise specific to edges.
from macroblocking and all P-frames and B-frames that This occurs when blocks contain some combination of
reference them perpetuate the artifacts. content like an edge and a background; the compression
algorithm may waiver from one compressed appearance
in a frame to another in an attempt to represent the
Packet Loss conflicting content frequency. This is called quantization
Blocking can also happen when there’s data loss, corrup- error.12 Mosquito noise is often mistaken for ringing which
tion or interruption whereby some small blocks of image is a spatial artifact.
frame information don’t make it to the final stage of decod-
ing and playback. This is called packet loss or dropout. While
distinct from normal blocking in the sense that it’s lost data, Color Bleeding
not overly simplified or compressed data, the visual effect The consequence of chroma subsampling, color bleeding
can be similar. It appears as small blocks, sometimes clus- might remind you of a child’s coloring book where the
tered together, getting “stuck” or failing to update for one colors don’t stay inside the lines. Since color information
or more frames. Packet loss is occasionally seen during live is sampled at a lower resolution and interpolated up to
broadcast digital television. As long as it’s a fleeting event match that of the luminance data, there can be a mis-
and it doesn’t happen frequently, most viewers do not match. Sometimes the interpolated color information
consider this artifact to be objectionable. When enough extends beyond subject edges and looks like the color is
data is lost or dropped either in transmission or because leaking or bleeding into other content. If seen at all, this
the decoder failed to decode at the correct pace, those spatial artifact is relatively small in the video frame.
small frozen blocks can turn into entire frames. This is called
freezing where the imagery on screen stops updating until

152
 c h a p t e r 6 : M ot i o n V i d e o F u n d a m e n ta ls

Jerkiness Short Form or Micro Video

Jerkiness, also called judder, appears as wobbly or incon- It’s impossible to ignore the popularity and “stickiness”
sistent motion from frame to frame. The main cause is the of animated GIFs and micro videos, that is, video content
conversion between 24 frames per second and 30 or 60 with very short durations.13 The video subformat saw
frames per second using an interpolation strategy called success with the now-extinct social media platform Vine
3:2 pulldown. Since neither 30 or 60 divide neatly into 24, (2013–2017). Vine encouraged users to create videos with
straightforward frame duplication or elimination is not pos- a maximum run time of 6 seconds. The videos looped
sible. The 3:2 pulldown technique creates synthetic frames indefinitely on playback. Similar features have cropped up
by combining portions of existing ones. The end result in other platforms like Instagram, Snapchat and Facebook.
may suffer from jerkiness that disrupts the natural smooth Mobile camera apps include modes to capture “living
motion appearance expected from video. photographs” that function as assets straddling between
photographs and full videos. They tend to be either GIFs
or traditional video formats with the latter offering better
Emerging Trends in Motion Video image quality and compression efficiency.
Short form videos may feature lower than normal
Motion video is a dynamic medium with a creative and frame rates (5–15 fps) to keep their data profile low.
technological momentum from the last decade that The looping characteristic means that very short events
parallels still photography. Smartphones, streaming video can translate to a digestible viewing experience. Once
services and a seemingly exponential growth of screens in considered a passing trend, micro videos have proven a
our lives means that video is a powerful presence. Here are formidable format in our evolving visual media lexicon.
a few areas where video is evolving or expanding beyond Short form videos or photographic animations can be
its traditional bounds. found in advertising, social media and fine art bodies of
work. What started out as a product of technological and
bandwidth limitations has now become a unique form of
Wide Color Gamut and High Dynamic Range temporal imaging.
4K televisions have been logical upgrades for living rooms
and are increasingly adopted by consumers. The increased
resolution alone doesn’t always bring with it a perceived Immersive Video
improvement in video quality or clarity due to its depend- Motion video is always defined as a two-dimensional visual
ence on proper viewing distance. Today, displays offer a medium viewed on displays and projection screens. This
combination of improvements, of which the increased may be changing as accessible technology now allows for
resolution is just one. Wide color gamut (WCG) promises a video capture rigs that record 180° and 360° field of view
wider, more saturated and varied color palette that breaks scenes in stereo (see Figure 6-10, overleaf). Ultra-wide-an-
from the color limitations that video has adhered to since gle lenses and action cameras like the GoPro Inc.’s HERO
the 1990s. High dynamic range video capture and display jump-started an era of immersive content and signaled
also promises to expand the tonal range to more closely exciting capabilities for taking tiny image sensors to places
match reality. This necessitates an increased bit depth from and perspectives that motion video hadn’t yet reached.
the standard 8-bits today up to 10- or 12-bits. The visual Depth data for stereoscopic or volumetric video is on the
impact when these new technologies are combined is horizon to combine three-dimensional video with com-
exciting and impressive but requires newer equipment and puter generated elements. Virtual and augmented reality
processing to generate content. We expect that motion technologies are blending the ideas of traditional viewing,
video will forge ahead with better color, wider dynamic user experience and interactivity. Immersive video is a
range and higher resolution and never look back. dynamic branch of filmmaking that is poised to augment

153
 s ecti o n 1: Captu r e

Figure 6-10 A student-built prototype 360° camera based on plans published by Facebook, Inc. called Surround 360.14 An array of 17 wide-
angle video cameras capture simultaneously and the footage is stitched and blended in post-processing. Photographs by Rochester Institute of
Technology photography alumnus Nick Franco

and grow our definition of what motion media looks like Creating-a-Recipe-for-Codec-Success-to-Please-All-
and how we experience and engage with it. Palates-128974.aspx.
8 “Recommended Upload Encoding Settings - YouTube
Help.” Google, Google, support.google.com/youtube/
Notes answer/1722171?hl=en.
1 Phillips, Jonathan B. and Henrik Eliasson. Camera Image 9 Unterweger, A. “Compression artifacts in modern video
Quality Benchmarking. John Wiley & Sons, 2018, p. 29. coding and state-of-the-art means of compensation.”
2 Salmon, Richard, et al. “High frame-rate television – BBC Multimedia Networking and Coding, 2012, 28–49.
R&D White Paper 169.” BBC, Jan. 2008, www.bbc.co.uk/rd/ 10.4018/978-1-4666-2660-7.ch002.
publications/whitepaper169. 10 Urban, John. “Understanding video compression artifacts.”
3 Raikes, Bob. “Artistic intent and aspect ratios.” DisplayDaily, Component, Biamp.com, Feb. 16, 2017, blog.biamp.com/
vol 24, no. 28, Aug. 1, 2017, www.displaydaily.com/article/ understanding-video-compression-artifacts/.
display-daily/artistic-intent-and-aspect-ratios. 11 Punchihewa, Amal and Donald G. Bailey. “Image and vision
4 Anil, C.R. and B.R. Krishna. H.264 Video Coding Artifacts – computing.” IVCNZ, Image and Vision Computing New
Measurement and Reduction of Flickering, www.diva-portal. Zealand: IVCNZ 02, ser. 197–202, 2002.
org/smash/get/diva2:833137/FULLTEXT01.pdf. 12 Unterweger, A. “Compression artifacts in modern video
5 Ozer, Jan Lee. Video Encoding by the Numbers: Eliminate the coding and state-of-the-art means of compensation.”
Guesswork from Your Streaming Video. Doceo Publishing, Multimedia Networking and Coding, 2012, 28–49.
2017, p. 212. 10.4018/978-1-4666-2660-7.ch002.
6 Vernitsky, Tanya. “2017 video developer report.” Bitmovin, 13 Pogue, David. “Why are micro movies so
Sept. 11, 2018, bitmovin.com/2017-video-developer-report/. popular these days?” Scientific American, May
7 “Creating a recipe for codec success to please 1, 2013, www.scientificamerican.com/article/
all palates.” Edited by Jan Ozer, Streaming Media why-micro-movies-so-popular-today/.
Magazine, Dec. 7, 2018, www.streamingmedia. 14 Briggs, Forrest. “Surround 360 is now open source.”
com/Articles/Editorial/Featured-Articles/ Facebook Code, July 26, 2016, code.fb.com/
video-engineering/surround-360-is-now-open-source/.

154
Section 2
Process
This next section covers the many aspects of image pro- The original, recorded brightness values at a given pixel
cessing that our captured image files undergo, from both location are far from rigid and there is great power in get-
a technical and creative viewpoint, before they’re ready for ting the most from our photographic data. A wide world of
output. This continues our exploration of the photographic software filters exist to further modify and expand captured
processes and technologies following our look at image image content: we look at the mechanics of sharpening,
capture. You’ve made countless decisions leading up to the blurring and blending operations as well as black and white
shutter button press and there are countless more to make conversion, special effects and filter masking for compos-
for functional and aesthetic objectives at the processing iting and retouching. Software filtering offers a dynamic
stage. potential for making photographs that go well beyond
We start with demosaicing and interpolation: the very traditional capture or objectively faithful reproduction.
first steps of unpacking the digital image information Understanding how a camera performs and how we
captured by the camera. Photographers are lucky enough define quality images means describing the imaging pipe-
to have computer scientists and commercial software line and deconstructing the tasks handled by the camera’s
products that make downloading and viewing raw image image signal processor. We review methods of testing
files a seamless experience. It’s valuable to understand all of sensor performance and identifying key image quality
the hard work and image processing that makes this pos- characteristics like color reproduction, dynamic range and
sible, particularly when there are decisions for improving noise. Since the sensor is one player on a team, we also
image quality that rely on knowing what happens under quantify lens quality and interpret lens performance.
the hood. Finally, we demystify image compression—a topic that
Practicing photography benefits from an informed is easily overlooked and underappreciated in its influence
and planned workflow called digital asset management. on quality, data storage and workflow efficiency. Pho-
Creating digital data necessitates a degree of literacy in file tographers should understand compression beyond a
management, computer hardware and storage solutions. JPEG quality slider; chapter challenges image-makers to
We also review file formats, metadata, image editing soft- consider forms of data redundancy and our strategies for
ware options and computer components. exploiting them to keep files small. We also review the
An image’s initial form is not its destiny: it may have its common image quality artifacts that result from image
tones redistributed, its content manipulated, sharpened, fil- compression.
tered or compressed (and possibly all of those combined).
 7 Demosaicing
and Interpolation

Photograph by Rochester Institute of Technology photography alumnus Josh Shagam

We explored the design of digital sensors including the Bayer pattern color filter array in
Chapter 3 and in Chapter 4 we looked at strategies for obtaining perfect photographic
exposures. We bring those concepts together here to explore how an image is transformed
from sensor signal to a processed image. Getting to the point where our digital image
approximates the appearance of the scene as we experienced it takes a considerable
number of steps. These steps are often invisible and unknown to photographers since they

157
 s ecti o n 2: P roc ess

happen behind the scenes in software processing, yet an option of reduced-resolution raw files for archiving,
they’re critical to the successful formation of a photograph. however, these simply throw away some of the native
Furthermore, the digital information in an image file is pixels. The reality is that the camera dictates the baseline
modified even going into its final outputs as interpola- amount of photo information you’ve got to work with in a
tion is used for upscaling or downscaling pixel data. This raw capture.
chapter dives into the details of a raw file’s contents and
how they’re translated into a functional image file as well as
strategies for resampling the data for output. Image Interpolation
In the analog realm, pressing the shutter exposes light
sensitive film to the scene. An exposure completes when Data interpolation is a method of creating new data based
the shutter closes. The resulting image exists as a latent on known data. Interpolation is used during two different
image on the film. A chemical change occurs but further stages in the digital photographic workflow: when convert-
events must take place before that image is visible to the ing a raw image file into an editable, viewable photograph
eye. These events involve chemistry and development and when resizing an image for output or further manipu-
solutions that change the photograph from a latent image lation. Before describing these use cases, let’s establish the
to a visible one that can be enlarged, printed and viewed. basics of image pixel data interpolation.
Despite a difference in materials and technologies, the life In the simplest terms, interpolation looks at the exist-
of a digital raw image shares some parallels with this trans- ing pixels and uses their values to make an educated
lation from potential to realized image data. guess at the unknown or missing pixel data. Despite
sounding straightforward, this is a mathematically
complicated operation. Consider just the green pixels
Contents of a Raw File in a raw image to start. Recall that the color filter array
provides every other green pixel value when capturing
A raw file is generically defined as the uncompressed, the raw photosite data. Once recorded, the missing data
minimally processed image data and associated metadata in between the green-filtered photosites must be recon-
captured by the camera. The raw file is not really an image structed. The missing data is represented by question
yet; it can’t be viewed without special software such as marks in Figure 7-2 (overleaf). Taking each line separately,
Adobe Camera Raw. A raw image is also inherently gray- we could average the pixels on either side to create the
scale, as shown in Figure 7-1, because the camera sensor missing values. The first line becomes all values of 24 and
detects and records an accumulation of light energy, not the second row all values of 36. What about the third row?
color. Raw image files must contain a header (the first The first unknown pixel on this row is between pixels
portion of the file’s content) that provides the information with values of 12 and 19. Averaging these two yields 15.5,
necessary to process it into a viewable and editable image. a problematic number because it’s not possible to have
This includes the sensor size, the configuration of the color non-integer pixel values in an image file. A decision must
filter array (CFA) and the color profile. Raw files also host a be made: is this unknown pixel going to have a value of
collection of metadata including all of the camera settings 15 or 16? An alternative approach is to take the average
and exposure details. Raw file formats themselves are often of all four surrounding pixels ([36+12+19+13]/4). This
proprietary to the camera manufacturer despite sharing yields a new value of 20 for the unknown pixel, making it
these fundamental components. brighter than three of the neighbors. This result may be
A raw file’s native pixel resolution is established at problematic, too.
capture based on the physical number of active photo- Now focus on the blue and red pixels: there are even
sites on the sensor. For example, a 45.4-megapixel sensor fewer of those to work with. How do we create the pixel
creates a file that’s 8256x5504 pixels. Adobe DNG offers values for an entire row of the image when there are no

158
 c h a p t e r 7 : D e m o s a i c i n g a n d I n t e r p o l at i o n

Figure 7-1 A raw Bayer image as it is recorded and stored by the camera.

159
 s ecti o n 2: P roc ess

Figure 7-2 The pattern of green, red and blue pixels in a raw image. The question marks represent unknown data needed for full color
channel resolution.

values from which to average or estimate? This is where color filter array is not used to capture an image, negating
interpolation and demosaicing become intertwined. Filling the need for demosaicing. For those wanting a deeper
in all of the missing pieces is a complex and nuanced set of explanation of the mathematics and sampling theory
operations. involved, consider available texts such as Digital Image
Compression Techniques (1991) by Majid Rabbani or Intro-
duction to Data Compression (1996) by Khalid Sayood.
Demosaicing Image Data We are much more sensitive to changes in luminance
(brightness) than we are to color. We are also more sen-
A mosaic is a picture or pattern created by placing small sitive to green than red or blue. This is a helpful starting
color pieces or photographs close together. When viewed point for a demosaicing strategy, so most algorithms start
at the proper distance, the individual components blend with the information recorded by the green-filtered photo-
together to reveal a unique image. When we take a pho- sites known as the green channel in the image file. The data
tograph through a color filter array, the result is a mosaic in this channel is analyzed for subject edges and brightness
comprised of either a red, green or blue pixel value at each trends usually through the use of variance calculations.
sensor photosite location. The process of demosaicing a Variance is a measure of how different the data is within
color-filtered raw image creates a red, green and blue pixel a small area around the missing pixel data. Depending on
value at each location. Demosaicing uses color interpola- the results, the algorithm makes a decision on how to inter-
tion algorithms to accomplish this task. These algorithms polate groups of pixels. A low degree of variance in either
are mathematical approaches to calculating and predict- the horizontal or vertical direction indicates the presence
ing pixel information. There is no single solution to the of an edge. The unknown pixel information is calculated
problem. Consequently, there is no singular, perfect inter- using only the vertical or horizontal surround pixels along
polation algorithm to employ and computer scientists and that edge. If the variance calculations indicate a gradient, a
image scientists continue to develop different approaches combination of surrounding pixel values is used to interpo-
in an effort to improve results. Some algorithms actively late the unknown pixels to closely approximate the likely
look to preserve edge detail and look for other patterns in values at these locations.
tone or color that may indicate content such as faces or Next up: the green channel information is combined
fabric. with the red-filtered pixels to fill in the red channel. The
Photographers don’t need to dive into the math of reconstructed green channel is used to find edges and
these algorithms, so we instead cover the key aspects of gradients in the image to make decisions to recreate the
images that demosaicing and interpolation algorithms missing red pixels. The blue channel is typically left for last
depend on. Also keep in mind that there are times when a because our camera sensors lack sensitivity in this region

160
 c h a p t e r 7 : D e m o s a i c i n g a n d I n t e r p o l at i o n

of the visible light spectrum. The raw exposure information appearance as the shadow areas are mapped into the
from blue-filtered photosites tends to have the most noise midtones and noise can become more visually apparent.
due to this insensitivity and using noisy data to make var- The second method is to treat the shadows, midtones and
iance calculation decisions proves difficult. The end result highlight areas separately to minimize the amplification
is a digital image file with a red, green and blue channel of the noise in the shadows and prevent clipping in the
brightness pixel value at every location. highlights. On-camera processing following demosaicing
The demosaicing process may also include color correc- may include some targeted exposure adjustments to satisfy
tion to compensate for the differing sensitivity of the three a preset “look and feel” like a high contrast profile.
color channels. This is due both to the sensitivity of the
silicon and the unique spectral characteristics of the color
filters used in a Bayer filter array. A color correction matrix Color Space Mapping
(CCM) corrects for this inconsistency and maps the image What’s the right color space to set for camera capture?
to a selected color space. First, we need to establish the concepts of color spaces. A
Demosaicing is one of the most computationally color space, in terms of a digital photograph, is the pool of
complex steps in the image processing pipeline. A highly colors available to represent the original scene. Cameras
optimized version of demosaicing happens after capture, typically offer two choices: sRGB and Adobe RGB. ProPhoto
on the camera, in order to render a preview image. Still, we RGB is a third color space used in photography more com-
rely on our more powerful desktop computer processors to monly seen in desktop editing rather than with on-camera
attend to the task in earnest when we’re back in the studio capture settings. Figure 7-3 shows all three color spaces
or home office. as a function of their color gamuts plotted on the CIE

Processing Steps Following

Demosaicing

Demosaicing the color-filtered photosite data is an initial

step in processing the raw image into something viewable.
Additional steps may include adjusting any or all of the fol-
lowing: exposure, color translation, white balance, contrast,
sharpness and noise reduction (both luminance and color).

Exposure Adjustment
Exposing your photograph correctly at capture always
provides the best results for the final image. However, there
is still hope if you failed to capture the perfect exposure.
Exposure (in the form of redistributing pixel brightnesses)
can be adjusted somewhat during the ingestion process
with most software packages. There are two general
methods for achieving this. The first method uses a simple
multiplication of all the pixel values in the image, effec-
tively shifting the histogram to the left toward brighter Figure 7-3 ProPhoto RGB, Adobe RGB and sRGB color spaces
values overall. This can have negative effects on the image plotted on a chromaticity diagram.

161
 s ecti o n 2: P roc ess

chromaticity diagram representing the gamut of human on its own for making photographic images. Viewing a
vision. photograph with a linear luminance mapping looks odd
The sRGB color space was developed in 1996 by Hewl- and incorrect to us due to the human visual system’s
ett-Packard and Microsoft Corporation for use with digital nonlinear response to light. Left unaltered, a properly
images. It is smaller than the Adobe RGB color space but exposed raw image looks dark and low in contrast relative
is the best choice if the intended output is display and to our visual experience of the scene. Therefore, a gamma
web use—for decades, most displays were only capable correction is applied to an image when it is processed
of reproducing the sRGB color space and it is a reliable using Equation 7.1.
lowest-common-denominator assumption. The Adobe RGB
color space was created by Adobe in 1998 to encompass Vout = Vingamma (Eq. 7.1)

all of the possible colors reproducible using commercial

printers. The Adobe RGB color space is approximately Where
35% larger than the sRGB. The ProPhoto RGB color space
was developed by Kodak and is by far the largest of the Vout = output luminance value (7.1)

three color spaces. This space exceeds the bounds of the Vin = input luminance value
chromaticity diagram. It is favored by some when editing
images because of its extra overhead, even though no This second gamma operation happens when the
output device is capable of reproducing data that sits out- image is displayed on a monitor or projector. The industry
side of an output device’s reproducible color gamut. standard for monitor gamma is 2.2. The gamma correction
Color spaces come into play more when we get into is conducted automatically by the computer’s operating
editing and output, yet it can be important to establish the system based on the display profile configuration.
right choice at capture. With a raw file, the color space is
specified during the demosaic process. Choosing the color
space setting when shooting raw offers a risk-free option Noise Reduction
of reprocessing in a different space if desired. Capturing in Noise, or random fluctuations of luminance and color, is
JPEG means that the camera’s output bakes the color space always present in photographs. Noise often negatively
into the file with no way to go back to a larger one. In impacts the reproduction of detail and can decrease con-
other words, forcing the decision to restrict a JPEG to sRGB trast. Excessive noise can also cause trouble with some
means limiting its potential color gamut from its earliest interpolation algorithms that use edge detection during
stages of life. the demosaicing process as it can lead to incorrectly iden-
tifying edges. There are many different methods available
to remove noise such as convolution with a low pass
Gamma Correction filter, gaussian filter or a bilateral filter. The explanation of
Gamma correction defines the relationship between the the statistics of these methods is beyond the scope of this
numerical pixel values in the image to the actual lumi- text.
nance. There are two gamma operations involved. The Noise reduction is a give-and-take operation, presenting
first is when the image data is processed in-camera and a tradespace between the degree of retained detail and the
is called gamma encoding. The camera sensor’s linear amount of noise present in the final result. Too much noise
response means that when a pixel receives twice the reduction softens the image by smoothing and averaging
amount of photons, it produces twice the amount of pixel variance that previously described true detail. Not
signal. The relationship between photons and recorded enough noise reduction can leave a distracting amount of
signal is linear. This is great if the sensor is used as a meas- noise. Noise reduction processes are often done prior to
urement device for scientific research but it is not as useful CFA interpolation to prevent the propagation of noise into

162
 c h a p t e r 7 : D e m o s a i c i n g a n d I n t e r p o l at i o n

the demosaicing stage. If left until after demosaicing, noise is the very first non-destructive decision made to raw files.
may be inadvertently exaggerated or amplified. As such, an old raw file can be taught new tricks if passed
through a newer or more advanced demosaicing algo-
rithm. An overly noisy photograph from today could look
Anti-Aliasing Filter much better five years from now as long as you hold on to
Most cameras using a CFA also use an anti-aliasing filter, the raw, unprocessed data (though don’t expect miracles).
called an optical low-pass filter (OLPF), on top of the imag- Applying interpretation algorithms to raw files allows us
ing sensor. We sample the scene at discrete points when the flexibility of reprocessing an image when algorithms
capturing an image. This is a functional approach for much change or to try an altogether different algorithm if we
of a scene, however, when we approach an edge or a aren’t happy with the initial results.
repeating pattern, the sampling rate (the number of pixels
on the sensor) may not be enough to recreate the original
details. Resampling an Image
This sampling rate failure can result in moiré artifacts
or what some call “jaggies” along edges and fine patterns. In addition to demosaicing, interpolation is employed for
These artifacts occur because the frequency of these resampling or resizing an image: either enlarging it (upscal-
objects in the scene exceeds the Nyquist limit of the ing) or making it smaller (downscaling). We stated earlier in
sensor. The Nyquist limit is the highest frequency continu- this chapter that the native resolution of a raw file is set at
ous signal that can be recreated by a discrete or sampled the moment of capture and dictated by the sensor. There is
signal. The OLPF effectively blurs the signal to smooth the no way to generate true image content beyond what was
artifact, leaving us with image data in need of baseline originally recorded. All we can hope to do is interpolate
sharpening. and estimate additional pixels—which, while not “true”
information, is often just as good or is unnoticeable to
the eye. An 8256x5504 pixel photograph is not destined
Sharpening to exist exclusively at this size because we can upscale or
A degree of capture sharpening is applied to the image downscale the data using interpolation.
following the demosaicing process. The amount is deter- Figure 7-4 (overleaf) illustrates linear interpolation.
mined by a default preset in most desktop software but The plot on the left illustrates a simple linear regression
is typically adjustable on a per-image basis. It is a global and is an example of data that forms a perfect line. A linear
image adjustment that considers all pixels in the image regression fits an equation to the data in the format of
frame. Sharpening is required due to the blurring that y = mx + b. In this case, we wanted to find the value of y
occurs in the optical system at multiple stages of image when x was known and the result is a perfect fit. In contrast,
formation: the lens, the IR-blocking filter and the optical the right plot in Figure 7-4 shows a line fitted to data that
low-pass filter. This type of sharpening is broadly catego- is not linear that serves as a realistic representation of pixel
rized as capture sharpening and is just the first of multiple values in an image. Linear interpolation does a poor job
potential processing filters that aim to enhance the appear- here, showing that we need to use a more complex inter-
ance of sharpness in the final product. polation method to correctly predict pixel values when
upscaling image content.
Upscaling is primarily useful when sending photographs
Reinterpreting Raw with New Algorithms to physical print media output. To enlarge or upscale an
Recall that working with raw image files is part of a nonde- image file, we’re effectively looking to create additional
structive workflow whereby any adjustments introduced pixels using the ones we already have. It’s assumed that the
are instructions that can be changed or reset. Demosaicing raw data is demosaiced and that red, green and blue pixel

163
 s ecti o n 2: P roc ess

Figure 7-4 Interpolating data offers varying degrees of success based on the original data interpolated and the interpolation method used.

values exist at all locations. Increasing the pixel resolution dynamic web design serves the best-fit option. This also
of the image means creating additional pixel data. The helps to ensure that someone browsing an image gallery
same principles apply as with demosaicing, although the on a smartphone screen isn’t wasting data downloading
algorithms can be tailored to create smooth gradients or ultra-high-resolution images intended for display on large
preserve details depending on your need. screen televisions. We prepare the image file variants
Enlarging an image using interpolation is marginally using the interpolation methods described later that yield
useful for on-screen output applications. We’re quick to the best looking results and let the web browser request
notice pixelation in image enlargements and unless you’re the appropriate version. The photographer can pick the
starting from a source that was already created as a final interpolation method and degree of compression for each
output product (like a thumbnail icon taken from a web- image size in the set, watching for unwanted compromises
site), you likely have plenty of pixels in your photograph to to image quality. It’s a bit more preparation work for the
display on the web. Enlarging for print is a different story photographer but it improves user experience and it keeps
involving viewing distance, discussed further in Chapter 16. control over reproduction in the hands of the image maker
On-the-fly pixel resampling is possible though not ideal instead of the web designer.
for photographic content or any time where you want to
assert quality control over the viewing experience. Modern
web design uses scaling and elastic layout behaviors Nearest Neighbor Interpolation
(sometimes called fluid design) to accommodate the wide One straightforward method for increasing image reso-
variety of device sizes and user interaction models. If left lution is pixel replication. This is called nearest neighbor
up to web developers, our images might get interpolated interpolation. If the image needs to be twice its original
and resized freely, prioritizing layout and fit over image size, replication takes each existing pixel and duplicates it
fidelity. It’s better to stock a family of prepared images in an adjacent position. Each new pixel is the same value
on a server at predetermined sizes (think Small, Medium as its nearest neighbor. This method works well for graphic
and Large like you’re ordering fast food) whereby the design elements or text because of their well-defined and

164
 c h a p t e r 7 : D e m o s a i c i n g a n d I n t e r p o l at i o n

Figure 7-5 The original image data (left), upscaled to twice the pixel resolution (center) and downscaled to half the pixel resolution (right) using
nearest neighbor interpolation.

geometric edges. Enlarging photographs with nearest Bilinear Interpolation

neighbor interpolation, though, leads to an increasingly Bilinear interpolation is a middle-of-the-road algorithm
pixelated appearance due to the visual effect of enlarging that often produces better results than a nearest neighbor
the pixels. Note that it only appears this way because pixel approach. It uses the four neighboring pixels on the left,
data is duplicated without new or unique image content; right, above and below an original pixel value to calculate
pixels themselves never change size. One original pixel the new pixel values. Another way to describe its approach
becomes represented by two identical pixels. Subject is double linear interpolation because the algorithm per-
edges appear more jagged and small areas of detail appear forms a linear interpolation in the x-direction and then
blocky as a consequence of this approach. Nearest neigh- again in the y-direction to produce the new values. Bilinear
bor interpolation is the quickest method, computationally interpolation can lead to a reduction in overall contrast
speaking, for upscaling a file. It is not subjectively desirable which in turn can make the image appear less sharp to the
for photographic content. viewer. Although this method may produce acceptable
This method can also be used for reducing an image results with continuous tone photographs, it should not be
size. Instead of replicating pixels, nearest neighbor inter- used with line art or graphics.
polation for downscaling simply removes every other
row and column of pixel data. Figure 7-5 is an example of
both upscaling and downscaling. The original and result- Bicubic Interpolation
ing images are rendered at the same size for illustrative Bicubic interpolation produces high-quality results with
purposes. The center version doubles the size in both the minimal artifacts. Bicubic interpolation uses a 4x4 grid,
width and height relative to the original, increasing the the 16 pixels surrounding the original, in its calculations
image size from 5x5 to 10x10, by duplicating the rows and to fit a surface and find the new value. It weighs the
columns. This works well due to the simple and graphic nearer pixels heavier than the farther pixels to produce
nature of the original data. The third image in Figure 7-5 a much sharper image than nearest neighbor or bilin-
shows a reduction or downscaling of the original. The ear interpolation methods. There are more calculations
nearest neighbor algorithm removes rows 2 and 4 and involved compared to other methods, so it does take
column 2 and 4. The result is missing information pres- longer to run. However, its high-quality result makes it
ent in the original and significantly changes the image’s the favored and default choice by many image editing
appearance. applications.

165
 s ecti o n 2: P roc ess

Figure 7-6 The original image data (left), upscaled to twice the pixel resolution using bilinear interpolation (center) and with bicubic
interpolation (right).

Figure 7-6 uses our same starting sample target to photosites is the primary method of capturing digital
create an enlargement using bilinear and bicubic inter- images with the robustness and high-quality results gener-
polation. Here again, the original and resulting images are ated by demosaicing software.
rendered at the same size for illustrative purposes. Both Using interpolation to upscale an image is where
methods provide improvements over nearest neighbor, caution is required to avoid potential image quality degra-
however the detail of the white pixels is lost in both. dation. With a silver-halide-based film negative, it seemed
These pixels perhaps represent specular highlights in a that enlarging was limited only by the size of your dark-
photographic image: small but important details that can room enlarger (in reality, there is a degree of magnification
be lost in the process of interpolating pixel data. Bicubic that results in objectionable grain). Upscaling digital image
interpolation produces the smoothest transitions between information, effectively creating data that is not in the orig-
colors, making it a good choice for photographic content inal capture, can be necessary for making particularly large
and problematic for graphics with crisp edges and color prints and it’s helpful for photographers to understand the
boundaries. available options for making it happen. Using interpolation
to downscale an image, effectively removing detail relative
to the original capture, is also likely in the life of a photo-
The Life Cycle of Image Pixel Data graph as we look to varying platforms, screen sizes and
applications on which our work is viewed. Creating output
There are many similarities between analog and digital assets specifically prepared for these different outlets can
photography; the processes of demosaicing and image help in maintaining the best possible image quality.
interpolation are not among them. The idea of using a Demosaicing is the critical first stage of image data
Bayer CFA and creating an image with two-thirds of the maturation, though it is hardly the last. Thanks to
input data missing seems like it’s destined to fail. Instead, well-developed image science and computer science,
demosaicing algorithms have evolved and the number photographers have an array of interpolation tools to
of pixels on our sensors has grown to the point where it translate, scale and generally manipulate digital image files.
is difficult to find artifacts in the resulting images. Making A given raw capture may be demosaiced and interpolated
color photographs with CFAs and brightness-detecting a number of times over the course if its life.

166
 8 Digital Asset
Management

Photograph by Rochester Institute of Technology photography student Svea Elisha

167
 s ecti o n 2: P roc ess

Digital asset management (DAM) is a repeatable and sus- objects are defined by equations rather than discrete pixel
tainable methodology for storing, sorting and working with samples. Vector files allow for printing graphics at any size
digital assets. In photography, this involves downloading or scale and are not limited by pixel resolution. Photogra-
and saving images, naming files, organizing, rating and phers may find themselves working with vector files when
adding metadata. Photographs are what we often think collaborating with designers for publication layouts or
of as the most important digital assets though they are advertisements. Vectors are also employed when creating
not the only ones. Complete digital asset management certain layer masks or text in image editing software. Pow-
includes all files related to your workflow such as editing erful in their own right, vectors are not suited to storing
presets, catalog files, software user configurations, storage truly photorealistic image data and thus serve a secondary
media and even invoices and emails. Properly managing role in the world of photographic assets.
all incoming and outgoing assets is fundamental to a
professional workflow. It also requires careful planning
for backups and fail-safes. Digital assets are valuable but Image Data and Quality
vulnerable; hardware and software can (and frequently do)
fail. Digital asset management starts with understanding Captured photographs begin as raw digital counts stored
the data and conceiving strategies for using and managing in image file containers. The information is untouched by
it while minimizing the chances of mismanagement or post-capture image processing decisions and adjustments.
catastrophic data loss. These raw files contain uncompressed data and are the
The best DAM workflows are thoughtfully built around starting point for a potentially long journey of optimization,
the specific requirements of an image-making practice, be artistic interpretation and delivery to an output medium.
it personal or commercial. All require solutions for han- Raw files are large because each pixel in an array of millions
dling an ever-increasing torrent of digital data. We detail has its brightness information recorded and the possi-
file formats, ingestion workflow, data transfer, storage and ble values are numerous. Smaller and more universally
computer hardware in the sections that follow. compatible image formats are employed for showing pho-
tographs on the web and to simplify the amount of stored
information. Raw files are the purest archival format for
Defining Image Data storing the data as it was originally captured. Raw interpret-
ers or editors use this data to generate derivative working
We handle all manner of digital data on a daily basis, so image files without permanently altering it.
what specifically defines image files? Photographs are Image data stored in image file containers is further
always a type of bitmap. Bitmaps are strings of bits (the characterized by the type of data compression used to
building blocks of digital information) that describe dis- pack it into the smallest storage footprint. Uncompressed
crete color and brightness values of each pixel along with a data does not receive any type of optimization or size
series of instructions (the map) that explains where each of reduction and is the least space-efficient way to store
those pixels resides in a two-dimensional array. A photo- digital information. Lossless compression means algorithms
graph can have every single one of its many millions of are employed to lessen file size with no permanent loss
pixels be a unique brightness, color or both. The world we of the original image data; it is a reversible process to
photograph is highly detailed, complex and even random. compress and uncompress. Lossy compression means that
Bitmapped data is fundamental to image files because of some image data is discarded or simplified in a way that is
this degree of specificity. Another term for storing pix- irreversible. This may have negative image quality implica-
el-specific image data is raster graphics. Graphic designers, tions or it may be visually unnoticeable with the benefit of
in contrast, often use vector formats that store visual infor- smaller files. The goal of both compression approaches is
mation in a more mathematically descriptive way. Vector optimizing storage and delivery of image data. We would

168
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

constantly need to buy additional hard drives and spend Table 8-1 Common raw file formats with the camera manufacturers
our work days watching loading bars without compression. that create and support them.

Uncompressed data has a place in a photographic work- Raw Format Manufacturer

flow, particularly in video production, if decisions regarding
CR3, CR2 Canon
compression or post-processing efforts must be deferred
NEF Nikon
until some point after capture.
ARW Sony
Chapter 12 elaborates on how lossless and lossy com-
pression algorithms accomplish their goals. For now, we IIQ Phase One

simply appreciate that decisions in our digital asset man- 3FR Hasselblad
agement workflow are influenced by our desire to retain RAF Fujifilm
the best possible quality images between capture and ORF Olympus
output. To that end, we employ specific image file formats
MEF Mamiya
based on the tradeoffs between maximum quality, minimal
X3F Sigma
storage space and bandwidth requirements.

Image File Formats digitally converted light signals and additional metadata
as raw files. They are the closest thing to the sensor signal
Each image file format has its utility and drawbacks for that we wish to handle. Raw images are not intended to be
storing and displaying image content. Photographers final, deliverable assets as they are large (uncompressed)
frequently juggle all of the formats outlined here and and incompatible with the average viewer’s software.
consider the right time and place for each. All image files There is no single file container or extension associ-
have header substructures that precede the body or pay- ated with raw files. Instead, it’s a category of a file format
load sections (where the image data resides) containing similar to how a hatchback describes a type of car but not
supplemental information. Headers include descriptors like a specific manufacturer or model. Table 8-1 highlights
bit depth and pixel dimensions that are frequently needed common raw formats. Camera manufacturers use propri-
at a glance by both users and software. The payload etary file types that often necessitate proprietary software
comprises the bulk of the total data in an image file. Let’s to read and edit them. All support 12-bit or 14-bit image
define the relevant formats: raw, DNG, JPEG, TIFF, BMP, GIF, data. Some include encrypted information that helps their
PNG and HEIF. associated software yield better looking interpretations of
the image data. Luckily, popular image editing software for
photographers translates and works with a variety of raw
Raw file formats.
The raw image format is generically defined as the uncom- Along with the image data from the sensor, a raw file’s
pressed, minimally processed image data and associated header section includes byte-ordering information, an
metadata recorded by the camera. Despite the name, raw identifier and the offset that indicates where the data is
files are not completely unprepped. Think of it like buying located in the file structure. It also contains camera meta-
chicken breasts from the grocery store as starting with data to provide cataloging software with the image sensor
raw ingredients even though there was some gruesome size, color filter array information and capture metadata like
preparation involved before making it to the shelf. We exposure settings, camera model and lens model. Addi-
like to think that we’ve made dinner from scratch while tionally, raw files often house a reduced-resolution JPEG
glossing over some preceding steps that are too messy to preview the raw contents without the heavy computa-
and cumbersome to deal with. Cameras neatly package the tional lift of interpreting them. This preview version is like

169
 s ecti o n 2: P roc ess

the picture of a cake on a box of cake mix; it’s one possible The DNG format recently gained an additional abil-
outcome of working with the box contents but it does not ity to use lossy compression and/or reduced resolution
directly represent the contents as they exist. image data. This reduces the storage stresses of otherwise
All raw formats have similarities and yet there is no large raw files while retaining a version of the image that
guarantee that software like Adobe Photoshop or Capture is untouched by post-processing adjustments. DNG is a
One can interpret their contents. It’s up to the software popular format for smartphone cameras with raw capture
developers to devise methods of decoding the packaged support.
raw data. This process can take time and resources. You
may have the experience of buying a new camera when
it first goes on sale and discover that your photo soft- JPEG
ware cannot open the files until an update is released. To The JPEG is undoubtedly the most ubiquitous image file
complicate matters, some manufacturers started offering format. It is a name known by even the most casual com-
raw format variants like CRAW, mRaw and sRaw that aim to puter users due to the highly image-dependent nature of
be a middle-ground between quality and file size. These the internet and visual media at large. Its name comes from
variants are compressed using lossy compression, reduced the acronym describing the standards body that created it:
resolution or both while maintaining characteristics of true the Joint Photographic Experts Group. JPEG describes both
raw files like higher bit depths and non-destructive editing the lossy compression standard and its file extension. JPEGs
support. support 8-bit image channel data and embedded color
profiles. They offer a practical balance of image quality
and file size, functioning as an end-product asset: great
DNG for emailing to a client or embedding in a website. JPEG
The digital negative or DNG is a raw image file format devel- compression offers compression levels of up to 1:100.
oped by Adobe, first released in 2004. The DNG is an open Variations of the JPEG including JPEG-LS (featuring lossless
source standard that can be used free of charge by camera compression) and JPEG2000 exist with some software sup-
companies. It supports lossless and lossy compression port, though the baseline JPEG remains the most widely
options. Though it is not universally used by cameras, many adopted and used.
proprietary raw formats can be converted to DNG using Photographers don’t use JPEGs exclusively because
Adobe software. Doing so offers some reduction to file size they come with a major caveat. In order for a camera to
and uniformity for photographers that work with more output JPEGs, it captures the raw sensor data, converts it
than one type of camera (and thus face juggling multiple to a digital signal and sends it through image processing
raw formats). The U.S. Library of Congress-funded Digital algorithms. Decisions about white balance, sharpness,
Photography Best Practices and Workflow project states that saturation and contrast are made before a JPEG file is gen-
“DNG files have proven to be significantly more useful erated. When this produces results that we like, there’s no
than the proprietary raw files in our workflow.”1 There is problem. It’s when we think that we can do better than the
no future-proof format for photographs; however, DNG onboard processing, or simply wish to have the flexibility
continues to be in the running for long-term support and to try multiple approaches at interpreting the data, that
adoption. the utility of the JPEG format falls short. It’s the difference
A DNG contains all of the same image data as a raw file between buying flour, eggs and sugar at the grocery store
from a particular camera manufacturer. The JPEG preview and buying a cake at the bakery.
image stored within is updated whenever the file is used JPEG files store image pixel data using lossy compres-
in editing software to reflect edits made by the photogra- sion techniques, helping to keep file sizes relatively small
pher, rather than being a static preview generated by the for fast transmission. They degrade in quality each time
camera. they are re-saved, often referred to as generation loss. This

170
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

• Working for photojournalism institutions that expressly

require straight-from-the-camera JPEGs for publication;
• When output is direct to a web-based platform and
in-camera presets are dialed-in to one’s liking.

TIFF
The Tagged Image File Format or TIFF is a popular and versa-
tile format created by the Aldus Corporation in 1986. Aldus
merged with Adobe in 1994, who now owns the copyright
to the TIFF format. It was originally designed for storing
black and white image data from scanners; subsequent
revisions of the format added support for color, additional
bit depth and compression.
TIFF is a bitmap image file format that supports CMYK,
RGB, Lab, indexed color and grayscale; 8, 16 or 32 bits per
channel are supported bit depths. The maximum size for
a TIFF file is 4 GB. The format provides support for tags,
multiple image layers, transparency and alpha channels.
These additional data components are compatible with
Adobe Photoshop layers, making them a convenient
choice for working copies that retain a degree of adjustabil-
ity to post-processing work. Layers in TIFFs appear flattened
in non-Adobe software. The TIFF format supports both
Figure 8-1 Noticeable JPEG artifacts due to lossy compression
algorithms.
lossy and lossless compression using JPEG and Lempel–
Ziv–Welch (LZW) algorithms respectively. A JPEG and TIFF
of the same photograph are visually indistinguishable
assuming both are compressed by the same amount, so
makes them great as end-of-the-line deliverables but poor the benefits of choosing one over the other are dictated by
as working assets. You don’t want to save a JPEG, go back how you intend to use the files. A TIFF is the better option
for additional edits, save it again and repeat. JPEG compres- for additional editing or print output at the expense of
sion artifacts can be distracting and undesirable (see Figure having to store, transfer or share a larger file.
8-1). The majority of image editing software allows for a TIFF’s extensibility and wide support across image
sliding scale of JPEG compression, sometimes labeled as a editing software make it an effective choice for scientific
“quality setting,” when saving a JPEG. The less aggressive applications. The Geostationary Earth Orbit Tagged Image
the compression, the better the quality of the resulting File Format, or GeoTIFF and BigTIFF, are specialized vari-
image. The downside is the lessened capacity for reducing ants for fields with specific needs not addressed by the
file size. traditional TIFF. GeoTIFF supports georeferenced and
There are logical scenarios for shooting/capturing in geocoding information for use in Geographic Information
JPEG over raw: Systems (GIS). BigTIFF supports 64-bit and image sizes up
to 18,000 petabytes (PB). It is used by GIS and medical fields
• Event photography that only requires basic image where scanners and cameras produce incredibly large
adjustments; amounts of data.

171
 s ecti o n 2: P roc ess

BMP PNG
The BMP format is also referred to as bitmap image file The Portable Network Graphics (PNG) file is a raster graphics
or a device independent bitmap (DIB) and uses the file file format created as a replacement for GIF. In some ways,
extensions .BMP or .DIB, respectively. The BMP is a raster the PNG offers a middle ground between GIF and JPEG. It
graphic format that stores a two-dimensional image. A BMP offers lossless compression but does not support anima-
supports bit depths ranging from 1 to 32 bits per channel, tion. PNG-8 can only store 256 colors in its palette similar to
alpha channels and color profiles. It can be uncompressed the GIF limitation. The PNG-24 variant allows for true color
or compressed, however a compressed BMP is often much images with support for transparency. PNGs are used by
larger than an equivalent JPEG. The BMP format is not com- both MacOS and Ubuntu operating systems as the default
monly used by photographers. file type for screenshots and are widely used for web
design elements. The PNG format is not typically used for
photographic images.
GIF
The Graphics Interchange Format, or GIF, is a variant of the
bitmap. GIFs use LZW compression. Originally designed HEIF
in 1987 for transmitting image content over limited The High Efficiency Image File Format (HEIF) is a newcomer
bandwidth internet connections, the format has a palette in a long line of attempts to dethrone the JPEG. HEIF was
restricted to 256 colors (or 8-bits in a single channel). developed by the Moving Picture Experts Group (MPEG)
This makes it a poor choice for full color photographs with its specification finalized in 2015. HEIF is based on the
but a viable option for some graphics or illustrations High Efficiency Video Coding (HEVC) video compression
on the web. The 256 colors are variable depending on standard. The storage structure is based on the widely used
image content, meaning one GIF might contain 256 ISO Base Media File Format (ISOBMFF).
unique colors from another GIF, but the total range HEIF is better described as a container for both image
within a single file remains limited. This limitation is what and audio data instead of a monolithic file format. One
helps keep file sizes to a minimum. The GIF’s utility as a container can contain a single image or a series of images.
photorealistic, small file format is diminished given the The format borrows tricks from video formats like inter-
prevalence of high bandwidth internet connections frame compression to efficiently store similar images
today. The format persists because of its support for from burst-fire or exposure-bracketed sequences. Image
multi-frame animations—a trick that JPEG can’t offer. The sequences saved in this format are, on average, half the
animation capabilities of GIFs are crude in terms of frame size of JPEGs with equivalent or better quality.2 This type
rate and yet social media and messaging trends have kept of storage efficiency is increasingly critical as the world
the format relevant. moves to storing and sharing visual media on mobile
The desire for an image format that supports animations devices with limited memory and transmission band-
or short, looping videos (sans audio) spawned alternatives width. The format also supports a strategy for dividing a
to GIFs that forgo the limited color palette restriction. There large image into tiles that are compressed and accessed
is no unified standard; the popular options are typically separately, allowing for high performance viewing of
built on an HTML5 video format. The best image format for high-resolution panoramic photos, for example. HEIF addi-
video-like content is a video format. GIPHY and GIFV are tionally supports storage of auxiliary data like audio, text,
examples of HTML5-based GIF alternatives that are reada- metadata and depth or disparity information for 3D imag-
ble by modern desktop and mobile operating system (OS) ing systems. It can potentially replace GIF for animations
web browsers. without quality loss and supports image transparency.
HEIF is an intriguing media format because it accommo-
dates many of the visual media products generated by

172
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

smartphones and offers extensibility for future imaging allows for non-destructive editing even across multiple
applications. sessions where the file is saved and reopened, avoiding
HEIF also mimics the raw file model of storing the permanent image quality loss. Adobe Photoshop provides
original image and all subsequent edits as non-destruc- the option to flatten all of the layers in order to save the
tive instructions. HEIF, like TIFF, supports either lossy or image as JPEG or TIFF as needed. A PSD can be as large as
lossless compression and bit depths up to 16-bits. This 30,000x30,000 pixels, 16-bits and 2 GB in file size. Image
allows for storage of images from 10-bit sensors with- documents larger than this can be saved in the Photoshop
out image degradation. The file extension used by HEIF Large Document or Photoshop Big format (PSB). This format
images is .HEIC. extends the potential image size to 300,000x300,000 pixels
The HEIF format is not yet the de facto image file at the with supported file sizes up to 4 exabytes (4 billion GB).
time of writing. Apple, Inc. adopted HEIF for image media There are several other software packages such as Corel-
storage on iOS in 2017. Even Apple’s decision to embrace Draw, Paintshop Pro and GIMP that read Adobe Photoshop
the format, however, is constrained by the reality that formats.
sharing with other devices and platforms offers uncertain
compatibility. HEIFs are converted to JPEG for exporting
or sharing with others. Popular web browsers do not yet Why Raw Works for Digital Asset
natively display HEIF media. Management

Now that you’re familiar with the options, why does the
Proprietary Editing Software Formats raw format come out on top for photographers? JPEGs
We accept that cameras are likely to generate their own offer excellent, expedient deliverables that don’t put heavy
flavor of raw file at capture. We also have to consider stress on memory cards or bandwidth. Most SLRs allow us
proprietary formats that specific software applications gen- to adjust contrast, saturation and sharpening preferences
erate as by-products of editing and output. These formats to be applied immediately after an exposure to saved
can be problematic because of their dependency on the JPEGs. Still, there are many advantages to the raw format
software that creates them. This sometimes translates to for a complete photographic workflow.
software-subscription-ransoms (i.e. if you stop paying for Setting a camera to record both JPEG and raw, we
access, you can’t work with your images) or a greater threat captured a properly exposed test target image. Figure 8-2
of obsolescence (the software company stops supporting (overleaf) shows the output with no adjustments made
their product or its development ceases at a certain point to the raw version. The raw image is less sharp and looks
for an operating system). On the other hand, these formats low in contrast compared to the out-of-camera JPEG. The
allow software developers to build in all sorts of instruc- target looks better and ready for prime time in the JPEG
tions and assets used by their program to accompany the image because it has the contrast and sharpness we’d
traditional image data. Proprietary editing file formats are a expect of a final image. However, not all images should be
necessary evil unless the industry rallies behind a universal processed the same and any other scene is likely to have
format; the best we can do is mitigate the risk and store characteristics that we want to fine-tune and adjust using
TIFF formats for long-term safe-keeping. the raw image data.
A common, proprietary format is the Photoshop Doc- A raw image lacks any processing for sharpness, con-
ument (PSD) generated and used by Adobe Photoshop. trast or color and contains the maximum dynamic range
The PSD takes full advantage of Adobe Photoshop’s image as dictated by the sensor’s capabilities. A JPEG is an 8-bit
editing capabilities, particularly those that are modifiable, image, storing 256 possible brightness levels per chan-
including image layers, layer masks, adjustment layers, nel. A raw image holds 12–14 bits of brightness data. This
vector and alpha channels. Editing with these features additional headroom means that some overexposed or

173
 s ecti o n 2: P roc ess

Figure 8-2 The unedited raw (left) and JPEG (right) images straight from the camera.

Figure 8-3 Highlight recovery of slightly overexposed image captured as a raw file (left) and as a JPEG (right). Only the raw file’s bit depth
retained enough brightness headroom to recover detail.

174
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

underexposed raw captures can be largely recovered with 8. Archiving all assets to long-term storage solutions,
post-processing as demonstrated in Figure 8-3. A JPEG of offline and remote
comparably poor exposure yields exactly the detail it shows
straight off of the camera; software exposure recovery
cannot bring back tone information that failed to write to Image Ingestion
the file format’s limited bit depth. Raw captures are ideal The act of transferring images from a memory card to a
for a photographic workflow because they are the least desktop storage drive can be a simple copy-and-paste
compromised assets we have. They do not degrade over operation. We perform this action all the time with text or
time or over repeated editing sessions and always allow us with miscellaneous files on our computing devices. Image
to return to their just-after-capture state. ingestion goes a step further in thoughtfully organizing and
If a raw file’s captured image data is the star of the reviewing the data. Basic copy-and-paste operations are
show, best supporting role goes to the non-destruc- heavily dependent on manual handling and lack a robust
tive image editing instructions. This data is sometimes strategy for long-term image maintenance. Let’s first review
stored in the raw file container or it may be written to a the broad strokes and dive into the details in the sections
separate, companion asset stored in the same location that follow.
called a sidecar file. This sidecar data comes in the form of We start by establishing a high-level directory for all
an Extensible Metadata Platform (XMP) file. XMP is an ISO image assets. All ingested files will be sorted into sub-
standard (16684-1:2012) created by Adobe in an effort to directories therein. A generic approach from here is to
uphold compatibility and interpretability of metadata while organize subdirectories by year and then calendar date
allowing for additional fields or pieces of information to be (YYYYMMDD format). Organizing files by capture time
added. Image editing software uses XMP-stored editing makes it easy to track down work through invoices, email
instructions to create an interpreted version of the raw correspondence or your memory. This approach lacks
image data, described later in this chapter. content visibility—just because you’ve got photos from
January 9, 2005, it doesn’t mean you know what happened
that day—we rely on catalog software to assist with that.
The Photographic Workflow Some photographers may prefer to use a single, temporary
dropbox directory for all incoming media that gets sorted
One’s workflow is the system of managing digital assets by software in a later step. Alternatively, a more descriptive
from capture to output. A generic workflow following folder naming system might work better for your profes-
image capture consists of: sional workflow. Consistency and long-term flexibility are
key considerations for your approach.
1. Copying the files to a primary workstation computer Downloading files from a memory card can be accom-
2. Ingesting them into cataloging software (can happen plished with drag-and-drop copying in an operating
concurrently with the previous step) system’s file browser. This approach may not catch prob-
3. Making initial selections, rating the stand-out frames lems like corruption or failed transfers. Instead, software like
from a shooting session ImageIngester Pro or Adobe Photoshop Lightroom handles
4. Making basic adjustments on those selections the transfer process with greater assurance of proper
5. Advanced image processing, compositing and cleanup data handling and consistent organization. They can also
on a subset of selections be configured to convert proprietary raw files to DNGs
6. Outputting edited, derivative files for print or digital or create backup copies to secondary locations. Using a
distribution dedicated memory card reader is recommended for image
7. Backing up all ingested and derived assets to local transfer. Connecting the camera via cable invites the possi-
storage bility of file damage if the camera battery dies. Card readers

175
 s ecti o n 2: P roc ess

offer potentially greater transfer speeds because they skip the scope of the impending image processing workload;
the middleman of the camera’s firmware. It’s also more making granular edits on half a dozen frames from a pho-
likely that you have a card reader and cable supporting the toshoot containing a few hundred is more realistic than
fastest available connection protocol. individually editing all of them.
Renaming files comes next. This is best done as a batch
operation, as no one wants to manually name files one at a
time. Basic metadata tagging is helpful to conduct as files Naming Files
are named and stored. This includes the photographer’s Developing a file naming convention is critical for keeping
contact information and usage rights. Descriptors for the track of images in a file system structure. Cameras use
entire image set, such as client name, location and a few default base filenames like “DSC” or “IMG” with frame num-
basic subject keywords, are valuable to introduce here. bers at the end (i.e. IMG_0001.cr2). After 9999 captures, the
We’ll detail types of metadata later in this chapter and counter rolls over to 0001. Alternatively, it’s common for
recommend the Digital Photography Best Practices and photographers to shoot with multiple cameras at the same
Workflow project as an additional resource.3 time, often from the same manufacturer. These scenarios
Additional processing instructions can be applied as easily lead to duplicate filenames unless actions are taken
images are ingested into cataloging software. General to make them unique. Even if you are careful to avoid put-
editing decisions can be made to all ingested files if it ting identically named files into the same directory, leaving
speeds up the process of preview, rating and selection. It’s the default filenames is dangerous when sharing images
not meant to replace the forthcoming step of hands-on with similarly lazy photographers.
editing decisions. If you conduct a photoshoot for a client File systems permit filenames as large as 255 characters.
that intends to print the final images in black and white, it’s Any file naming convention should use numbers, letters
helpful to set all ingested files to a basic black and white and hyphens or underscores (no other symbols or non-
adjustment. This offers quick visual feedback for an entire Latin characters). Using non-standard symbols or characters
image set before committing to specific selections and invites incompatibility from operating systems and edit-
edits. ing software. For example, the “/” indicates a directory
Once files are transferred, renamed and injected with an separation in many operating systems. The three letter
initial round of metadata, it’s wise to back it all up. We often file extension (i.e. “.JPG”) must be included at the end of
think that backing up files comes way later in the process, filenames.
yet immediately following ingestion is one of the best It’s recommended that file naming templates include
times to do it. Backups of captured images lessen the risk your initials or last name, a calendar date and unique
of losing irreplaceable or unrepeatable photo shoots; we’d numbers (four digits long) that restart for every directory
rather lose a few hours of editing work than lose the sole set. You may also wish to use an additional, unique number
set of raw files from a one-time shoot. Generating a primary identifier for invoices or project codes appended to calen-
and backup copy of images downloaded from a memory dar dates. In coordination with your subfolder organization,
card allows you to erase the card or otherwise write over its dates should follow the YYYYMMDD format. Craft a custom
stored files on future shoots without hesitation. naming template with these ideas in mind. As with every
Culling and rating images is a helpful follow-up step. step of a DAM workflow, the file naming system must
We delete very poor-quality images that are unlikely to be be applied with consistency every time you ingest new
of any use. This includes extremely blurry images, missed work. It’s tempting to make directory names descriptive
exposures or catching a model with their eyes closed. We and specific to a job or location. However, your memory
then use a method of flagging or indicating value (a star is not nearly as capable as a cataloging program is when
rating, a color code, etc.) as we quickly review all captures. it’s tapped into the numerous data attributes that describe
An initial round of rating and filtering can narrow down every image file.

176
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

Figure 8-4 Example directory and file naming structure.

The last item to consider is appending filenames with A catalog provides the ability to organize, label and
a consistent vocabulary for differentiating between the locate every image. If image directories are like chapters
original and derivative versions. For example, an RGB in a book, the catalog acts as the index. It documents the
image converted to CMYK for printing could add the string locations of all ingested media files and their metadata
“_cmyk” to the end of the filename (see Figure 8-4). We did such that your entire photography library can be searched
this for every image printed in this book. by many dozens of attributes. It can be an overwhelming
task when first establishing such an important organiza-
tional backbone but it gets easier to maintain once it is
Cataloging Software incorporated into your workflow. The cataloging software
Software packages such as Adobe Photoshop Lightroom, can take the reins on image ingestion to ensure auto-
Capture One Pro, ImageIngester Pro and Canto Cumulus mated, consistent treatment as new content comes in.
are professional go-to’s for cataloging and editing images. The catalog contains supporting information like the star
These are designed as workflow streamliners and catalog ratings or color coding mentioned earlier. It also includes
builders. What is an image catalog and why is it important tools for descriptive keyword tagging that become
to diligently maintain one? searchable criteria for future referencing. Keywords and
Let’s digress for a moment: I had stacks of negatives ratings are leveraged by filter queries to find exactly what
when I was a photography student in a previous but inten- you need, even if the thing you need is not reinforced by
tionally unspecified decade. I placed each developed film your file directories or naming conventions. For example:
roll in a negative holder, carefully labelled with their subject the highest rated images of Mary Smith, a regular model
and capture date and created a contact sheet. It was of yours, taken over the last five years is easily found in an
relatively easy to find any negative I wanted later by using image catalog system.
this system. Digital images changed everything; instead of Cataloging software typically requires a single, dedi-
36 exposures per roll of film, we capture as many images cated catalog file that describes and documents access
as our memory card can hold during a given photo shoot. to the image data. The catalog assets store critical docu-
Sorting, labeling and filling every single frame by hand in mentation for your library’s contents and where to find
this new image creation paradigm would take so much them across any number of directories and storage drives.
time as to eliminate all the time we otherwise enjoy going If the folders or files stored within are moved or deleted,
out and taking the photographs. the catalog software loses its link to the data. If the drives

177
 s ecti o n 2: P roc ess

are offline, the software continues to index their existence. you insist on using pixel editing, archive the original image
Conveniently, cataloging software creates previews upon files and work exclusively with derivative, working copy
ingesting the full raw images. This allows for viewing a versions. This way, though your editing is irreversible from
version of the image even if the full resolution file is on an a given editing session, you can always fall back to the
unavailable or disconnected drive. This is very different unedited original capture and start over. Adobe Photo-
from accessing files via the operating system’s file browser, shop is a pixel editing software, though advanced users
which can only present data it sees connected in that leverage layers and other special functionality to allow for
moment. much more malleable and reversible edits in a working
The image catalog is vital to asset organization and it document.
should be backed up just as often as your images. We have Parametric image editing does not alter the original
witnessed students losing their catalogs on too many occa- data. Every edit made to the photograph, from cropping
sions to shake the paranoia; it’s not a pleasant experience to exposure adjustments and color balance, is recorded as
to start the whole process over from scratch. an individual instruction or parameter value in metadata.
Trends in cataloging and editing software point to a The software knows how to adjust contrast, for example,
move to the cloud. This provides businesses, large or small, it just needs to know by how much. We edit by specifying
to easily share data across the internet and synchronize new values for different image adjustments like satu-
all ingested and edited content. In addition, software as a ration, sharpness and brightness. In addition to global
service (SaaS) systems provide editing software as well as changes, advanced instructions can specify pixel locations
DAM options, allowing employees access to assets from for local adjustments to portions of an image. Software
anywhere including mobile platforms. generates a temporary preview of those adjustments as
we edit yet the raw data is preserved. In the future when
software and our editing skills improve (or our aesthetic
Image Editing tastes shift), the changes can be undone and new editing
It is seldom that a photograph goes completely unedited decisions applied to the original image data. Parametric
between capture and output. Whether it is removing a image editing software writes these instructions to its
small piece of dust or completely altering the look and own documentation or to XMP sidecar files written along-
feel, image editing is an inevitable stage in a photograph’s side the images. We primarily use this technique with
development. There are two overarching but fundamen- raw images yet JPEG and TIFF files can be parametrically
tally different approaches to image edits: pixel editing and edited since the instructions are written independently
the previously mentioned parametric image editing. from the source file.
Pixel editing modifies the original pixel values in an A great capability of parametric editing is in storing
image. The edits are permanent once the file is saved. multiple, unique editing sets for a single raw image. Similar
This is often referred to as “baked in” edits. If you save to how every grandmother has her own apple pie recipe
the edited image using the original filename, effectively despite the same standard starting ingredients, a photo
writing over the data, the unedited version cannot be can have unique parametric editing instruction sets applied
retrieved. Consider the following scenario using a pixel to create different interpretations (see Figure 8-5). One
editing approach. You make an exposure adjustment that simple use case: keeping both a color and black and white
darkens all pixels. You save the image, figuring that the rendering of a photograph. Instead of duplicating the
change was for the better and worth preserving. Upon image and doubling the storage burden, parametric image
printing, you notice that the shadows are blocked up and editing software simply writes two recipes and keeps them
detail-less where there was previously tone variation and with the raw file. It’s only when we export a version that the
texture. The exposure adjustment made before printing software generates a new, derivative and “baked in” image
cannot be undone—the shadow detail is lost forever. If with the editing instructions applied.

178
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

Metadata and Keywords

Metadata is simply data about other data. Photography
metadata is information about your images. It can be
stored in the image itself and can therefore be available
to you or another user whenever the image is available.
It may also reside in a sidecar file. Image metadata can be
organized into four categories: technical, administrative,
descriptive and processing.
Technical metadata is the capture information and
gets stored in the Exchangeable Image File Format (EXIF)
portion of an image file. While raw formats lack standard-
ization, EXIF metadata is consistently used by raw files to
record details such as the date and time the photograph
was captured, GPS coordinates, shutter speed, aperture
setting, ISO, if the flash fired, the camera make, model,
serial number and the lens used. The bulk of this is gen-
erated automatically and we leverage it for cataloging,
filtering and documentation purposes. We typically don’t
need or have a reason to modify EXIF metadata. EXIF
information is also supported and stored by the JPEG
format.
Administrative metadata includes International Press
Telecommunications Council (IPTC) fields. This set of fields
is typically entered after capture with desktop software,
though some cameras can be configured to automatically
add it to every capture. Here is where contact information
is entered, such as name, address, phone number and
email address. It’s a good idea to populate the “Usage”
IPTC field with a statement such as “No usage without
express and prior written permission. All rights reserved.”
This tells others that they don’t have free reign to use the
photos. IPTC metadata also includes fields for the year
of copyright and the copyright holder’s name. However,
Figure 8-5 Three unique edits of the same raw file. The raw image simply having something stored in this part of an image’s
is stored once with three unique instruction sets when working in
parametric image editing software. It’s only upon export that new, full
metadata does not make the copyright official. For that,
resolution image files are generated. you must register the work with the Library of Congress.
Administrative metadata is easily added en masse to files
as they are ingested into a catalog since the information is
applicable to entire photo shoots or events.
Descriptive metadata is added by hand after ingest and
it’s where we add image content descriptions. This can
include the title or caption, the concept, the location (an
actual name rather than GPS coordinates) and keywords.

179
 s ecti o n 2: P roc ess

Keywords are specific words that help to locate photo- Image Data and File Sizes
graphs within your library or when included in any sort
of large, searchable image database. Example keywords The pixel is the smallest and most basic visual building
include a client’s name or a specific product shoot. Soft- block of a photograph. A pixel is a type of digital informa-
ware tools exist to aid in crafting keywords for optimizing tion with its own building blocks: binary digits (ones and
search engine results. Google’s Adwords is one example of zeroes). All digital data, from photographs to music to text,
a keyword optimization tool. Rating systems, like stars or boils down to binary information read and stored by a
thumbs up/down indicators, are also descriptive metadata computer. The amount of binary digits needed to describe
concepts that function like circling frames on a contact even a single pixel of a photograph is so large that it’s not
sheet as used to be done with darkroom development. practical to directly count them. We use units of higher
They are useful when tagged and reviewed by a single orders instead. For every eight binary digits (or bits) there
user, perhaps less so when attempting to communicate is one byte. The conversions between units to quantify file
nuanced opinions or sorting ideas with others. size are shown in Table 8-2.
Descriptive metadata is the most granular and
image-specific form of metadata. Descriptive keywords Table 8-2 File size unit conversions.
can be generic terms like “landscape” or as specific as is Unit Symbol Number of Bytes Binary
needed, such as “Mary Smith” to tag a model. Parent and
Byte B 1 2
child keywords describe hierarchical keyword groupings
Kilobyte KB 1,024 210
where “Rochester, NY” is a child keyword for “New York”
Megabyte MB 1,048,576 220
which itself sits nested under the parent keyword of
“Places.” Sometimes it makes sense to use hyper-specific Gigabyte GB 1,073,741,824 230

child keywords while other times it’s reasonable to stick Terabyte TB 1,099,511,627,776 240
to higher-level ones. Ideally, descriptive keywords should Petabyte PB 1,125,899,906,842,624 250
make thoughtful use of a controlled vocabulary such that
you consistently use “architecture” instead of haphaz-
ardly using “buildings” one day or “manmade structures” The practical units relevant to photographers are
another.4 Machine learning is increasingly leveraged by kilobytes (KB), megabytes (MB), gigabytes (GB) and terabytes
digital photo album software and social media networks to (TB). Sidecar XMP files and small JPEGs are a few dozen to
determine descriptive keywords (including facial recogni- a few hundred kilobytes a piece. Single raw image files are
tion) without human input. tens of megabytes. Collections of image assets like those
Finally, processing metadata is parametric image edit from a single photo shoot can be multiple gigabytes. A
information. This information can be things like instructions photographer’s catalog of photoshoots, composite files
for cropping, color correction or contrast adjustments. The and exported JPEGs or TIFFs tends to be on the scale of
specific controls and values for such editing adjustments terabytes (and this is where the panic sets in as we price
may be specific to a particular platform or software; one out storage solutions). The more photographic informa-
program might understand “exposure: +1.0” while another tion we want—more layers, channels, reproducible tones,
requires “exp val 15.” Such variation is supported because versions—the more the files grow in size. Digital image file
XMP is an extensible metadata structure. sizes are calculated with the following steps:

1. Calculate the total number of pixels.

Total pixels = number of vertical pixels x

number of horizontal pixels

180
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

2. Determine the number of bits needed. generates three channels (R, G, B) of bit depth information.
Consider the same photograph after a raw engine demo-
Total number of bits = total pixels x bit depth saiced it and saved a full color, 16-bit TIFF file, calculated in
Equation 8.3:
3. Divide the total number of bits by 8 to determine the
number of bytes. ((8,000 x 5,320) x (16x3))
= 255,360,000 bytes (Eq. 8.3)
8
total number of bits or 243.53 megabytes
File size (in bytes) = (1.7)
8 bits/byte
The TIFF version of the image is much larger now that
4. Divide the file size in bytes by 1,024 to determine the file there are three color channels, each of which contains up
size in kilobytes. to 16 bits of information. One takeaway from Equation
8.3 is that storing raw files over unedited TIFFs is noticea-
File size (in bytes)
File size (in kilobytes) = (1.7) bly more space-efficient despite the comparable lossless
1,024 bytes/kilobyte
characteristic. The catch is that the raw data requires the
5. Divide the file size (in kilobytes) by 1,024 to determine appropriate decoding software (the raw engine) to turn it
the file size in megabytes. into a workable photograph. Conversely, the file size of an
8-bit JPEG derived from the raw file is calculated in Equa-
File size (in kilobytes)
File size (in megabytes) = tion 8.4:
1,024 kilobytes/megabytes

These steps are summarized into Equation 8.1: ((8,000 x 5,320) x 8)

= 42,560,000 bytes (Eq. 8.4)
8
or 40.59 megabytes
[(Total number of pixels) x (Bit depth)]
8
2,048
File size (MB) = (Eq. 8.1) It’s even smaller in practice because JPEG introduces
2,048 size reduction via lossy compression not accounted for in
our calculation. Estimating a file size with this consideration
This formula is true for the raw data in an image file and is more difficult due to the content-dependent nature of
does not consider compression applied or any embedded JPEG compression efficiency.
metadata. The greater the number of pixels and/or the
greater the bit depth, the larger the resulting file size. We
estimate the size of a raw file from a 42.5MP sensor using File Storage Options
the number of pixels (8,000x5,320) and the sensor bit
depth (14-bit) following the previous steps, as summarized We reviewed the options for camera capture storage in
in Equation 8.2: Chapter 3. Never rely on files stored on capture media for
anything other than short-term storage between capture
((8,000 x 5,320) x 14) and ingestion. We strongly recommend formatting camera
= 74,480,000 bytes (Eq. 8.2)
8 storage cards through the camera’s firmware before every
= 71.03 megabytes
new shooting session. Memory cards write to a File Alloca-
The raw file may be slightly larger than 71 MB once tion Table (FAT) that acts like an address book for where the
additional pieces of information are added alongside the data resides on the card. Selecting “delete” in a camera’s
image data like proprietary processing metadata. Note firmware menu typically deletes this table without deleting
that the raw image has a bit depth of 14 as it is brightness the image data itself. The image data is only truly deleted
information for every photosite location. It is not yet a full when the card runs out of space and writes over it with
color image and undergoes a process of interpretation that new data or when you format its contents. This is good

181
 s ecti o n 2: P roc ess

news for recovering accidentally deleted data—there’s a convenience. Storage capacities of optical discs are listed in
chance it’s on there—but bad news for the stability and Table 8-3.
health of the capture storage medium. Errors or corruption We rarely use optical discs for listening to music or
can occur when data is written over other data. transporting assets between computers anymore. Com-
Once you’re safely back in the studio or starving artist puter hardware, particularly laptops, increasingly forgoes
studio apartment, it’s time to offload all those photos. This optical disc drives altogether and thus requires an external
means transferring the images from memory cards to your peripheral to read and write to disc. Still, the technology
laptop or desktop computer. How safe is that data once it offers a cheap backup storage solution for media files. Discs
resides on your laptop hard drive, the laptop that has taken can be scratched and damaged when handled, so cases or
more than a few spills off of coffee shop tables and couch disc sleeves are necessary for anything more than short-
cushions? When your digital assets have high monetary term transfers. Scratched discs mean an impaired ability to
and artistic value, you must get serious about storing them. read the data previously written to them. CDs have an esti-
The main options are optical disks, disk drives and offsite mated lifetime of approximately 100 years under moderate
cloud storage. storage conditions.5

Optical Disc Media Hard Disk Drives

Optical disc media storage includes Compact Discs (CDs), A hard disk drive (HDD) stores information on a rotating,
Digital Versatile Discs (DVDs) and Blu-ray Discs (BDs). They magnetic platter. A mechanical read/write head sits just
store binary information by laser-cutting microscopic pits above the platter’s surface and detects changes in mag-
into the surface of an alloy material. The information is netic polarity. Though visually reminiscent of a needle and
written in a circular, spiral pattern from the center of the record on a turntable, these components do not touch. The
disc outwards. An optical disc drive laser reads these pits faster the platter spins, the faster information stored on the
and lands (the raised or level areas) as the disc spins at high platter can be read and written to. Blunt force, especially
speeds. CDs are an earlier generation optical disc format, when the disk is spinning, can irrecoverably damage the
followed by DVD and most recently, Blu-ray. Each has platter. This is a particular risk for laptop hard disk drives
increased storage capacity in turn by finding ways to write and external, portable hard drives. Exposure to strong earth
smaller on the disc surface or by writing at different depths magnets can also damage the platter.
in the material. The two primary specifications for an HDD are speed
Optical disc media can be write-once-read-many (number of revolutions per minute or RPM) and size, in
(WORM) or rewritable (RW). Rewritable discs made sense gigabytes or terabytes. Common hard disk drive speeds
for transferring files back and forth by reusing a single disc are 3200 RPM (found in some laptops and all-in-one
but flash drives quickly surpassed this option in speed and desktops), 5400 RPM and 7200 RPM (the best option for
desktop computers). An HDD that tops out at 3200 RPM is
reasonable for a web-browsing machine but insufficient
Table 8-3 Optical disc media and their storage capacities. for heavy image or video editing work. Some specialized
Media Type Storage Capacity HDDs used for data servers offer speeds of 15,000 RPM.
HDDs come in standard sizes of 2.5” and 3.5” and top out
Compact Disc (CD) 725 MB
in capacity at 10 TB.
Digital Versatile Disc (DVD) 1.8 GB
Even extremely fast rotation speeds can’t change the
Dual Layer DVD 3.6 GB
physical reality that the read/write head can only be at one
Blu-ray Disc (BD) 25 GB place at a time. Accessing information that resides across
Dual Layer BD 50 GB multiple locations on the platter is time-consuming for the

182
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

drive. This limitation leads to the recommendation that Solid state drives come in standard sizes: 1.8”, 2.5” and
a computer’s operating system and program files should 3.5”. Figure 8-6 shows an SSD alongside an HDD. Smaller
reside on one drive while all working media assets should form factors are used in laptops when space is at a pre-
be stored on a separate drive. mium, though they can be more expensive. Internal SSDs
communicate with the motherboard via Serial Advanced
Technology Attachment (SATA) or mini-SATA (mSATA)
Solid State Drives connections.
Solid state drives (SSDs) are like grown-up USB flash drives or
memory cards; the same microchip technology is used to
store digital data with non-moving parts. This technology Disk Configurations
is called NAND-based flash memory and it is managed by an Data storage can be a nightmare. As your career pro-
onboard processor called a controller. gresses, the number of photographic assets you need
SSDs historically demanded a significant premium in to store grows at an incredible rate. It’s unlikely to shrink,
price-per-megabyte compared to HDDs, though recent save for a major data loss! There are two configuration
years have seen prices drop. Solid state drives range in size approaches for data storage on disk drives.
from 64 GB up to 10 TB (though this upper end capacity The first is JBOD, short for just a bunch of disks. This con-
rises every year). Since all of the data is stored on flash figuration allows you to attach a collection of disk drives
memory, it doesn’t have the physical limitations of an to your computer. A low-tech implementation of JBOD
HDD’s head and platter movement. On the other hand, stops here—if you have spare ports, you can plug in more
flash memory isn’t great at reading and writing many tiny drives—though this is not a sustainable practice. Each disk
files at a time. Additionally, flash memory can degrade. All appears as an independent storage drive to the operating
drives, as a result, are built with more storage than listed system, necessitating manual data management. A fully
and write to new areas as old ones wear out. Shooting realized JBOD system, on the other hand, adds a layer of
video is a great use case for choosing an SSD over an HDD: software that does this management for you. This software
the volume of data and large file sizes combined with the groups the collection of storage disks to appear as a single
need to load and view it all quickly make SSDs the superior drive. Such mediation may solve the question of “which
storage choice. drive did I put that image on?” but it does not provide any
redundancy for your data storage. If a sector or a drive fails,
all is lost. That data exists in a singular location that is sus-
ceptible to hardware failure and user error. JBOD systems
are appealing because they can make use of old or differ-
ently sized hard drives that tend to pile up or collect dust in
a closet. Still, they threaten to monopolize desk space and
are vulnerable to data loss.
The second configuration approach is a redundant array
of independent disks or RAID. A RAID also uses multiple disks
and adds intermediary software to appear as a single drive
to the computer. The major departure from JBOD is in its
ability to add data storage redundancy. In a redundant
configuration, files are written in several different loca-
tions across the array of disks. This way, if there is a sector
or a disk failure, all the data is preserved and recoverable
Figure 8-6 A 3.5” hard disk drive next to a 2.5” solid state drive. from one or more of the healthy siblings. There are seven

183
 s ecti o n 2: P roc ess

standard levels of RAID configuration: RAID 0 through RAID Backup Solutions

6. RAID 0 is essentially a JBOD system that offers no redun- A popular idea in backup solutions is the 3-2-1 rule: three
dancy with unique data existing in a single location on a copies of your data on two different media with one
single disk (confusingly, you can have a RAID configuration stored at an alternate location.6 You want one copy of your
that is simultaneously a JBOD). Data redundancy starts at image data available to work on and a second copy close
RAID level 1, a mirroring of the data in one additional loca- at hand in case you accidently delete a file or encounter
tion. Above level 1, bit- and byte-level striping and parity an unrecoverable file error. The third copy resides at a
are used. These increasingly complex mechanisms offer different physical location—a beach house, perhaps, if
additional data protection. With each ascending RAID level, your photography career is a lucrative one. Keeping a copy
the amount of available storage space decreases relative somewhere else protects the data against natural disasters,
to the total amount of disk storage in the array. A RAID 1 break-ins or other unforeseeable events that can wipe out
configuration, for example, might have two 2 TB drives. all electronics residing at a location (one of the author’s
Only 2 TB is available to storage data so that it all exists on homes was struck by lightning, damaging a computer).
one physical disk drive and can be mirrored (duplicated) on Using two forms of storage media hedges your bets as
the second. every type of media has a different likelihood of failure and
Network-attached storage (NAS) hardware is used by degree of future-compatibility-proofing.
photographers to house and manage RAID-configured Maintaining these copies of your work is a full time job
drives (one of ours is shown in Figure 8-7). NAS units without the help of backup software solutions. Two such
reside on a desktop or they can be connected to the local solutions are ChronoSync or Carbon Copy Cloner, though
area network; they function as file storage servers. Net- there are many options.7,8 Operating systems and photo
work-attached storage solutions range in size based on cataloging software also offer a degree of automated
the number of disks they hold. They often include auto- backup behaviors to make the process painless. Cataloging
mated backup and drive health monitoring software that is software may offer the option to immediately make dupli-
accessed through a browser-based interface on a desktop cates and store them on a separate drive upon ingestion.
computer. RAID-configured storage solutions like NAS are a Otherwise, scheduled automatic backups of entire drives
valuable DAM investment. They can stick with you for years makes backup plans something to establish once. After
by periodically upgrading the disk drives to larger capaci- that, the systems should remain in place to regularly write
ties as your asset library grows. new data to all secondary storage drives.
The last piece of the puzzle is knowing the difference
between a backup and an archive. A backup is meant for
assets that you are currently working on. The backup is
what you turn to if you accidently delete a file or the file
becomes corrupted. The backup must be easily accessible
so that valuable work time is not lost. An archive means
storing files offline for long-term safe-keeping. These are
files that are no longer changing, like in the case of finished
client assignments, and it’s unlikely that you need to access
them again in the near future. The files are accessible with
some effort and transferred to a convenient, local working
environment if they’re needed in the future.
There’s no such thing as a truly archival format or strat-
Figure 8-7 A network-attached storage solution with multiple egy for digital media. Generations of societies, archivists
drive bays. and librarians have found ways to maintain and preserve

184
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

all sorts of physical media: paintings, sculpture, text. The either controlled access or access to anyone on the web.
medium of photography (starting with analog processes), File hosting can be read-only specified if needed. Many
in the context of humanity’s timeline, is brand new. Digital cloud-based storage sites also have file hosting capabilities.
technology is an even more nascent phenomenon whose
long-term storage and archiving solutions are in the
earliest stages of development. You may not be thinking Data Recovery
about your photographs as the next Mona Lisa—though You will lose photographs. It happens to everyone. Memory
we believe that photography exists amongst the halls of cards get corrupted and you find yourself desperate to
traditional art—but you would like to believe that work recover photographs you’ve taken but didn’t yet transfer
from the early days of your career survives ten, 20 and even to a computer. A hard disk drive sputters and fails to start
40 years down the line. Archiving of digital media is an up one day. Hope is not lost! There are scenarios where
effort requiring regular tending, moving or updating files the photo data is on the storage device, inaccessible but
to new storage solutions, formats and computer hardware unharmed. This is where software solutions like Photo
to ensure readability and compatibility. Backup and archive Rescue or Disk Warrior can dig through the rubble of your
solutions must be regularly revisited and maintained. memory card or hard drive disaster. Sometimes it’s the
storage drive’s index that gets garbled, not the data itself,
in which case clever software can recover some or all of
Remote Server File Hosting and a drive’s data. Best practices to minimize the chance of
Storage Services data loss include minimizing movement or transportation,
An increasingly prevalent option for managing digital avoiding extreme temperatures or moisture in storage and
assets is cloud-based storage. Cloud storage and services ensuring good air circulation around disk drives. Memory
put all the cost of hardware purchase and maintenance cards are particularly susceptible to dust and debris on
and computer knowledge for setup into someone else’s their contacts, as they are easily pocketable items and are
hands (the service provider). The user pays to effectively moved between devices on a regular basis.
rent the offsite storage space that is accessible with
an internet connection. The provider is responsible for
maintaining your data, making it available, and providing Avenues for Data Transfer
backup and recovery. Cloud storage can be used as a
primary backup, an archive or as an offsite copy in case Bandwidth describes the maximum data transfer rate of a
of disaster. There is a monthly cost that may outweigh network or connection protocol. It’s commonly measured
the expense of investing in additional storage on-site in a in bits per second (remember: 8-bits is 1 byte of digital
home or studio. It’s also not susceptible to your cat knock- data) and gives us a sense of how much digital informa-
ing the drive array off your desk. The biggest limitation of tion can be moved from one place to another, be it two
cloud-based storage is transfer speed as a function of your computers, two drives within one computer or from a
internet service provider (ISP). Leaving files to upload over- local network to the internet. Data transfer rates fluctuate
night is convenient for backups; downloading dozens of based on a host of variables including the number of files
images you needed 5 minutes ago is not a viable workflow. and signal stability. For example, a large number of very
Cloud-based storage is best leveraged for asynchronous small files takes longer to transfer than one large PSD file
activities. containing the same total number of bits. Additionally, a
File hosting services are a great option for delivering storage medium is usually only accessible one sector or
data to a client. Such services enable you to upload files area at a time. Different connection standards are available
to server-based storage through which clients are granted for data transfer, independent of where it’s headed or what
access to download. There are many options available with the data payload looks like.

185
 s ecti o n 2: P roc ess

Data Transfer Connection Standards Table 8-4 Data connection types and their potential bandwidth for
data transfer.
Here we offer a rundown of available connection stand-
ards. Our intention is to make you familiar with each Data Connection Type Maximum Data Rates9
without dipping into the electrical engineering or stand- Gigabit Ethernet (1000BASE-T) 1,000 Mb/s
ards documentation. Just like with every other commercial
10-Gigabit Ethernet (10GBASE-T) 10 Gb/s
technology, standards and protocols ensure compatibil-
WiFi IEEE 802.11b/g 11 Mb/s, 54 Mb/s
ity and versatility in our image-making toolbox; cables
WiFi IEEE 802.11n 600 Mb/s
and connection ports are a tangled reality of our digital
ecosystems. Connection standards vary in their potential Bluetooth 4.0/5.0 25 Mb/s, 50 Mb/s

bandwidth and present potential bottlenecks in data trans- eSATA 3 Gb/s, 6 Gb/s
fer, be it moving raw files from camera to computer or from FireWire 400/800 400 Mb/s, 800 Mb/s
internal storage to an external NAS. Photographers find USB 1.0 12 Mb/s
themselves in variable environments from field shooting,
USB 2.0 480 Mb/s
studio work, editing in transit or remotely reviewing photo
USB 3.0/3.1 5 Gb/s, 10 Gb/s
sets from mobile devices. In short, there isn’t a single,
Thunderbolt 1.0 10 Gb/s (x2 channels)
neatly determined setup for your gear. The best bet is
taking advantage of the best technology available for the Thunderbolt 2.0 20 Gb/s

photographic workflow task at hand. Table 8-4 lists data Thunderbolt 3.0 40 Gb/s
transfer protocols and their bandwidth potential. There are DVI Single/Dual 5 Gb/s, 10 Gb/s
more variants than the ones included here and standards DisplayPort 1.0/1.3 10.8 Gb/s, 32 Gb/s
are regularly updated with backwards compatibility (at the
HDMI 1.0/2.1 5 Gb/s, 48 Gb/s
slower speeds). We’ve put network connections alongside
peripheral connections to appreciate the relative speeds
despite their different usage.
Internal storage drives communicate with other com- or medium format cameras that offer ethernet ports built
puter components using SATA cables. This connection into their bodies. This setup is called shooting tethered
offers high transfer rates. External serial advanced technology because there’s a cabled connection between the camera
attachment, or eSATA, connections use this high-speed bus and a computer workstation. Large raw images can be
in service of connecting drives external to the computer transferred as they’re captured to a storage medium and
motherboard (i.e. an external drive used with a laptop). The into cataloging or tethering software that displays imme-
connection requires a peripheral component interconnect diate previews. Using ethernet in this scenario is the fastest
express, or PCIe, card to make the connection possible way to move the image data off camera. The niche use
though these cards are more common on PC machines case means that ethernet ports are only offered on high-
than Macs and may need to be installed separately. end professional equipment.
Ethernet is used for local area networks (LANs), a WiFi is a wireless network connection protocol that
protocol originally developed by Xerox Corporation transmits information through radio frequencies. Wireless
and standardized in the early 1980s.10 We use ethernet communications are critical for network connectivity with
connections to hardwire our desktop and sometimes our many portable devices and we often expect non-wired
laptop computers to a local network (a router or modem). information transfers. Their ubiquity stems from con-
Ethernet connections are fast and reliable for large data venience, not superior speed. We highlight this to stress
transfers as they are not subject to environmental inter- consideration of the right tools for the job. When a client
ference in the way that wireless protocols are. Studio is waiting on images from your laptop set to go to print
photographers benefit from shooting with high-end SLRs on short notice, the right course of action is plugging into

186
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

an ethernet cable—not transferring over potentially slow compatibility with newer ports and cables. FireWire is a
WiFi. WiFi standards have increased potential bandwidth serial bus interface developed by Apple that provided very
capabilities over time. The IEEE standard for WiFi, 802.11, fast speeds relative to USB in the 1990s and early 2000s.
has a, b, g and n variants. It came in two bandwidth variants: 400 Mb/s and 800
One of the most commonly known and encountered Mb/s (named FireWire 400 and Firewire 800, respectively).
connections is the Universal Serial Bus, or USB, protocol. FireWire is no longer supported and has no direct or back-
USB is used for connecting a wide variety of peripherals to wards compatible descendant.
computers and has been around for decades. Its standard A wireless technology standard designed for transfer-
connector shape (USB Type A or Standard-A) is a small ring data over short distances, Bluetooth, was invented by
rectangle; as devices grow smaller and thinner, compact Jaap Haartsen in 1994 and it uses radio waves to transmit a
variants like miniUSB, microUSB and others have created a signal.11 Its appeal is a low power draw, making consumer
world where our friends never seem to have the exact right products like wireless cellphone earpieces possible despite
connector to charge our phones. USB Type-B, the often the devices’ small batteries. Bluetooth has never been
overlooked connector, is found on printers. The USB 2.0 particularly useful for data transfer of the scale required for
standard was published in 2000 and touts speeds up to 480 photo or video files, but that could change. The proto-
Mb/s. USB 3.0 “Super Speed,” published in 2008, promises col’s increasing bandwidth, reliability and accompanying
up to 5 Gb/s. USB connections can be used to shoot teth- software to pair communication between devices led to
ered to a workstation or laptop but the transfer speeds lag its use in dozens of household electronics like wireless
behind ethernet tethering. As of this writing, USB-C (a form speakers, headphones and tablet computers. Bluetooth is
of USB 3.0) is gaining in popularity for a range of portable categorized by class with available classes of 1–4 in order
devices, particularly mobile phones, laptops and tablets. from greatest power draw to least. At the same time, each
It provides power to directly charge connected devices, successive class decreases the potential signal range of
supports display output signals and features a symmetrical the connection. Class 2 Bluetooth is often found in mobile
connector design. This last point is critical for user sanity: phones, though some have started using Class 4 for the
it eradicates the all-too-familiar frustration of incorrectly lower power consumption benefits.
guessing the orientation of a traditional USB-A cable when Separate yet related connection standards, Digital Visual
attempting to plug it in. Interface (DVI), DisplayPort and High-Definition Multimedia
Thunderbolt is a protocol developed by Intel and Apple Interface (HDMI) are designed for digital video data transfer
that functions similar to eSATA with a PCIe connection, to computer monitors and other external displays. Displays
offering direct and fast access to the motherboard. Thun- are output hardware that require some form of digital
derbolt 1.0 and 2.0 used a Mini DisplayPort connector; video signal; you won’t find HDMI ports on external hard
Thunderbolt 3.0 uses a flavor of USB-C. Thunderbolt offers drives, for example. Displays have seen similar evolution
the ability to daisy-chain devices together, something over the years in increasing throughput needs: a computer
that USB cannot do. Identifying the difference between a sending a 640x480 video stream to a desktop monitor has
Thunderbolt and USB 3.0 port can be difficult since they lower bandwidth requirements than a computer sending
both use the USB Type-C style connector despite internal 4K video with surround sound to an OLED television. The
differences. Thunderbolt is primarily found on Apple hard- former is possible with a DVI connection, the latter an HDMI
ware and external hard drive model variants. 2.1 connection.
We’ll mention USB 1.0 and FireWire if photographers
find themselves inheriting older hardware, though both
standards are superseded by newer and faster ones. USB
1.0 offered slow transfer speeds compared to the latest
versions, of course, though its connector design retains

187
 s ecti o n 2: P roc ess

Computer Workstation Hardware Graphics Processing Unit

The graphics processing unit (GPU) is a processing chip sim-
Despite computers saturating every aspect of our lives, the ilar to the CPU specifically designated for graphics-related
inner-workings and specifications of computer hardware computations like rendering on-screen video, video game
are not necessarily well known to its users. Let’s briefly engines and image processing calculations. Dedicated
describe the major components considered in relation to GPUs and software that directs tasks to the GPU instead
image editing and storage. We’ll limit mentioning specific of the CPU are increasingly common. GPU capabilities are
numbers because they never age well in printed textbooks. often lesser on laptop hardware than desktop hardware
We’d rather not have readers from the distant future laugh- due to size and power requirements. GPUs are particularly
ing at how quaint our computer processing capabilities good at running computations in parallel. Image editing
seem to them. software can see improved performance and speed when
The backbone of every computer workstation is the configured to leverage a GPU.
motherboard, a circuit board that pumps digital informa- Desktop computers support installation of video cards
tion through the many subcomponents that read, process consisting of GPUs with their own, dedicated RAM. This
and store it. It offers connections and ports for expanding memory is loaded with the relevant data (an image layer
input and output capabilities. The motherboard also con- in a Photoshop document, for example) so that it can be
tains the memory slots that provide random access memory passed immediately to the GPU for calculating a trans-
(RAM) that all software needs to efficiently operate. formation or filter on that data. The speed and amount
of onboard RAM of a computer’s GPU are important
specifications for photographers because complex image
Central Processing Unit processing tasks demand significant computational power.
The central processing unit (CPU) is the brain of a computer If you want to open a gigapixel panorama with three dozen
in that it handles the bulk of the computations related to adjustment layers at 100% while expecting buttery-smooth
running the operating system and any software therein. scrolling, you’ll need a dedicated GPU with a lot of RAM to
When buying a computer intended for image and/or video match your performance expectations.
editing, the specifications of its CPU (clock speed, number
of cores) are of primary importance. Each CPU core func-
tions as a unique processing unit, leaving it up to software Random Access Memory
optimization to split up tasks and computations across Random access memory can be thought of like the
available cores. The more cores the better, assuming that human brain’s short-term memory. You keep things
the software is written to take advantage of them. Clock readily available for use at a moment’s notice, but may
speed or clock rate is measured in gigahertz (GHz) and move certain information into long-term storage when
describes the frequency at which the CPU core operates. it’s not needed right away. Your brain might keep names
While a faster clock rate is theoretically better for com- and relationships handy at a social gathering, but not the
puting performance, it’s difficult to pin it on this number particular details of how a car engine is built. RAM is how
exclusively. CPUs generate heat and often require heatsinks the computer keeps track of the different images, the
and fans in their immediate surroundings. This explains different programs and the endless number of browser
why your computer can get either very hot to the touch or tabs you leave open. It’s a form of flash storage housed on
loud during editing sessions when the CPU is taxed more removable circuit board chips connected to the mother-
than casual web browsing usage might. The main CPU board. Just like any other storage medium, an important
manufacturers are Intel and AMD. number to know is how much storage it can handle;
RAM is commonly described as a function of gigabytes.
Unsurprisingly, then, the more RAM a machine has to

188
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

store this information, the less time you spend watching it laptop or desktop computer through a port (USB being the
think and search for what it was doing. RAM is a computer most common). We trade portability and convenience for
component that indirectly impacts performance: adding speed; an internal drive can read and write data faster than
RAM can result in a snappier, more responsive experience on a plugged-in external drive. External drives can be pow-
when editing photos. It’s a relatively inexpensive hard- ered through the data connection or using a second cable
ware upgrade that can keep an aging computer on pace connected to a power source. Because of the increased
with your workflow. power demand of operating at higher speeds, USB-only
RAM is most effective when it is added in matched pairs portable drives tend to have slower read/write speeds
(i.e. installing two sticks of 16 GB RAM for a total of 32 GB). compared to externally powered units. Portable drives
We recommend setting Adobe Photoshop’s preferences allow us to transport images from one location to another
to allocate up to 70% of available RAM to its processes. easily without requiring that we lug an entire computer
When Photoshop uses all of the allocated RAM to do its along with them.
calculations on complex image transforms or filters, it uses It takes more than the previously mentioned compo-
empty hard drive space as temporary RAM called scratch nents to build a modern computer workstation. However,
disk space. This can also be specifically allocated in its pref- most of us aren’t looking to assemble one from scratch.
erences. Some photographers take unused drives (after an Instead, we should look at key specifications like CPU, GPU,
upgrade to larger ones, for example) and use them as ded- RAM and storage capacity to make informed purchase
icated scratch disks. All of these strategies aim to increase decisions of off-the-shelf machines. Always buy the best
efficiency when editing large images. machine that you can afford and make that purchase last
by planning for incremental hardware upgrades. Consider
peripherals for extensive editing sessions like ergonomic
Storage Drives mice, drawing tablets and customizable MIDI boards to
Computer drives store everything: the operating system, help with comfort and workflow efficiency.
user files, software data and everything else you can think
of. Without a place to store the information the computer
needs to operate, we’d be stuck temporarily loading infor- Proof of Concept: Considering File
mation in every time we turn it on. Types and Image Content
Ideally, a workstation has a dedicated drive for the
operating system, a dedicated drive for media assets and Let’s build an appreciation for image file types and their
possibly a third drive as an internal backup of the second. digital asset storage footprint. This exercise highlights what
An operating system reads program information from the photographers juggle after capturing a great raw photo:
drive it’s installed on. A program reads and writes informa- how will it be stored long term? How can it be shared over
tion from the drive that contains the image assets. Having email or the web without compromising image quality?
these read/write actions spread out over multiple drives After all the derivative versions are saved, how much data
means that there are fewer bottlenecks in getting infor- does a single photograph generate?
mation when it’s needed. Using a solid state drive for the
operating system is ideal because its faster speeds afford 1. Find a raw image that contains a lot of texture or detail.
quick boot-up and general performance boosts. Hard disk Consider this Image A. Find another raw capture that
drives are better suited to large storage archives or backup is simpler and minimal in detail, like a smooth subject
drives because they can be cheaply stockpiled. against a plain background. Consider this Photo B (see
Photographers often use external or portable drives for Figure 8-8, overleaf).
media storage. These can be HDD or SSD and come in a 2. Open Photo A in the raw image editing software of your
variety of form factors. A portable drive is plugged into a choice. Save a derivative version as DNG (lossless).

189
 s ecti o n 2: P roc ess

Figure 8-8 Photo A (left) and Photo B (right).

3. From the same editor, save a 16-bit TIFF version with

LZW compression.
4. Save an 8-bit JPEG version at the highest-quality setting:
usually the top end of a 0–100 or 1–10 scale.
5. Save a series of 8-bit JPEGs, changing the JPEG quality
setting by an interval of your choice, stepping down
until you reach the lowest setting available. Append
the filenames with the selected quality setting for easy
reference.
6. Repeat steps 2–5 for Photo B. You should have a set of
files for each scene: the original raw, a DNG raw, a TIFF
and a series of JPEGs.
7. Record the data size of each file in the sets for Photos
A and B using your operating system’s file property
window. Plot them in spreadsheet software to visualize Figure 8-9 File sizes (in megabytes) for different image formats saved
from the raw versions of Photo A and Photo B.
the relative sizes.

Figure 8-8 shows our example photographs. Photo A is we generate them and delete the originals while retain-
a beach landscape with considerable detail in the fore- ing their non-destructive editability. The TIFF of Photo
ground rocks and background foliage. Photo B is a simple, A is much larger than the raw. Where the raw file stores
monochrome still life of cosmetic supplies. Both images mosaicked, 14-bit brightness data, the TIFF now houses the
started out as proprietary format raw captures before being demosaiced, three-channel (full color) image and grows
saved as DNG, TIFF and JPEG. We’ve plotted our recorded in size as a result. Photo B’s relative content simplicity and
file sizes for both sets in Figure 8-9. lack of color means that its TIFF doesn’t follow suit and is
Looking at Figure 8-9, we draw initial conclusions actually somewhat smaller than the raw. The 8-bit JPEGs
about saving images in different formats. Writing raw decrease in size as increasingly aggressive compression is
image data to DNG sometimes yields a smaller file, but not applied. These are useful end-product assets but lose the
always (Photo A remained nearly the same size, Photo B editing flexibility of the raws, so they’re well-positioned as a
slightly shrunk). DNGs offer a small space-saving benefit if secondary asset options.

190
 c h a p t e r 8 : D i g i ta l A s s e t M a n a g e m e n t

Figure 8-10 A close crop of Photo B as it appears in its native raw capture format (interpreted by a raw engine) and as the lowest-quality JPEG.

Chapter 12 dives deeper into image data compression, Notes

its methods and shortcomings. For now, evaluate the 1 Anderson, Richard. “Raw file formats.” File Management
different versions of your two photographs up close. Figure Overview | DpBestflow, Sept. 22, 2015, www.dpbestflow.org/
8-10 is a close crop of Photo B. The difference between the node/305.
original raw and the lowest-quality JPEG are surprisingly 2 “Technical information.” HEIF Technical Information – High
Efficiency Image File Format, nokiatech.github.io/heif/
minimal from a visual standpoint. Of course, we understand
technical.html.
that the utility between the two is vastly different: one
3 Krogh, Peter. “Metadata overview.” File Management
is the purest form of a digital image from which to edit,
Overview | DpBestflow, Sept. 22, 2015, www.dpbestflow.org/
where the other is a fully baked, final output version.
node/298.
Questions to consider in your digital-asset-managed 4 “What is a controlled vocabulary, and how is it useful?”
workflow based on your findings: Controlled Vocabulary, www.controlledvocabulary.com/.
5 Allen, Elizabeth and Sophie Triantaphillidou. The Manual of
• Is it worth it to evaluate JPEG compression settings on a Photography. Elsevier/Focal Press, 2017, p. 341.
per-image basis or default to a lesser setting at the cost 6 Russotti, Patti and Richard Anderson. Digital Photography
of file size? Best Practices and Workflow Handbook: A Guide to Staying
• Is the highest compression setting (lowest “quality” Ahead of the Workflow Curve. Focal Press, 2009.
number) a visually acceptable output for the images 7 “Synchronize. Backup. Bootable Backup.” Welcome to
used in this exercise? Determine the quality level at Econ Technologies’ Store, www.econtechnologies.com/
which the compression artifacts become objectionable. chronosync/overview.html.
8 “Mac backup software.” Mac Backup Software | Carbon Copy
• What real-world use-cases would you use each of these
Cloner | Bombich Software, bombich.com/.
formats for? How do you organize derivative versions of
9 “List of Interface Bit Rates.” Wikipedia, Wikimedia
the same photograph as you create different versions?
Foundation, 2 Feb. 2019, en.wikipedia.org/wiki/
List_of_interface_bit_rates.
10 Roberts, Gideon. “The history of ethernet.” YouTube, Dec. 14,
2006, www.youtube.com/watch?v=g5MezxMcRmk.
11 “High tech traveler.” Web Archive , Verizon Wireless, 2011,
web.archive.org/web/20110111075514/http://www.
hoovers.com/business-information/--pageid__13751--/
global-hoov-index.xhtml.
191
 9 Digital Tone
Reproduction

Photograph by John Compton

It’s exceedingly rare that the raw capture as it is rendered in-camera conveys the range and
interplay of tones, colors and details present in the scene or in our creative vision for that
scene. We believe strongly in the power of craft that all photographers should seek to per-
fect when shooting yet even the best exposed frames are just the starting point for image
editing and enhancements. Optimizing digital tone reproduction is a necessary and core
tenet of image adjustments that goes a long way in making that great capture into a great
print or digital file.
It’s an informative exercise to study the relationship between the tones in the original
scene and the corresponding tones in the reproduction, even when the final product is
intended to bear little resemblance to the original. By studying the factors that influence

193
 s ecti o n 2: P roc ess

and the tools that control the reproduction of tone, we

can create images that satisfy both technical and creative
requirements.
A photograph is a reproduction of recorded light first
experienced in-person by the photographer. This original
experience is made possible by the human visual system,
as is the experience of viewing a photograph in reproduc-
tion. With this in mind, perceptual phenomena including
brightness adaptation play a significant role. The deter-
mination of the objective tone reproduction properties
required in a photograph of excellent quality is related
to the perceptual conditions at work when the image is
viewed. The final image showcases the cumulative tone
reproduction characteristics of the camera’s sensor, its
optics, image processing decisions and the output medium
(print or screen).
Figure 9-1 An objective tone reproduction curve for a reflective print.

Objective and Subjective Tone

Reproduction the effect of viewing conditions on the perception of light-
ness and contrast. It’s an established fact, for example, that
There are objective and subjective aspects to consider a photograph appears lighter when viewed against a dark
when determining the tone reproduction properties of background than when viewed against a light background.
the photographic system. In the objective phase, meas- Similarly, the contrast of a photograph appears lower when
urements of light reflected from the subject are compared viewed with a dark surround than when viewed with a light
to measurements of light reflected or transmitted by the surround. Understanding these psychological realities can
photographic reproduction or light emitted from a display. help you decide on things like frame colors and website
These reflected-light readings from the subject are called backgrounds. This is explored more deeply in Chapter 13’s
luminances and are accurately measured with a photom- section titled The Roles of Light and Color on Perception and
eter with a relative spectral response equivalent to that of Perspective.
the average human eye. The logarithms of these lumi- The subjective tone reproduction curve plots the rela-
nances are plotted against the corresponding reflection or tionship between the perceived lightnesses of image areas
transmission densities in the photographic reproduction and the measured luminances of those corresponding
and the resulting plot is the objective tone reproduction areas. Since the perception of a photograph’s tonal qual-
curve as shown in Figure 9-1. ities can vary with the viewing conditions, the American
The perception of the lightness of a subject or image National Standards Institute and the International Stand-
area corresponds roughly to its luminance. However, unlike ards Organization have published ISO 3664:2009 Graphic
luminance, lightness cannot be directly measured. The Technology and Photography – Viewing Conditions, the
perception of lightness involves physiological as well as standard for the viewing conditions under which photo-
psychological factors. An area of constant luminance may graphs should be produced, judged and displayed. The
appear to change in lightness for various reasons, including influence of viewing conditions should not be underesti-
whether the viewer is adapted to a high or a low light level. mated when considering the specifications for optimum
Psychological scaling procedures are used to determine tone reproduction.

194
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

The Evolution of Tone Reproduction the film plane. Quadrant II was the negative material, using
the image luminance values and the characteristic curve
For those that have been involved with photography from of the film to determine the densities that resulted on the
the days of film, tone reproduction reminds us of the clas- film. The film densities were used in Quadrant III along with
sic four quadrant plot that tracks light through the entire the characteristic curve of the photographic paper used
imaging pipeline in log space. Quadrant I started with the to create the print. This resulted in the relationship of the
subject’s luminances and took flare and lens transmission tones on the final print to the original scene luminances.
into account, resulting in the image luminance values at Just writing about it brings back memories of hours spent

Figure 9-2 Four quadrant plot for a digital tone reproduction study.

195
 s ecti o n 2: P roc ess

Luminance for Digital Count Maximum Brightness

in a darkroom with a sensitometer followed by hours with a Monitor Density = (Eq. 9.1)
Luminance for Displayed Digital Count
densitometer and a French curve.
This classic method of tone reproduction analysis Quadrant IV provides the relationship between the final
has perhaps gone away in the digital world, yet the four print made from the image on screen and the original
quadrant plot concept retains utility. The four quadrant scene luminances. The final print densities can be obtained
tone reproduction plot can still be used to study a dig- using a reflection densitometer.
ital system; however, the quadrants take on a different For all practical purposes, the response of our digital
look. Studying a digital tone reproduction system means sensor in Quadrant I cannot be changed. We can only work
understanding the scene brightness characteristics, the to properly expose our images. In contrast, all aspects of
response of the imaging sensor, the performance of the our digital image data can be edited on a properly profiled
display used for editing and the output paper characteris- display. The rest of this chapter concentrates on the ways
tics (see Figure 9-2). that we modify and change the digital tone values to pro-
In a digital tone reproduction study, Quadrant I shows duce a desired output result.
the tonal transfer between the subject log luminances and
the resulting digital counts, often referred to as the detector
response. An image sensor, either CCD or CMOS, exhibits Revisiting the Image Histogram
a more linear relationship to the subject luminances than
to the log luminance. As a result, Quadrant I provides the Recall that an image histogram is a breakdown of the fre-
same input-output relationship that a film system once did, quency of pixels at a given tone value and a representation
only now the output values are digital counts. of an image’s tonal range. Every element of a digital image
Quadrant II represents the ability to alter the image channel (of which there are three) is a discrete brightness
characteristics after the image is captured; with digital data value. We first encountered histograms in the context of
this means processing or altering digital counts. There are photographic exposure. Histograms are great for evaluat-
many possible alterations, such as adjusting the contrast, ing exposure settings and getting a sense of the scene’s
color balance or sharpening. Although these operations are tone characteristics. Interaction with the histogram at the
done with complex math, all of the operations can be com- capture stage is limited to adjusting exposure settings to
bined into one lookup table or LUT. An LUT provides a new ensure that the scene’s tones are recorded effectively; it’s a
output digital value for each input digital count. Quadrant passive feedback mechanism.
II, then, is represented as an LUT to apply tonal alterations. Histograms work double duty as tools for tone repro-
Quadrant II is plotted as a straight line if no modifications duction efforts during post-processing, this time as an
to the image take place: the input digital count values are interactive and malleable representation of changes made
equal to the output digital count values. to tone, contrast and color. The concepts of luminosity
It’s reasonable to assume that an image is viewed on histograms and color channel histograms were introduced in
a display like a monitor prior to printing. This is where the Chapter 4. Let’s unpack those further now that we’re taking
tone reproduction study ends for some use-cases, as a a close look at how this visualization helps in post-process-
final print may not be required. Quadrant III illustrates the ing edits.
characteristics of the display device upon which the image Generally speaking, the range of brightnesses in an
is viewed. The input values are digital counts that resulted image is the tonal range. The darkest tone is black, the
from the application of the LUT in Quadrant II. The output brightest is white and the tone perfectly between those
values describe monitor density as defined in Equation 9.1. two is midtone gray. These points along the tonal range
Several factors affect this quadrant and the measurement also help us divide histograms into three sections: high-
of the luminance on the display such as the settings for lights, midtones and shadows. A luminosity histogram
brightness and contrast. better represents the distribution of perceived brightnesses

196
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Figure 9-3 Many of the midtone values from the raw file (left) can be reassigned to dark shadow values to create a rich, contrasty final product
with a strong visual impact (right). The histogram’s changed appearance echoes the visual changes in the photo.

when all three color channels are combined. It gives more Watching the histogram when editing helps to spot
weight to the green channel in correspondence with our a number of potentially detrimental changes and we
visual system sensitivity rather than being an additive recommend having a histogram window or panel open
version of the three channels of data. If you prefer to view at all times. The first of those changes is clipped highlight
the image information as additive RGB channels, use an or shadow detail. Clipped tones are indicated by sharp
RGB histogram. This is what the camera firmware typically cut-offs at either end of the histogram where it appears
uses to preview captured exposure information. Luminosity that the data would continue if it wasn’t abruptly ended.
histograms do not necessarily show specific color channel Some small amount of clipped highlights is usually
clipping, particularly in the blue channel as it is weighted acceptable depending on the scene just as crushing some
the least. shadow areas to solid black is not objectionable. It’s usually

197
 s ecti o n 2: P roc ess

Figure 9-4 A low contrast scene and its histogram (left) and the same scene after a contrast adjustment (right). Notice how the post-edit
histogram shows pixel values spanning a broader tonal range with small gaps throughout.

desirable to maintain all the descriptive image data in the considered flat or low contrast. A black cat sitting on a pile
edited file after working hard to avoid clipping at capture. of coal is one inherently low contrast scenario. Adding
In addition to keeping a watchful eye on the histogram, contrast can help emphasize texture, tonality and the
image editing software provides overlays or warnings to appearance of sharpness. Contrast adjustments are like the
show where parts of the image are clipped. Figure 9-3 caffeine of image edits: it’s easy to get hooked and start
shows a well-exposed image with a reasonable starting thinking that every situation calls for more. You eventually
histogram. Introducing a darker black point means adding overdo it and learn that there’s an effective time and place
some much needed tone contrast, moving the shadow for adding contrast. Combing in a histogram occurs when
tones to even darker values with a small amount of shadow an image that has already been processed is adjusted again
clipping. The minor loss of shadow detail is unlikely to be or when a low dynamic range image is stretched to adjust
noticeable. Though the raw image version may accurately contrast. Pushing a flat scene far beyond its available tones
represent the scene as it appeared in person, modifying its leads to combing. The name stems from the histogram’s
tone reproduction provides a visual punch, commanding appearance after such a drastic change: instead of a full
attention and making the subject more appealing. Always mountain range of tones, the histogram shows frequent
keep in mind that tone reproduction is not strictly con- gaps similar to a hair comb (see Figure 9-4). This happens
cerned with objective or numeric optimization. Subjective because the additional processing leaves some pixel values
and thoughtful consideration for how an image should empty and can signal visible posterization in the edited
look largely drives this process. image. Shooting in raw and saving images to an uncom-
A photograph with very little range of tones is pressed file format during editing helps avoid combing.

198
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Interpreting the Tone Curve Figure 9-5 is example is an implementation of a

user-controllable tone transfer curve. There is also the
A transfer curve shows the mapping of input data to option of using tone transfer compensation (TTC) profiles to
output data. A tone curve is a transfer curve specific to automatically edit files if the workflow is highly controlled
image pixel values representing photographic tones. (scanning film from a specific camera and lighting setup,
During camera capture, the tone curve maps the input for example). TTC is a method of selectively expanding or
scene luminances to the recorded digital counts. At the compressing specific portions of an image’s tonal range
editing stage, the tone curve represents the starting digital using lookup tables to remap each input pixel to a new
counts mapped to the final digital counts after post-pro- output value.
cessing adjustments. Figure 9-5 shows a tone curve as seen
in Adobe Photoshop Lightroom. The 45° line is the default
mapping prior to any adjustments such that the input and Using the Curves Tool to
output values are equivalent. Any change made to this line, Modify Tones
either with an editing preset or through targeted edits, is a
form of tone mapping. This term is also used in the context Tone reproduction can be controlled with the Curves tool,
of High Dynamic Range imaging when the input data an intimidating but valuable software interface for editing
consists of many more discrete tone values than can be the image pixel values. By default, the tool offers a plot of
reproduced on screen or in print. Here we use the term to input–output values and is not a curve at all—it’s a straight
describe any instance of reassigning pixel values. diagonal line from left to right as shown in Figure 9-6
The x-axis in Figure 9-5 represents the input digital (overleaf). This line indicates that any pixel with a value of
counts as they exist in the unedited image. The y-axis 0 before the user changes it is a value of 0 afterwards (and
represents the digital counts after editing. The image so on through all available tones). Put another way, there
histogram is underlaid for reference with the 0 digital count is a perfect 1:1 relationship between input and output.
representing black on the left and the digital count 255 The Curves tool remaps tones based on changes made
representing white on the right. to this diagonal line so that the output is changed from
the input. Any abrupt or erratic jumps within small tone
ranges (say, dark gray to slightly lighter gray) are unnatural
in a continuous tone photograph and as such the curve
behaves more like a tightrope than a zig-zagging rigid
line. This is to encourage a natural ramping or gradation
between user-introduced tone changes. A tightrope is also
an appropriate metaphor that serves as a reminder to use
Curves adjustments cautiously and with forethought to
the destination.
Modifying the diagonal line to form a slanted “S”
shape is a common Curves tool adjustment and is
accomplished by adding pivot points as shown in
Figure 9-6. An S-curve darkens the shadows, lightens
the highlights and impacts the midtones very little. It
also avoids clipping on either end by virtue of ramping
down or dampening the changes at the extreme ends of
Figure 9-5 An example image tone curve with a histogram reference the tone range. This tool allows fine control over gray-
behind it. level distribution; we can add multiple pivot points that

199
 s ecti o n 2: P roc ess

Figure 9-6 The Curve tool’s tone reproduction plot with a 1:1 reproduction line (left) that has no impact on the image
compared to an S-curve (right) that darkens the shadows and lightens the highlights.

Figure 9-7 Curves tool adjustments go beyond s-curves. Here, the original image is very high key and flat. A tone adjustment pulls down
deeper shadows and introduces contrast overall.

200
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Figure 9-8 The original image (left) and the tone-optimized version (middle). Targeted tone corrections bring out the best of varied scene
content. The layers panel (right) shows how different tone changes are isolated to the trail, the grass and the sky.

impact the contrast of pixel tones on either side of each subjects like the ones shown in Figure 9-8 is the modern
point like in Figure 9-7. Adding and moving pivot points equivalent of darkroom dodging and burning and offers
on the Curves plot can take some getting used to, as it the greatest amount of nuance and control of the final
has a rubber-band-like behavior where pulling on one product.
area makes changes to another. It’s tempting to add lots
of points to isolate affected areas. This invites unintended
or unwieldy changes that are hard to manage. The Inverting Tones for Negative Film Scans
Curves tool is often considered to be one of the most Complement mapping inverts all values in the image: black
powerful and most difficult to master tools in tone repro- becomes white, white becomes black and gray values are
duction editing. Consequently, the tool includes presets mapped to their complementary tones above and below
for assisting the novice user or to provide a convenient a middle gray. This is accomplished with a diagonal line in
starting point. the Curves tool that starts at the top left and travels down
Curves adjustments can be treated as global image to the bottom right (the exact opposite of its default form).
edits in that the rules and changes they establish apply There aren’t many scenarios where this approach to pixel
to every pixel in the file. Global tone adjustments are remapping is valuable yet we include it here because you
often necessary yet are unlikely to be the one-stop solu- might scan film negatives at some point. A complement
tion for tone reproduction in a photographer’s workflow. mapping turns scanned negatives into positive images as
Instead, treating Curves modifications as separate layers shown in Figure 9-9 (overleaf).
(called “adjustment layers” in Adobe Photoshop) is most
flexible when combined with layer masking. Mask-
ing tone adjustments to specific parts of the frame or

201
 s ecti o n 2: P roc ess

Figure 9-9 Complimentary mapping applied to a color negative film scan.

The Zero-Sum Game of Tone Reproduction problem is convincing your boss that your “binge-watching
Tone-remapping editing is a zero-sum game when it uniform” meets the workplace dress code. The difference
comes to the pixel data. We’re redistributing contrast between an 8-bit photograph and a 16-bit photograph
across different portions of the tone curve and there’s risk that exhibits a similar dynamic: 8-bits are enough but
in spending more than what you start with. We always there’s no room for adjusting tone or contrast without
hope that there’s enough data to achieve our aesthetic potential quality degradations; 16-bits works equally well if
objectives for the final product. However, going too far no adjustments are needed and really save the day when
with the adjustments or starting out with a sub-par set image processing is introduced to tweak tones.
of data—like having an underexposed raw image and In the motion video world, we’re quickly entering a shift
attempting to show beautiful shadow detail in the edited from 8-bit to 10-bit video capture, editing and playback.
version—reminds us that digital tone reproduction is not Professional video production teams have been shooting
magic. Capturing at higher bit depths and preserving it high bit depth capture footage for years for the same
throughout your workflow make this a less risky endeavor reason that still photographers shoot raw and edit 16-bit
when working to maintaining image quality. files. The bigger change is that High Dynamic Range and
Here’s an analogy to bolster an appreciation of bit Wide Color Gamut video standards require at least 10-bits
depth in tone reproduction. You’ve got a pair of dress of tone storage to deliver the improved image quality
pants that just barely fit and a pair of sweatpants with an experience to home viewers. Televisions, combined with
elastic waistband. You gain a few pounds around the waist streaming media services, are increasingly offering 10-bit
after the winter holiday festivities. The dress pants that fit video content that promises better tone reproduction.
snuggly before all of the holiday cookies are no longer an
option; the sweatpants provide a much-welcome flexibility
to accommodate your newfound love handles. The only

202
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Using the Levels Tool to Modify Tones applying it more heavily toward the shadows, midtones
or highlights. Alternatively, our knowledge of the Maxwell
Pixel point processing uses a mathematical function to triangle takes the guesswork out of manual color correc-
map the original pixel values to new output values, light- tion: determine the starting color cast, find its opposite on
ening, darkening or changing the contrast of an image. The the triangle, and either decrease the color of the cast or
simplest form of remapping values is slide mapping. Adding increase its opposite. An image with a blue cast, for exam-
or subtracting a constant value to all pixels in the image ple, can be neutrally balanced by introducing more yellow.
alters its overall brightness. This has the effect of moving Another approach to pixel remapping is stretch map-
all of the pixel values in the histogram to the right or to the ping. Stretch mapping uses multiplication or division by a
left (brighter or darker, respectively). The drawback to slide constant value to stretch or contract the tonal values over
mapping is the ease with which one can clip highlights or the histogram. This can be used to stretch a low contrast
shadows. For example, adding 50 to all pixel values means image over the full range of values to produce normal
that all the pixels whose values are 205 or higher become contrast as shown in Figure 9-10.
255. Slide mapping is a blunt approach previously found in The Levels tool, as it is called in Adobe Photoshop (sim-
Adobe Photoshop’s “Brightness/Contrast” legacy tool from ilarly available in other software), is like a direct histogram
the early days of digital tone reproduction and editing. editor. If the histogram is like a sandbox filled with our
A constant value applied to only one of the color chan- image pixels, the Levels tool is like handing photographers
nel levels changes the overall color balance of the scene. toy shovels and rakes to use in that sandbox. Our zero-sum
Adobe Photoshop’s color balance tool takes the slide observation still applies: there’s no adding sand to the
mapping approach while weighing the constant value, sandbox, only redistributing what’s there at the start. Work-
ing with Levels means using the feedback provided by the
histogram to distribute or reassign pixel tones. A single
luminance histogram is available for modification as well as
the separate red, green and blue color channel histograms.
On either end of the Levels histogram are black and
white points or anchors. The position of each dictates
which part of the starting histogram is assigned to values of
0 and 255. Just like the tone curve plot, the adjusted posi-
tions of these anchor points dictate the remapping of pixel
values between input and output (before and after the
change). The middle-gray point adjusts gamma. It impacts
the midtones while leaving the white and black points
untouched. Instead of applying a constant value, the mid-
tones are changed using a Gaussian (normal) distribution,
with greater amounts of change in the middle values and
decreasing amounts toward the highlights and shadows.
Moving the middle-gray point to the left lightens mid-
tones and moving it to the right darkens them. Figure 9-11
(overleaf) demonstrates using only the middle-gray point
adjustment to improve the overall contrast of an image.
While the Curves tool allows users to create an anchor
point anywhere along the tonal range, Levels only offers
Figure 9-10 An example of stretch mapping. the white, midtone and black points as adjustable anchors.

203
 s ecti o n 2: P roc ess

shoulder-rub of pixel remapping, we turn to image editing

controls (often taking the form of adjustment sliders). The
commonly encountered ones are described here. Try each
control in your software of choice while watching the his-
togram as it often reveals the tool’s underlying behavior.
The Exposure or Brightness slider targets the brightest
areas of the image. It’s useful when looking at the overall
lightness or darkness of the scene. Even with an adequate
capture exposure, minor changes to the brightness are
often desirable. Using this adjustment tool can compress
or expand the highlights. Clipping is easy to introduce
here, so consider taking advantage of any clipping warning
overlays available.
The Highlights/Shadows slider avoids the midtones
entirely and instead looks to remap the brighter highlight
or darker shadow portions of the image histogram. This is
equivalent to pulling up or pushing down a small area on
a Curve plot with the intention of modifying just the small
range of tones considered to be highlights or shadows.
Concurrently brightening the highlights and darkening the
shadows effectively introduces additional image contrast.
The Whites and Blacks sliders adjust the white and black
Figure 9-11 An image before (top) and after (bottom) a tone change points. Whereas the Exposure slider stretches the histo-
using only the middle-gray point adjustment of the Levels tool. gram toward the highlights (altering the entire histogram),
the whites slider moves the white point while generally
maintaining the overall shape of the histogram. The black
slider performs the same way for the other end of the tonal
range.
Common Tone Reproduction Applying a positive Contrast adjustment to an image
Software Controls darkens the darker midtones and lightens the brighter mid-
tones. It also widens the entire histogram in the process.
Levels and Curves are great tools once photographers Adding contrast can be easily overdone. Alternatively, neg-
develop an intuitive sense of how they can be controlled ative contrast flattens the tonal range and begins to make a
to improve photographs. After all, we’re visual people and potentially wide range of shadows and highlights into sim-
the idea of using graphs and numbers can be a confusing ilar midtones. Controls called clarity or texture are contrast
intermediary between what we want to do and what we controls for a narrow set of midtones rather than the wider
see. While mastering both tools is recommended, there are range of highlights and shadows. These tools impact local
also times where such deep-tissue pixel massaging is too contrast but do not add or remove global contrast in the
time consuming or unnecessary. This is particularly true same way that a traditional contrast control does.
when a set of images needs the same sort of treatment Finally, dehaze or aerial haze adjustment sliders rein-
and it needs to happen quickly such as portrait sessions or troduce contrast in areas where aerial haze inherently
event series. Spending minutes or hours on one image file lessens it. Aerial haze is most apparent in landscapes
is unrealistic in these circumstances. For the quick-and-easy with objects that extend out to the horizon; the farther

204
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

aspect ratio as full frame DSLR cameras. The target design

includes evenly spaced neutral patches between white
and black and the additive and subtractive primaries. Four
patches represent common objects (light skin, dark skin,
blue sky, foliage) that color scientists deemed critical for
evaluating tone and color reproduction. The other patches
represent colors that were particularly challenging to
reproduce on film (such as a specific blue flower). These
colors do not necessarily pose the same challenges for dig-
ital sensors or modern photographic output but tradition
Figure 9-12 An outdoor landscape often exhibits reduced contrast and general utility have meant that use of the classic Color-
to distant elements due to aerial haze (left). A dehaze tool can Checker patches persists. Modern updates like the Digital
reintroduce contrast to these elements (right).
ColorChecker SG supplement the classic set with a large
number of new color patches attuned to sensor character-
istics and to provide additional data points for automated
away things are, the more that the atmosphere scatters software evaluation.
light and reduces tone contrast. Dehaze tools attempt to Setting the proper white balance when shooting is
add contrast back in these parts of the scene and they always the best practice and is easily done with an X-Rite
require some knowledge or educated guess regarding ColorChecker. Or, if circumstances don’t allow for the time
scene depth to do so. This is accomplished with con- it takes to set a custom white balance, a photograph of
text-sensitive estimation algorithms (taking factors into the reference target can be used at the editing stage with
consideration like focal length and local contrast in the equal effectiveness. These uses make the ColorChecker a
original image) or with the help of depth maps generated staple piece of equipment for every photographer. X-Rite
from dual camera systems. Software can reintroduce con- also makes a smaller version, the ColorChecker Passport,
trast to groups of pixels that it believes are farther from that comes in a hard case to toss into an equipment bag
the camera. Such tools may also introduce saturation or without fear of damage. Using the target for accurate color
color shift changes. Figure 9-12 shows a before and after balancing starts by taking a reference photograph during a
dehaze adjustment example where the distant mountains photo shoot like we did in Figure 9-14 (overleaf).
see a tone improvement. A dehaze slider may be adjust-
able in the negative or inverse direction to intentionally
add a haze effect if desired.

Color Correction and the

ColorChecker

The ColorChecker was invented in 1976 and was designed

for use with analog film.1 The goal was to have an object
that could easily be compared with its image either visually
or through objective measurements to ensure proper color
reproduction. The ColorChecker target design features 24
unique patches proportioned to fill the frame of a 35mm
frame (see Figure 9-13). This is, conveniently, the same Figure 9-13 The X-Rite ColorChecker.

205
 s ecti o n 2: P roc ess

that they should be neutral. The algorithms are not always

right, regardless of the method used. The ColorChecker
includes neutral patches to leverage with eyedropper-style
color correction tools. Such tools tell software that a small
group of pixels identified by the user should be neutral.
Any color cast exhibit by these pixels is counteracted with
a global change that impacts all pixels. The neutral patches
on the ColorChecker are great to lean on for this task since
we always know that the pixels recorded at these patches
should be neutral in the processed image. It’s conceivable
to use a random subject or scene area with an eyedropper
tool to establish a neutral point, however, a truly neutral
surface is not always available. Worse, you might incorrectly
recall a subject or scene element to be neutral when it was
not.
Clients are unlikely to be interested in photographs
of a color patch target. Instead, the suggested workflow
involves shooting reference images with the ColorChecker
in every new lighting environment. These images are
white balanced and their color corrections are copied
or replicated on all related captures from a shoot. This is
particularly effective in a parametric image editing work-
flow where such corrections are efficient, metadata-based
instructions that can be quickly applied to dozens or even
hundreds of frames at once.
Figure 9-14 Having a ColorChecker Passport handy in never-to-be-
The ColorChecker patches are also valuable as refer-
repeated scenarios can be a color correction lifesaver when it comes
time to edit. ences when a specific color cast or color treatment is
desired. If the intent is to create an aesthetically pleasing,
warm color tone, for example, then watching the known
Camera firmware typically includes an option to set color patches and memory colors shift is like having a
custom white balance during a shoot. This is helpful if the visual landmark. Knowing where you have been is just as
lighting is unusual and previewing color is important as constructive as knowing where you are going with creative
the shoot progresses. The firmware requires a captured, color reproduction.
reference photograph containing a neutral object. Going
this route is convenient but not altogether necessary, as
a similar set of steps can be done at the post-processing Color Balancing Using Color
stage. Channel Curves
Automatic color balancing tools are found in almost all
image editing software. These algorithms are proprietary The absence of a neutral target or a neutral object in the
but use one of two methods to establish color balance. scene doesn’t mean that color balancing is impossible.
They either assume that the scene, when averaged, We can turn to the image channel data to clue us into a
produces a mid-level gray or they find the brightest areas color imbalance between the red, green and blue channel
of the scene, like bright specular highlights, and assume data. Keep in mind that neutral tones are represented by

206
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Figure 9-15 Adjusting the red, green and blue channel curves independently to maximize the tonal range in each helps to color balance.

207
 s ecti o n 2: P roc ess

equal amounts of red, green and blue pixel brightnesses. Split Toning
Any deviation from this relationship creates a non-neutral The same strategy of independently modifying channel
color. curves is useful for introducing split toning to an image.
Throughout this chapter thus far, we stuck to the Split toning introduces color casts to shadows and high-
single luminance curve that represents the combination lights independently and does not modify luminance. The
of all three color channels in a color image. Once you’re goal is to add a color cast (or two) rather than removing
comfortable manipulating the luminance curve, the next one. Split toning can be applied to color or black and white
place to hone your skills is by working independently photographs. A popular combination sees the shadows
with the red, green and blue channel tone curves. For the made cooler and blue and the highlights a warmer yellow
task of color balancing, we recommend setting a Curves as shown in Figure 9-16. We include a black-to-white gra-
adjustment to a “Color” blend mode in Adobe Photoshop dient below the images to better show the toning change,
to ensure that the changes impact color separately from independent of subject content. This yellow/blue combi-
luminance. nation contributes to our innate association of light and
Figure 9-15 shows a starting image, a low light long warmth and its absence as cooler in temperature. Subtle
exposure and the adjustments made to each of its three toning of shadows and highlights is a purely subjective
channels with the Curves tool to produce a color bal- practice.
anced version (shown to the right of the original). The
blue channel histogram is the smallest, indicating that
there’s less blue overall and confirming the original Proof of Concept: Color Correcting
image’s visibly yellowish cast (recall the Maxwell triangle Film Scans using Curves
to appreciate the relationship that makes this true). We
start by adjusting the black and white points of each There are three approaches to color correcting an image:
channel histogram in order to stretch each one to maxi- going strictly by-the-numbers with RGB pixel values and
mize the available tone information recorded. This often attempt to make the image neutral; using educated
gets us close to a desired final product as it makes it so trial-and-error and visual subjectivity; embracing the lack
that pixels representing neutral content have equivalent of color correction and create an intentionally unnatural or
or very similar amounts of brightness across all three inaccurate aesthetic choice. These methods can co-exist in
channels. Be aware that this can reveal noise that wasn’t your editing toolkit. We use a mix of the first and second
as visible before such changes. From here, we pull up or approaches in this exercise.
push down each curve from its midpoint to add that color Color film scans are great subjects on which to practice
or its complement, respectively. We pull the blue and color correction techniques. Unlike raw digital captures,
green channel curves down a touch and leave the red color film frames can have significant color casts that don’t
channel mostly as-is in our example. The color-corrected have an easy toggle to alternate through white balance
version shows the concrete statue with a more realistic, settings. Older film can also show color degradation due
neutral appearance compared to the warm tungsten cast to natural fading and exposure to the elements. We dug
in the original. up the Kodachrome slide shown in Figure 9-17 from an old
The Curves tool is alternatively accommodating in family photo collection. Despite being taken in the 1960s,
those instances where we do have a neutral reference the act of taking one’s cat for a walk on a leash is forever
patch or target, offering “eyedropper” functionality to fashionable. Follow along with our example or use your
automatically set black, white and midpoints on an image own scanned image.
histogram. We can perform this tone mapping to either
the combined RGB luminosity histogram or the separate 1. Set your Adobe Photoshop workspace surrounding
channels. color to a neutral, midtone gray by right-clicking it and

208
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Figure 9-16 Split toning an image is a subjective practice that serves aesthetics over objective
accuracy. Here, the shadows are rendered with a blue cast and the highlights with a yellow cast.

Figure 9-17 The color and tones are an unedited rendering from a film scanner (left). The color-
corrected version (right) is improved thanks to modifying the tone distribution of the separate color
channels.

209
 s ecti o n 2: P roc ess

Table 9-1 Highlight and shadow pixel values from sampled areas.

Sampled Points L* a* b* R G B

Sampled Highlight Before Color Correction 56 0 0 147 125 117

Sampled Highlight After Color Correction 73 0 0 179 179 178

Sampled Shadow Before Color Correction 13 5 -1 43 37 30

Sampled Shadow After Color Correction 25 -2 -5 59 63 68

selecting Select Custom Color. Set R, G and B fields to 128. 8. Watch the Info panel where the highlight and shadow
This is critical as a neutral reference point for our eyes to samples’ RGB values are displayed. Try to match the R,
make accurate color assessments. G and B values for both samples while visually evalu-
2. Use the Color Sampler tool to define two points of ating the changing appearance of the overall image.
interest: one in a highlight area and another in a shadow Always keep in mind that your subjective opinion on
area. Try to find areas in the scene that should be neutral color reproduction holds weight alongside any numeric
like white walls, concrete pavement or clothing details. balance. Our finished Curves adjustment is shown in
The Color Sampler tool holds the selected pixel areas Figure 9-18.
and displays their CIELAB and RGB values in the Info
panel. The red circles in Figure 9-17 indicate where we
measured the pixel values in our image.
3. Observe the sample points’ CIELAB and RGB values (ours
are recorded in Table 9-1). Consider that any truly neu-
tral area in an image has near-identical R, G and B values
once it is color corrected. Alternatively, neutral areas
have near-zero a* and b* values.
4. Add a Curves Adjustment layer. Use the dropdown
menu to switch to the Red channel histogram and
curve.
5. There is tone information (the histogram peaks) in the
red channel yet the black and white points may not be
matched with the dynamic range of the image. Adjust
the black point to the location where the darkest pixel
information exists to introduce stretch mapping to the
channel data. Do the same with the white point.
6. Repeat this process with the Green and Blue channels.
7. Cycle through the channel curves again. In each, click
on the diagonal line and drag the curve up to add more
of that channel color; pull it down to subtract it. Adding Figure 9-18 The resulting Curves tool adjustment made
multiple points to a curve allows for independent to our Kodachrome film scan.
adjustment of shadows, midtones and highlights within
a color channel, i.e. adding green to the shadows while
taking away green in the highlights.

210
 c h a p t e r 9 : D i g i ta l To n e R e p r o d u c t i o n

Compare the color-corrected version against the

original. Leaving the adjustment layers set to a Normal
blend mode means that the changes affect both color and
tone. The original and final versions should show notably
improved color and tone reproduction. Working with
raw image data is often easier than film scans but can be
equally challenging when scenes contain multiple light
sources or brightly colored and reflective surfaces. It’s here
that using Curves adjustments and carefully considering
shadows, midtones and highlights independent of one
another that the best results can be achieved. Of course,
having a ColorChecker in the scene from which to set
sample points is the ideal scenario to aid in manual color
correction.

Note
1 McCamy, C.S., H. Marcus and J.G. Davidson. “A color
rendition chart.” Journal of Applied Photographic Engineering,
vol. 2, no. 3, 1976.

211
10 Software
Filters

Photograph by Rochester Institute of Technology alumnus and Lecturer Dan Hughes

This chapter is an overview of the digital filters we apply to images after capture, during
our post-processing workflow. In Chapter 5, we discussed using optical filters like neutral
density and UV filters placed in front of our lens when capturing an image. Optical filters
dictate characteristics of the image recorded relative to the scene itself. Software filters
take the digital information recorded and work to modify its characteristics for different
visual effects. Software filters allow photographers to express their creativity in ways
never possible with film, offering opportunities to enhance the raw image data in both
broad strokes and with extreme, pixel-level precision when desired. Thoughtful and
skillful use of filters on image files elevates a photographer with a good eye to a pho-
tographer that produces amazing photographs. The software filters or image processing

213
 s ecti o n 2: P roc ess

filter techniques discussed are sharpening, blurring, blend- the term “illusion” carefully as not to imply falsehood or
ing and black and white conversion. We’ll conduct some trickery: sharpening plays to the behavior of our visual
applied sleuthing at the end of the chapter to reverse-engi- system. You’re going to really stress your good qualities
neer a combination of edits made by a mobile app image during a job interview while glossing over the dubious
processing filter. details of your work history. It’s not inherently malicious or
misleading, it’s making the best of what you’ve got. The
same is true for sharpening image data. We perceive sharp-
Sharpening an Image ness and detail when we have local contrast to delineate
where one structure begins and another ends. The deline-
Sharpening filters enhance the appearance of sharpness ation between a building edge and the blue sky behind it
or clarity in images. That is, they improve acutance or should be clear-cut and distinct: the pixels representing the
how quickly the edges in an image transition from one building are one tone and color while the blue sky pixels
contrasting tone to another. Before you get the idea are another (see Figure 10-1). There shouldn’t be blue
that sharpening filters are miracle workers, know that no pixels breaching that building edge if the image is sharply
amount of post-processing sharpening will save a blurry resolved.
or out-of-focus capture. Excessively blurry or out-of-focus Sometimes even a well-recorded edge isn’t enough
frames are not restorable despite what crime scene inves- to appear as sharp as we’d like in the final, edited version.
tigation television shows may have us believe. A subject Here we have the option to exaggerate edges by creat-
edge recorded as a blurry smear cannot be reconstructed ing extra contrast, well beyond what is recorded by the
as an edge. Additionally, we closely associate sharpness sensor. Figure 10-1 shows what this looks like in practice.
with optical resolution whereby the camera resolved fine We overshoot edges by artificially darkening a darker edge
detail. No post-processing filter exists that can restore a and lightening the lighter side of the edge. The transitions
combined lack of resolved detail and acutance. become more extreme or abrupt as a result. Despite it look-
Sharpening filters follow mathematical rules and behav- ing exaggerated and unnatural up close, this strategy can
iors. Don’t let that lull you into a false sense of security. Both contribute to a sense of sharpness at the proper viewing
under- and oversharpening are very real possibilities in distance. The risk of overshooting edge contrast is in creat-
post-processing and being at either end of this spectrum ing halo artifacts. Like the name implies, this image quality
is detrimental to the perceived quality of your images. degradation looks like bright rings or outlines around sub-
Knowing when, where and by how much to sharpen is an ject edges. Avoid sharpening to the point where obvious
acquired skill empowered by a strong understanding of the haloing appears.
way in which sharpening filters work. There are three points in a photographic workflow
Sharpening software filters locate the edges in your where sharpening should occur: directly following camera
scene and increase the contrast along those edges to capture, during creative enhancement and immediately
create the illusion of additional image sharpness. We use preceding print or screen output.

Figure 10-1 A simplified illustration of a building edge against a blue sky. On the right, the image is sharply focused and the boundaries are
distinct. The center image is blurred and the building edges appear to transition from the sky over multiple pixels. In the left image, the building is
oversharpened to show overshoot.

214
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Capture sharpening happens right after capturing the Capture sharpening is a global enhancement. Global
image or as the initial phase of demosaicing and process- enhancements, such as sharpening, are applied to the
ing. The need for capture sharpening is a result of the entire image: every pixel is evaluated and modified with
anti-aliasing filter placed in front of the digital sensor to the same rules. This does not mean that every pixel is
help reduce the appearance of stair-stepping along object affected the same way, only that the entirety of pixels is
edges and avoid moiré artifacts. The anti-aliasing filter considered. Since the anti-aliasing filter is installed over the
reduces overall image sharpness at capture, an acceptable entire area of the digital sensor on most cameras, global
compromise. Capture sharpening is applied in-camera capture sharpening is necessary.
when shooting directly to JPEG and is destructive (mean- Creative sharpening is a subjective strategy of enhanc-
ing that it cannot be reversed). Some camera firmware ing specific parts of the scene for emphasis. For example,
allows for user adjustments to the amount of JPEG cap- it is quite common to apply some creative sharpening to
ture sharpening, others do not. Capture sharpening for the eyes in portrait photographs. The eyes may be just as
raw images is available in image processing software and objectively sharp as every other detail in the image after
should be applied at an early stage of image ingestion or capture, yet we want to further sculpt and direct the atten-
basic adjustments. As with all raw editing adjustments, tion of the viewer. Creative sharpening is employed as a
it is non-destructive. It’s common for raw image editors local enhancement for this reason. A subset of pixels are
to default to a preset for capture sharpening because we targeted and filtered with specific intent and control. All
expect that the raw image data needs sharpness enhance- other pixels are left unmodified by the software filtering.
ment right out of the gate. You may find that modifying Creative sharpening is key to achieving the aesthetic you
this default capture sharpening filter brings your images up want for your image. There are no exact rules or instruc-
to a good baseline level of sharpness before moving on to tions to dictate the most effective use of this processing
other filters and adjustments in your workflow. An example stage as it is heavily context dependent and creatively
of capture sharpening settings is shown in Figure 10-2. open-ended.

Figure 10-2 Capture sharpening controls in Adobe Camera Raw introduce a base level of sharpness that is inherently lost at capture and
demosaicing.

215
 s ecti o n 2: P roc ess

Figure 10-3 Unsharp mask sharpening applied to an image without care to adjusting its
parameters can lead to extreme, exaggerated sharpening effects.

Let’s highlight three types of creative sharpening via becomes. The Radius parameter controls how many of the
software filters, though there are many others: unsharp neighboring pixels are affected. A larger radius means that
masking, smart sharpening and high pass sharpening. All the amount of sharpening extends to pixels farther away
three are strong options for creative sharpening, however, from the edge. A smaller radius enhances small details;
they can alternatively be used as global enhancements for a large radius impacts all details and possible edges. The
output sharpening. radius and amount settings affect each other: if we increase
Unsharp masking holds many bad memories for anyone the radius, the amount should be decreased to avoid gross
who studied film photography. The name comes from oversharpening. The third adjustable parameter of unsharp
creating an unsharp or blurry film negative to mask the mask filtering is the Threshold value. The threshold dictates
original image in the darkroom. It was a difficult and how different the adjacent pixel values must be in order
frustrating process. Unsharp masking in the digital world to be sharpened. Setting the threshold to 0 sharpens the
is, thankfully, much easier, though based on the same idea. entire image with no distinction between edge pixels
Taking a blurred (or “unsharp”) version of an image file and and non-edge pixels, making this a global adjustment
subtracting its pixel values from the original leaves behind (see Figure 10-3). This tends to over-emphasize image
the photograph’s edges. This subtracted, edges-only prod- noise. Increasing the threshold will leave low contrast
uct is used as a mask or guide to dictate where the edge areas (which are likely to be non-edge content) alone. We
contrast enhancement takes effect. typically adjust the radius first to get a sense of which areas
The unsharp mask filter commonly offers three adjust- are going to receive the sharpening, dial in the amount and
ment parameters. The Amount parameter controls how finally tweak the threshold to minimize oversharpening in
much contrast is increased between adjacent pixels. It is areas where we don’t want it.
closely tied to the overall perception of sharpening; it has Smart sharpening offers more control than unsharp
a direct and coarse relationship to how sharp the image masking. The options are similar, however smart sharpen

216
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Figure 10-4 The result of a high pass filtered image. This can be blended or overlaid with the
original to introduce edge sharpening.

allows the sharpening to be reduced or faded in the shad- Let’s mention a bonus software filter to round out
ows and highlights. This reduces halos and deemphasizes our creative sharpening options. Midtone sharpening also
noise that is always present in the shadows. Smart sharpen goes by names like clarity and structure depending on
has options to correct blur, motion, lens or Gaussian blur the image processing software used. This method targets
(we define Gaussian later in this chapter). Unsharp mask- midtone pixel edges exclusively rather than going after all
ing can only correct Gaussian blur. Lastly, if you are using edges across the shadows, midtones or highlights of the
smart sharpen in Photoshop, the settings can be saved as a scene. Midtone sharpening tools may use the same types
preset, which can speed up workflow in the future. of adjustable parameters (amount, radius, threshold) as
The last method we’ll cover is high pass filtering. While it the previously described filters. Content in the midtones
is not inherently a sharpening filter, many photographers is likely to be fine texture or small details that get a boost
find that employing it as such is one of the most effective when their edges receive additional contrast. Instead of
sharpening strategies. Recall that we create the illusion emphasizing building edges or large object boundaries,
of sharpness by increasing the contrast along the edges. midtone edge contrast enhances the perceived sharpness
Successful sharpening affects the edges only, leaving all of elements like skin and hair. This gives scenes a pop of
other areas in the image untouched. The high pass filter is perceived detail or clarity and can be combined with addi-
an edge detection filter and its output is shown in Figure tional, global sharpening.
10-4. When using Adobe Photoshop, high pass filter results Output sharpening is necessary because the final desti-
are combined with the image using the luminosity blend nations and output processes for your photographs affect
mode to sharpen edges without negatively affecting the their appearance. When printing, paper surfaces accept ink
rest of the image. The luminosity blend mode helps avoid differently. Newsprint, as an example medium, absorbs ink.
color shifts that can otherwise arise with the high pass An ink applied to newsprint spreads out as it is absorbed
sharpening method. into the material. This spreading means that distinct edges

217
 s ecti o n 2: P roc ess

begin to blur. If your photograph’s output destination the image’s appearance) makes this a blunt and limiting
is newsprint, the file must be prepared with significant strategy. Chapter 16 covers print media and the appro-
amounts of sharpening before it is printed. Photographic priate considerations for optimal photographic quality.
papers, on the other hand, are designed so that the ink Output sharpening takes practice and a familiarity with
does not spread significantly when absorbed into the your equipment to get great results.
paper. Output sharpening strategies for printing on photo
paper are driven by the paper surfaces, such as matte or
glossy, and the anticipated viewing distance (as a function Blurring an Image
of print size and venue). A highly reflective glossy paper
requires less sharpening than a diffuse reflecting surface The use of blur in a photograph starts with decisions at
such as matte paper. If your final destination is display, you capture. Blurring strategies include using aperture settings
may not need any additional sharpening at all. to limit depth of field and panning on a moving object to
Output sharpening is not tied to a specific filter or blur the background. Introducing blur in post-processing
approach, so the previously mentioned sharpening tools is another approach, either in combination with or com-
are valuable here. The difference is that output sharpen- pletely independent from capture decisions. Maybe we
ing is only intended as processing in anticipation of the want to minimize the visual weight of a distracting object
image’s final presentation on a medium with sharpness or simply blur the entire image to use it as a background.
considerations. Output sharpening is primarily global in Blur can also imply motion or a feeling of uncertainty and
nature. Image processing software typically includes the ambiguity.
option to sharpen for output in the stages immediately There are three common low pass filter approaches for
preceding printing. This means that we don’t necessarily creating blur with software processing: a mean filter, a
save the output-sharpened result for later; it’s treated as a weighted average filter or a Gaussian filter. The mean filter is
temporary, as-needed and in-context software filter. Also the simplest of the three. It takes a sliding window (typically
note that camera firmware offers sharpening options when 3x3 pixels) passed over the image and replaces the center
shooting directly to JPEG, typically only in vague scales like pixel with the average of all the pixels in that window.
“low, medium, high” as the presumption is that these JPEGs Expanding the window to 5x5 pixels or larger introduces
are skipping most of the processing pipeline. These JPEGs even more blurring. A mean filter can be applied to the
are suitable for printing, though the lack of fine control entire image or a locally masked area such as the scene
over the sharpening approach (and every other aspect of background. An example of a 3x3 average filter is shown in

Figure 10-5 A sample averaging filter.

218
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Figure 10-5. A mean filter must be odd ordered (3x3, 5x5, Blending Images
9x9 and so on), all of the filter elements must be the same
and sum to 1. To have the elements sum to 1 using the Blending pixels from two or more sources is commonplace
custom filter tool in Adobe Photoshop, we need a scale in photographic compositing and retouching. Consider
value equal to the sum of the filter elements. In Figure 10-5, image files as two-dimensional arrays of pixel data that can
the sum of the elements is 9, so we have to divide our filter be stacked or layered together. The next step in working
by 1/9. This means using a scale factor of 9. with these layers is deciding on the rules dictating their
The weighted average filter places greater importance interactions. Simply placing one image on top of another
on the center pixel value relative to the surrounding will completely obscure the bottom-most image like milk
pixels. This blurring technique offers more control over the foam layered on top of a cappuccino. We blend elements
amount of blur introduced. A weighted average filter must of both image layers to achieve some constructive creative
meet three conditions. The filter must be odd-ordered (3x3, output from the exercise, otherwise we simply see the
5x5, 9x9), all filter elements surrounding the center must be top-most image. Blend modes or blending interaction rules
1 and the center value must be larger than 1 which gives it use the numeric representation of pixels—their luminance
the weight or greater influence on the image. In order for values between 0 and 255 (across three color channels)—
the elements to sum to 1 using Adobe Photoshop’s custom to establish predictable outcomes from their combination.
filter tool, we need a scale factor equal to the sum of all the Common blend interaction rules are as follows1:
elements. The sum of the elements in Figure 10-6 is 13, so
we divide our filter by 1/13. In Photoshop, this makes our 1. Darkening
scale factor 13. A darkening blend mode always prioritizes the darkest
Lastly, we have the Gaussian blur filter, also referred to as pixel at a location of the blended (or active) image layer
Gaussian smoothing. Recall from math class that a Gauss- over pixels on layers below. Simply stated, this blend
ian distribution is a normal or bell-shaped curve. Jogging behavior takes the darkest (lowest luminance value)
your memory further, a normal curve can have different pixel at a given X,Y coordinate and has the effect of
widths or standard deviations. A Gaussian blur filter takes darkening at least part of the image relative to where
on the shape of a two-dimensional gaussian curve; it’s used it started. Some forms of darkening blending multiply
to reduce the presence of noise and smooth details. The pixel values of the blended layers to create new pixel
larger the kernel, the larger the deviation of pixels blurred values. See Figure 10-7 (overleaf) for an example of
in the filtered image. blending using a darkening interaction rule; notice how

Figure 10-6 A sample weighted averaging filter.

219
 s ecti o n 2: P roc ess

the white background of the plant image is treated for example, takes the saturation component of the
differently based on the blending strategy. active layer pixels and applies it to the luminance and
2. Lightening hue of the image layer beneath.
A lightening blend mode is the opposite of a darkening
blend mode whereby the brightest pixel (highest lumi- We’ve framed the concepts of blending filters in terms
nance value) across stacked image layers is used. Some of two unique image layers. This is just one of many possi-
forms of lightening multiply pixel values of the blended ble ways to employ blend modes or filters; entire groups of
layers to create new pixel values. Figure 10-7 also shows image layers (and non-photo pixels, like design elements)
an example of this interaction. or small subsets of images can be combined when doing
3. Contrast heavy photographic compositing. Blending pixels from two
A contrast blend mode is a mix of both the darkening image layers is just the tip of the iceberg. The previous list
and lightening blend modes that introduces additional covers the five categories for blend mode interactions—
contrast by examining each pixel. If the pixel is brighter there are often multiple variations within these categories.
than 50% or midtone gray, it is lightened. Conversely, if Finally, working with degrees of image layer opacity adds
it is darker than midtone gray, the pixel is darkened. additional complexity to pixel interactions. Blending image
4. Compare elements is both an art and a science and we’re often
Comparison blend modes compares pixel values between encouraged by the results of experimenting. Naturally,
active and inactive image layers and performs different Adobe Photoshop professionals have considerable docu-
mathematical evaluations like subtraction or division to mentation and insight on this topic and we point you to
derive new pixel values. These blend modes are particu- them for further information.2
larly difficult to anticipate or previsualize because the
behavior is not as visually intuitive as lighten or darken
blend modes.
5. Component
Component blend modes separately change hue, satura-
tion, color and/or luminosity. A saturation blend mode,

Figure 10-7 The left two images are stacked and blended using a darkening blend mode to create the next image. The rightmost image is
blended using a lightening blend mode.

220
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Black and White Conversion Let’s refresh our memory regarding color filters and their
functions. A color filter is named for the color it passes, for
One might say that we are returning to our roots where example a red filter passes red light and to a lesser extent
photography began by taking color images and convert- yellow and magenta while blocking cyan, green and blue.
ing them to black and white. There are three options for Analog photographers routinely use color filters when
creating black and white images in the digital era. The
first option is to set your camera to monochrome mode,
available in many DSLRs’ camera firmware, though this
only applies to JPEG shooting. Raw image files will always
retain all of the color information even if the camera
generates a black and white preview. A second option is
to convert the digital color image to black and white after
capture with software filters. The third option, which we
won’t expound on here, is using a dedicated, feature-built
monochrome-only camera. Such cameras are offered sans
color filter array for shooting exclusively in black and white.
Their niche functionality tends to make them expensive
and uncommon equipment. Figure 10-8 Black and white conversion controls in Adobe Photoshop
Lightroom.

Figure 10-9 A color image and two unique black and white conversions made by independently adjusting the brightnesses of yellows and reds.

221
 s ecti o n 2: P roc ess

shooting with black and white film. Landscape photogra- Digital Filtering for Special Effect
phy using black and white film benefits from using a yellow
filter to darken bright skies, enhancing the separation With the ever-increasing processing power of chips in all
between the sky and white clouds. On-camera black and cameras, from smartphones to high-end DSLRs, comes the
white modes simulate this use of color filters. Removing ability to introduce in-camera digital filtering. These filters
color from the process allows the artist to concentrate include special effects like grainy black and white, soft
on composition and lighting. For complete control of the focus, fish-eye distortion, watercolor painting and minia-
tonality rendered in a black and white image, most profes- turization. All of these can be applied directly to images as
sionals and experienced editors opt for conversion later in they are captured or when reviewing them on the camera’s
post-processing. preview screen. They require no extra training or separate,
Converting a color image to black and white in editing off-camera software. The effects can range from a simple
software allows for much more creativity and flexibility. black and white conversion to abstracted interpretations.
Editing software provides preset options as well as the Moving to desktop software offers additional creative
ability to adjust individual color brightnesses as shown options; special effect filters available in Adobe Photoshop
in Figure 10-8. This provides endless options for black or DxO’s Nik Collection are seemingly endless. It’s possible
and white conversion from a full color image. Figure 10-9 to add a realistic soft-focus like the one shown in Figure
highlights an example of two very different black and 10-10 or simulate different media ranging from charcoal
white interpretations relative to the original by increasing drawings to stained glass (see Figure 10-11). Some special
the tone brightness of some color pixels and darkening effect filters offer a host of tweakable sliders and settings to
others. finely tune the outcome. Some are kitschy, one-trick-pony
affairs while others are valuable aesthetic options when
used in the right context.

Figure 10-10 DxO’s Nik Collection soft focus filter.

222
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Figure 10-11 Adobe Photoshop’s stained glass filter.

Filter Masking selection tool identifies a local set of pixels for filtering or
other modifications; we never see the reference grayscale
We described the difference between global and local file used to define them and the software stores it in tem-
enhancements earlier in this chapter. Global enhancements porary memory only.
are easy to explain because we can increase the brightness A robust selection strategy for local filtering and pro-
value of every single pixel in an image, as one example. cessing includes using an alpha channel layer called a layer
There are no coordinates or other complex instructions mask. Photoshop displays this mask as a black and white
needed to grasp which pixels change when applying the layer of pixels, though it is not a visible set of pixels in the
adjustment or filter. Local adjustments do require this final image. Instead, the mask information dictates the
specificity, so how can we define selections of pixels with degree of “on” or “off” a filter or adjustment exhibits on the
minimal impact to file size, processing requirements and image pixels. Think of a layer mask like a dimmer switch
do so in a non-destructive manner? controlling a lamp. In addition to being switched off, the
Making selections in software such as Photoshop is dimmer allows for a set of increasingly brighter settings
a dense topic because it offers a variety of ways to do before reaching maximum brightness. Working with a layer
so. Some selections are temporary while others can be mask means modifying a grayscale file that defines active
recorded and called up at any time, even in a later editing and inactive areas for local image adjustments. On one end
session. Under the hood, though, all local selections use of the grayscale, black pixels indicate completely inactive
the same central programming concept: the selections areas. Combined with a filter or adjustment, these are con-
are grayscale bitmap or vector files used to designate sidered “off” and the original image pixels are unaffected.
active and inactive areas of the image frame. A temporary White pixels, on the other end of the grayscale, are fully

223
 s ecti o n 2: P roc ess

Figure 10-12 Masking image adjustments non-destructively is accomplished with an overlaid but invisible mapping of which pixels are and
which are not modified by the filter.

“on” and these areas receive the filter or adjustment at full by white areas of the mask will receive the brightness
strength. Everything in between represents a gradient of adjustment. All areas in black will remain the original
strength or influence that a local adjustment has over the brightness. The mask’s instructions at specific pixel loca-
underlying image pixels. tions are a form of pixel addressing using X,Y coordinates
Since Adobe Photoshop and similar software offerings in our two-dimensional image frame. Pixel addressing is a
use this method of local selections, expert-level image software filter strategy that allows for local, isolated pixel
editors take full advantage of their tools by treating layer changes.
masks as if they are the images themselves. The grayscale Alpha channel masks do add to file size. They’re like
mask histogram can be modified to add contrast, its overall additional image layers, only in shades of gray and often
definition can be blurred or sharpened and local changes with less detail than a continuous tone photograph. They
can be made with paintbrush tools. Furthermore, a photo- can be either 8-bit or 16-bits, the latter offering much more
graph can be converted to grayscale and made into a layer granular control over mask opacity via the extended gray-
mask itself—a powerful strategy for complex selections, scale palette. Expect them to add to your overall file sizes
especially when the image has color or brightness contrast when saving working-copy documents like PSDs. Lever-
along subject edges. aging editable selections for future use makes masking a
Layer masking is ideal for software filters and any image flexible post-processing skill.
adjustment in general. Our intention in Figure 10-12 is to
lighten the eyes in a portrait. The filter mask we defined
acts as instructions: only the pixel positions indicated

224
 c h a p t e r 10 : S o f t w a r e F i lt e r s

Proof of Concept: Reproducing an the app processing of our choosing and compare it to the
Aesthetic Filter Process output (see Figure 10-13). This renders a processed version
of the original that we like and would love to have as an
Effective implementations of software filters for quick and image processing preset in Adobe Photoshop Lightroom
effective image processing can be found in popular social for future use.
media apps like Snapchat and Instagram. Catering to a Ending our investigation here provides limited insight.
different audience, perhaps, their mobile apps include What does this filter do with colors and tones not present
advanced image processing filters that turn a standard in this single example image? It’s best to use an image
image into something with serious visual punch without target with known colors and tones that cover all possible
requiring users’ time or expertise. The challenge with apps’ subject content. Our deduction skills are greatly aided by
look-and-feel filters is that their inner workings are locked an ability to directly compare and contrast, so we build a
away in proprietary, special sauce source code. Photogra- reference target and a Photoshop document to drop in
phers may enjoy using a particular filter without knowing the before and after versions (see Figures 10-14 and 10-15,
how to reproduce it with traditional image processing tools overleaf).
on the desktop. We set out to explore a method of decon- Our target includes useful patches including skin tones
structing and emulating existing filters so that they can and highly saturated colors. It also uses a tone gradient to
become a regular part of our photographic workflow. help identify contrast and adjustments like white point or
Extensive research and computer scientists are regularly black point changes. We encourage you to make a custom
finding new ways to truly reverse-engineer image process- target to your liking that includes similar elements to make
ing steps with nothing more than the final result using the reverse-engineering easier. The comparison tool is
deep convolutional neural networks.3 Our approach here is simply an Adobe Photoshop document with the baseline,
relatively simple: take a known input image, feed it through unprocessed target image and a layer mask that cuts out

Figure 10-13 A photograph before and after the mobile app processing. We like the overall effect and recognize a few key trends in what was
done, but the more precise we can reproduce it, the better.

225
 s ecti o n 2: P roc ess

Figure 10-14 A custom reference target before and after the smartphone app filter.

Figure 10-15 A layer mask applied to the original target image allows us to insert the processed version as a layer
underneath. The mask allows the patches from both versions to appear side by side for easy comparison.

226
 c h a p t e r 10 : S o f t w a r e F i lt e r s

half of every patch. The layer mask hides half of the patch • Spend time in the HSL panel using the Targeted Adjust-
and leaves a transparency under which we insert the pro- ment Tool to select specific color patches and shift their
cessed version of the target as its own layer. This offers a hue, saturation, luminance or any combination therein.
split before and after comparison of the filter effects. Some of these filters make extremely targeted process-
Notice how Figure 10-15 reveals key characteristics of ing adjustments.
the app filter right away. The white patches in the corners • Make use of the Compare view mode in the Library
get much darker and yet the lighter neutral patches near module of Adobe Photoshop Lightroom to put your
the center do not darken to the same degree, indicating a preset head to head with the real app filtered version.
vignetting effect was introduced. The solid tone patches
now show a textured noise, almost like film grain, where
they were previously uniform. Color and saturation get Notes
trickier. However, using the Eyedropper tool reveals RGB 1 Eismann, Katrin, et al. Photoshop Masking & Compositing. 2nd
and L*a*b* pixel values that Chapter 13 helps make sense ed., New Riders, 2013.
of. Overall saturation appears lower and the greens and 2 Kost, Julieanne. “Adobe Photoshop CS blend mode magic.”
blues in particular are almost completely blanched. Pay Julieanne Kost Principle Digital Imaging Evangelist, Adobe,
jkost.com/pdf/photoshop/past_events/Blendmodemagic.
close attention to trends in brightness as well as changes
pdf.
to shadows relative to highlights (like different color
3 Hu, Yuanming, et al. “Exposure.” ACM Transactions on
casts).
Graphics, vol. 37, no. 2, 2018, pp. 1–17., doi:10.1145/3181974.
Try creating an Adobe Lightroom Edit preset that closely
emulates the app filter effect on a photograph to see
how close you can get using the raw image editing tools.
These filters tend to make global adjustments to tone
and contrast but very specific tweaks to color. Periodically
take a version and drop it into the Photoshop comparison
document. Swap out the original, unprocessed target with
the app filtered image and compare it against your Adobe
Lightroom interpretation. As long as we’ve got the ability
to send an original through the black box image process-
ing, we have this capability to compare and match results.

Strategies and tips:

• Use a comparison PSD file tool. The top half of each

patch shows the “original” and the bottom half is the
“processed.” Look closely for changes in tone, color and
texture.
• Create your preset using the color target before trying it
with an example photograph, as the target patches pro-
vide a much clearer picture as to the types of changes
going on.
• Do you see evidence of global or selective sharpening?
What about blurring? How might you replicate these
types of edits?

227
11 Camera
Characterization
and Image
Quality

Photograph by Rochester Institute of Technology photography alumnus Dave Kelby

229
 s ecti o n 2: P roc ess

Image quality matters to every photographer. This chapter and creates a usable photographic image or video file
reviews the pillars of image quality and camera character- as output. The ISP can be a dedicated piece of hardware
ization to better understand the results that we get—or known as a system on a chip or SoC. The path from light to
fail to get, as the case may be—from a camera. The state image file displayed flows from scene to optics to ISP to
of high-quality photographic equipment today is one storage. The pipeline includes the following steps (with
of abundance, yet every photographer has moments of variations depending on camera hardware, shooting mode
frustration or disappointment when they see a flawed and user settings):
image on their screen. Characterizing an imaging system is
a means of identifying its capabilities as well as its short- • Automatic exposure
comings. We start by providing an overview of an imaging • Automatic focus (including face or object recognition)
system pipeline and then head straight into the many • Automatic white balance
facets of quality as a function of appearance, technical • Correcting sensor defects such as stuck pixels and
accuracy and usability for photographers’ image-mak- non-uniformity
ing practices. These facets include noise, dynamic range, • Noise reduction
non-uniformities, sharpness and color. • Demosaicing
Jonathan B. Phillips and Henrik Eliasson define image • Gamma correction
quality in their book, Camera Image Quality Benchmarking • Correcting lens imperfections such as vignetting or
as “the perceived quality of an image under a particular distortions
viewing condition as determined by the settings and • Bracketed exposure HDR processing
properties of the input and output imaging systems • Filter effects or image processing styles
that ultimately influences a person’s value judgement of • Global and local tone reproduction adjustments
the image.”1 We are partial to this definition because it • Color space conversion
encapsulates the many variables that impact our criteria • Compression, encoding and storage
for good image quality. It ultimately comes down to an
individual’s assessment and the standards set by one’s use Figure 11-1 maps out a generic flow diagram for an imag-
of the imagery. The judgment of both the photographer ing pipeline. There are some steps that circuitously repeat
and the viewer are critical but not necessarily based on during capture such as auto focus; other steps may be
objective metrics. Modern photographers must keep one omitted based on how the imaging system is setup before
foot in the technical realm and the other in the subjective capture. Notice how the raw image bypasses many of the
art of recording light. steps that interpret and potentially improve image quality
shortcomings. These steps are instead offloaded to the
post-processing stage of a photographer’s workflow and
The Imaging Pipeline are typically conducted by raw image editor software.

The earliest stages of a digital photograph’s life include

critical decisions, calculations and conversions in the imag- The Many Tasks of the Image Signal
ing pipeline that dictate aspects of image captures before Processor
and after they’re made. The image signal processor (ISP)
is a dedicated computer chip in a camera that processes The ISP’s primary jobs are demosaicing the data and man-
live preview video feeds, handles autofocus, exposure and aging the automatic features that work in a feedback loop
other essential tasks. It is sometimes alternatively termed to set sharp focus, proper exposure and accurate color.
the digital signal processor or image processing engine. It These are known as the 3A’s: auto focus, auto exposure
takes the raw light signal detected by the sensor as input and auto white balance (AF, AE and AWB). These are highly

230
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

active behaviors, particularly when photographing events

or environments with rapidly changing lighting, subject
distance and camera movement. In smartphone camera
systems where automatic functions and ready-to-share
image assets are desired immediately after capture, the ISP
can be additionally responsible for face detection, red-
eye reduction and other output processing steps. The ISP
constantly calculates statistics from the incoming signal to
compute informative information about brightness, color
and sharpness just to name a few.
Photographers generally consider the raw file to be the
purest, no-processing-added form of image data in the
life of a photograph. This is true from a post-processing
perspective because raw files stop short of making per-
manent, irreversible editing decisions. However, raw files
are truly minimally processed image data given that some
decisions are made by the ISP before saving out the data to
file. Decisions at the earliest stages of a digital image’s life
must be made immediately, including signal amplification,
analog to digital conversion and automatic estimations for
white balance. Truthfully, most photographers shouldn’t
get in the weeds with this degree of signal processing and
are better off picking up where the basic image processing
pipeline ends: when we’ve got a viewable, full color expres-
sion of minimally processed data.
The ISP has work to do following image capture even if
the output is a raw image. A raw file includes an embed-
ded, processed JPEG for on-camera preview and for
desktop file browsers. The details of the JPEG processing
chain depend on the camera and any previously estab-
lished user settings (compression quality, noise reduction,
sharpening, filter effects). Pipelines designed for consumer
and professional cameras alike aim to match human
perception.

ISP Task: Auto White Balance

Figure 11-1 The imaging pipeline from scene to image file. Automatic white balancing algorithms additionally
evaluate the signal strength of each channel and apply
analog signal gain to the signals that it deems necessary
to boost. If it assumes a gray-world, that is, that every scene
has an average midtone reflectance, then its job is simply
to apply a fixed gain to the channels. Complex automatic

231
 s ecti o n 2: P roc ess

white balancing algorithms rely on deeper data analysis ISP Task: Demosaicing
or estimated guesses about subject content and lighting Demosaicing occurs next and then global adjustments
conditions from which to draw conclusions about the most can be made. We outlined the process of demosaicing and
accurate relationships between red, green and blue signals. interpolation in Chapter 7. Many DSLR cameras offer some
This can be done prior to the analog to digital conversion control over key parameters such as sharpness, contrast,
to minimize noise but is frequently also adjusted following saturation and color tone. These adjustments are applied
A/D conversion. to the image prior to saving the JPEG image. If the camera
has picture style presets to choose from such as landscape,
portrait, neutral or faithful, the selected style adjustments
ISP Task: Black Level Subtraction are applied at this stage. These interpretations of the raw
A form of noise reduction also happens prior to dem- image data are non-destructive and reversible on the raw
osaicing. Here the image black point is set. Black level file but baked into its preview JPEG until a new one is ren-
subtraction determines an offset for every photosite so dered and saved by desktop software. Demosaicing may
that when zero light is received, zero signal is recorded. also leverage insight into stuck pixels to interpolate around
Noise reduction is then applied. Aggressive noise reduc- them.
tion can reduce image sharpness and detail rendered by
the optics. Noise will always be present in every image due
to the nature of electronics and its forms are discussed ISP Task: Compressing and Saving
later in this chapter. The system risks perpetuating and When a capture is made and the earlier tasks are com-
potentially exacerbating noise if noise reduction is left pleted, it follows that the image data must be saved to
until after demosaicing. memory. Raw files are created and written to solid state
storage or sent over a data connection to another form of
storage; they contain the mosaiced image data along with
ISP Task: Stuck Pixels metadata from camera firmware and the ISP. This includes
Just like with television or phone screens, it’s not uncom- exposure information and timestamps. Writing JPEG image
mon for the occasional photosite in an array of many files requires applying a JPEG compression algorithm and
millions to get “stuck” or suffer from manufacturing defects. output sharpening. JPEG quality must be set prior to cap-
In a screen, this appears as a stuck or dead pixel. In a pho- ture in the camera firmware. More recently, cameras offer
tograph, a sensor’s equivalent is a single pixel with a stuck raw file variants that feature lossy compression or reduced
value that never changes from image to image (it is not spatial resolution. Data corruption or loss is likely if this step
necessarily black). This is a sensor defect. Fortunately, cam- is interrupted or the RAM buffer fills before the images can
eras can evaluate their sensors for these shortcoming and be permanently saved. The processor speed combined
mask or hide them before writing photosite data to the raw with the available RAM buffer can dictate the volume and
file. Another solution is to map and document the coordi- duration of rapid-fire burst exposures it can sustainably
nates of these photosites as instructions for the demosaic process and store.
algorithms to specifically interpolate from neighboring
pixels so as not to perpetuate the junk pixel data. The larger
the image sensor, the more opportunities for flaws during ISP Task: Video Recording
manufacturing, which in part contributes to higher price The ISP also handles video capture in imaging systems that
tags for larger sensors. support it. Video capture is a computationally and memory
intensive task to maintain throughput at frame rates of 24
frames per second or higher. The ISP will typically subsam-
ple the image sensor to determine the 3A’s and to provide

232
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

the user with a live preview stream. The ISP can also handle A photographic image with a linear response would
video stabilization and compressing the video stream with appear unnatural. Gamma correction is applied to all pho-
a specified codec, though encoding may be handled by a tographic images in the processing pipeline as a result. The
dedicated graphics processing unit (GPU) once handed off image data may be converted to a different color space
by the ISP. such at CIELAB before gamma correction. Around the same
time as linearization is some form of tone mapping to
optimize the image data in a subjectively pleasing way. The
Characterizing a Camera tone mapping is designed primarily for JPEG assets as they
are considered a final output version. Tone mapping and
Understanding the flow of light information and the many tone reproduction at both the global and local levels were
tasks handled by the ISP fosters an appreciation for how discussed in Chapter 9.
many facets of hardware and software work together to The result of one method for testing sensor linearity is
function as an imaging system. Characterizing a camera shown in Figure 11-2. Gamma correction must be turned
means different things to different people; here, we are off. Begin by setting up a ColorChecker target with even
interested in determining the performance of an imaging illumination. Focus and frame to fill the camera field of
system as identified through testing. This can be done view with the target. Capture a raw exposure series from
with expensive targets, lighting and software, or, alterna- very underexposed to very overexposed to obtain the
tively, with a ColorChecker, some affordable or printed-out entire range of digital counts. DCRaw is free software that
targets and your image editing software of choice. Charac- can convert the raw files to 16-bit linear TIFFs. Record and
terizing is about knowing the capabilities and limits of your plot the pixel values for the white, midtone gray and black
equipment so that you know what to use when and to patches against the captured exposure times. These plots
pave the way for effective post-processing. The rest of this help to indicate if the sensor behaves in a linear fashion:
chapter highlights key areas of camera performance that changes in exposure time should closely match the change
photographers should explore with their hardware. in recorded pixel values. A linear regression was fitted to

Testing Sensor Linearity

CCD and CMOS image sensors provide a linear output.

Simply put, this means that doubling the number of
photons collected by a photosite also doubles the output
pixel brightness values. This characteristic is invaluable
when using a sensor for scientific, quantitative purposes.
It also allows for combining multiple exposures to extend
the dynamic range of a sensor for high dynamic range
scenes. However, the linear response of our image sensors
does not match the human visual system’s detection and
perception of light intensity.
Notably, the linearity of a sensor cannot hold at low light
levels where the signal can be overtaken by the noise pres-
ent in the camera electronics. This is further complicated in
Figure 11-2 A linearity check for the red channel. The data is taken
low light situations when higher ISO settings are used. The from the midtone gray patch on a ColorChecker. Each channel should
increase in gain amplifies both the signal and the noise. be examined separately.

233
 s ecti o n 2: P roc ess

the data so the R2, goodness of fit value, helps to determine gain capabilities and then extrapolating the data to
if the camera is performing linearly. higher ISOs. This often causes quality-degrading levels
of noise, though it may be acceptable when the alterna-
tive is entirely missing the shot. Some cameras only offer
Evaluating Bright and Low Light extended ISO settings when shooting JPEGs due to the
Performance additional firmware processing the image undergoes
in-camera. A good rule of thumb is to only resort to the
DSLRs and other technically inclined cameras offer a range extended ISO range when there is no other option (a
of selectable ISOs. They commonly range from ISO 100 or noisy photo may be better than none at all). Extended ISO
200 up to 6400, with most offering extended sensitivities as modes can also affect contrast, dynamic range, sharpness
low as 50 and as high as 409,600. Keep in mind that even or resolution.
though those numbers seem incomprehensibly huge, each Evaluating the performance of both bright and low light
doubling of the ISO value equates to a single stop of sen- performance of an imaging system means shooting across
sitivity, therefore ISO 50 to 409,600 is a 14-stop range. The the range of available ISO settings and lighting conditions
native ISO range tends to be around the 100–6400 bracket; while keeping a close eye on accurate exposure, sharpness
this range is where the manufacturer tested and evaluated and image noise.
image quality to be very good with minimal offensive noise
characteristics. The extended ISO settings are available
for more extreme lighting scenarios and photographic Categorizing Image Noise
exposure flexibility. They also come with the caveat that
the results exchange increasing sensitivity for decreasing It’s easy to bring to mind what noise sounds like when
image quality. referring to audio: it’s not something pleasant, desirable
Chapter 4 established that ISO is a function of signal or constructive. The same applies to photography’s flavor
amplification because the true, native sensitivity of an of noise; the less the better. Visually, image noise appears
image sensor is fixed. Though a range of 100–6400 is stand- as static-like pixel values that speckle areas that should
ard, the sensor’s inherent sensitivity likely falls somewhere be smooth, continuous tones and colors. A basic form
around ISO 100, give or take a stop. It is this native sensitiv- of image noise can be described as “salt and pepper” in
ity that theoretically offers some of the best performance appearance. Any variation from the signal we intend to
of the camera sensor. Evaluating performance here means capture is a form of noise. Recall that most image sensors
shooting in bright light conditions, either natural or with employ red, green and blue filters over individual photo-
artificial sources, and confirming proper exposure. Bright sites to record light. The image file that results after capture
light conditions represent a best-case scenario and allow and demosaicing contains luminance data across three
for setting lenses to moderate aperture openings and rela- color channels corresponding to this same red, green and
tively fast shutter speeds. blue set. The materials used to make the color filter array
Offering a lower-than-native ISO setting like ISO 50 over each photosite, combined with the inherent sensitiv-
requires some engineering creativity. When used, the ity of the photosites themselves, mean that the red, green
camera intentionally overexposes and pulls the exposure and blue signals are not equal when recording a scene
back in firmware. The consequence is decreased dynamic illuminated by white light even if it’s a perfectly gray scene.
range with less perceptible noise in shadows regions This is accounted for in the analog to digital conversion
relative to the native ISO exposure equivalent because the process whereby the photosites with lesser sensitivity (i.e.
tones are brought down. the blue filtered photosites) are amplified more than the
On the other end, extending the ISO is achieved by green filtered photosites. This is in an effort to reduce a
pushing the camera sensor to the high end of its signal color bias or color cast in the full color image. The lesser

234
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

There are different types: fixed pattern, dark, read and

photon noise. Visual noise shows variations in luminance
(brightness) and variations in chroma (color).

Fixed Pattern Noise

Figure 11-3 Luminance noise differences between red, green
and blue channels. The surround is a midtone gray, meaning that Fixed pattern noise is named for the way that it appears in
the subject reflected equivalent amounts of light across the three an image: as an ordered, structured pattern along rows or
wavelength bands.
columns in the image array (see Figure 11-4). If we photo-
graph a uniform source where each photosite receives the
sensitivity, followed by the signal’s relatively increased same amount of light, the resulting image should have the
amplification mean that the blue channel in a photograph same pixel value everywhere. That is not the case when
can exhibit a greater amount of noise than the other two fixed pattern noise is present. There are two causes of fixed
(seen in Figure 11-3). pattern noise, dark signal non-uniformity (DSNU) and pixel
It’s convenient to talk about noise as a singular, visual response non-uniformity (PRNU).
artifact in our photographs. However, that generalizes a DSNU is seen when taking an image in total darkness
collection of activities happening in our imaging system. (or with the lens cap on). We expect every pixel value to be

Figure 11-4 Fixed pattern noise seen in underexposed areas of an image. This photograph’s shadows are boosted to show the noise
characteristics.

235
 s ecti o n 2: P roc ess

zero, however, DSNU presents as an offset between zero random variations in color, as well as luminance because
and the actual pixel value. DSNU noise can be removed the filtered photosites are demosaiced and it’s difficult
by subtracting a dark image (one taken in total darkness to identify the true color signals that failed to record
with the shutter closed) from the captured image. A dark clearly. Read noise affects the dynamic range of a camera;
image may be taken in a fraction of a second before the reducing the read noise lowers the minimum number of
true exposure in order to provide the ISP with a baseline photons needed to detect a signal. This results in higher
reference point. Causes of dark noise are discussed in the dynamic range in the captured image. Special designs
next section. of the electronics can minimize read noise for low light
PRNU is due to each photosite’s slight variation in imaging applications.
response to the same amount of light. We expect each
photosite to detect and record the same digital count
when exposed to an equal number of photons, yet the Photon Noise
reality is not so exact. The variations that occur between Photon noise (also called shot noise) is brightness variations
photosites is due to an inherent heterogeneity in sensor recorded due to the randomness of photons hitting the
manufacturing; it can be corrected by applying an offset photosites. Imagine setting up four buckets side by side
and gain to each pixel. and dropping 100 ping-pong balls above them. A perfect
distribution of ping-pong balls would have 25 land in each
bucket. That is not likely to occur in reality, though. There
Dark Noise may be 24 balls in one bucket, 26 in another and 25 in
Dark noise (also called thermal noise) is the variation in the remaining two. Photons and the photosites detecting
digital count resulting from thermally generated electrons them exhibit this same characteristic. The flow of photons
within a pixel. Electrons accumulate over time and are from a light source over time follows a Poisson distribution.
recorded as signal by the photosite, ultimately introducing This fluctuation of photons results in random variance
noise in the image. The amount of dark noise present is noise in the image. There is no direct method for correcting
dependent on temperature: the lower the temperature, the this image quality artifact beyond averaging pixel values in
lesser the thermal noise. Some digital backs feature fans uniform areas.
to keep sensor temperatures low. In scientific applications
where low light levels are an issue, such as astrophotogra-
phy or microscopy, there are multiple techniques used Signal to Noise Ratio
to lower the sensor temperature to minimize dark noise There is always going to be a degree of noise, both lumi-
including heatsinks and liquid cooling. Longer exposure nance and chroma, in a photograph. Instead of considering
times causes dark noise to become more prominent as the noise in isolation, it’s measured based on the signal to noise
camera electronics heat up. ratio (SNR), that is, the amount of noise present relative to
the amount of signal. The measurement does not delineate
between the types or sources of noise as outlined earlier, it
Read Noise simply provides a measure of variance. Noise becomes less
Read noise is the most prevalent noise in our camera visually apparent and distracting with a very strong signal.
systems. It is generated by the camera’s electronics, A high value for SNR means there’s a lot of good image
particularly the analog to digital (A/D) converter. This information relative to the noise, while a low SNR value
noise is introduced into the image signal during the means the opposite: any useful signal recorded by the
readout processes. Read noise is highly amplified when camera is getting drowned out by random noise, obscur-
shooting at high ISOs and is therefore the type we often ing our subject. The base or native ISO of a camera sensor
see when shooting. It can appear as chroma noise, or is set by most manufacturers such that the black level is

236
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

suspended in an emulsion. The crystals are the image

building elements of a photograph just like photosites
are for a digital sensor, however, the crystals are randomly
distributed. The appearance of grain increases with higher
ISO film because the crystals are made larger for better
low light sensitivity. These larger crystals are more visually
apparent and resolved to the human eye at normal
Figure 11-5 The amount of noise in a raw image file as ISO increases,
sans noise reduction. viewing distances and look like static, speckled variations
of tone. Your eyes notice the texture just like you can feel
the difference between ultra-fine powdered sugar on
just above the point where noise becomes objectionable. funnel cake and coarse, granular sugar crystals sprinkled
The signal to noise ratio is measured in decibels (db) as on a muffin. Film grain truly is the observable grains of
it describes a form of noise power. The SNR of an image silver halide crystals. Digital imaging, on the other hand,
cannot be improved after capture (if “signal” is defined develops different types of noise as the gain or sensitivity
as the original signal we hoped to record), only made increases. It is different because it is not a reflection of
worse. Noise reduction algorithms are a valuable means any changes to the photosites and it is a uniquely digital
of minimizing the visual appearance of noise, though they artifact.
are effectively introducing blurring or interpolation to the There are post-processing algorithms available to
image data. In this way, noise reduction filters or processing minimize digital noise. Figure 11-6 (overleaf) shows
can further degrade detail if aggressively applied. Equation a high ISO capture with considerable luminance and
11.1 offers a basic SNR calculation assuming we can meas- chroma noise visible upon zooming in. Software controls
ure the mean pixel values and the standard deviation in a for lessening the appearance of both offer sliders dictat-
uniform portion of an image: ing noise reduction strength, thresholds or tolerances
for how much variation is considered noise and even
Mean
Signal to Noise Ratio = 20 x log (Eq. 11.1) edge-preserving filtering. The background area is greatly
Standard Deviation
improved from a noise reduction standpoint, though
Figure 11-5 shows the quality and appearance of noise aggressive reduction can easily smooth out true high
at three ISO settings when photographing a gray card. The frequency detail and leave the subject looking overly soft
SNR values are calculated below the samples. Note that and featureless.
larger SNR values mean higher signal and therefore lower There are occasions where digital image noise is
noise. There are other approaches to calculating signal to preferred, even if it’s not entirely desirable. Random
noise ratio taking other variables into account. We tend variations in color and luminance help to disguise
to use this simple equation to get a general sense of pixel posterization in smooth-toned areas like the sky at
variance in our images. sunset. Posterization is visually distracting, as the discrete
jumps in color or tone are distracting when the transi-
tion should be gradual and imperceptible. Adding a bit
Delineating Noise from Graininess of noise (or restraining from applying noise reduction)
The terms noise and grain are often used interchangea- makes posterization artifacts less obvious or degrading to
bly in describing image characteristics. This stems from the perceived image quality. We can avoid the possibility
analog film’s equivalent to digital photography’s image of posterization by capturing and editing image data at
noise, though they are not the same. All films have higher bit depths.
grain. Graininess occurs because film’s light recording
abilities come from microscopic, light-sensitive crystals

237
 s ecti o n 2: P roc ess

Figure 11-6 A photograph captured at ISO 12800. The close crops show before (top right) and after (bottom right) luminance and chroma
noise reduction.

Considering Sensor Dynamic Range brightest highlight with detail, that we aim to record.
Sensor dynamic range is the range of tones from a scene
Recall that bit depth dictates how many distinct that the camera actually records. Sensor dynamic range is
tones are available to describe brightness. Bit depth largely dependent on a sensor’s bit depth—its ability to
and dynamic range go hand in hand even though delineate and record a wide range of photon signals and
they are not precisely equivalent characteristics of a then convert that to digital values. A sensor with a small
camera system. Scene dynamic range is the range of or limited bit depth is unable to record a scene image
brightnesses present, from the deepest shadow to the with a large tonal range of incoming signal. Therefore,

238
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

we practically consider that bit depth dictates the poten- + + +

ITUHDR36 GrayWorld Chart (C) Imatest LLC

+ +

tial dynamic range of a sensor. 1 4

UHDR UHDR

The camera manufacturer states a sensor’s dynamic

range performance in an ideal situation and can therefore
5 10

be misleading. In real-world use, the lens adds flare, the

IR-blocking and anti-aliasing filters add blur and noise is
always present. These elements lessen the overall dynamic
range of the image formed on the sensor which effectively
lowers the recorded bit depth. Sensor dynamic range is
also a function of sensor design and sensitivity: a sensor
that can successfully record unique values for 150 photons 27 32

versus 160 photons, for example, can exhibit better low

33 36

light recording capabilities. We’ve made up these numbers ITUHDR36 GrayWorld Chart (C) Imatest LLC

but consider the principle; a sensor that is finely attuned to + + + + +

low amounts of signal but that can also handle huge del- Figure 11-7 A standard test chart from ISO 12233.
(Image courtesy of Imatest LLC)
uges of photons in bright light makes for a wide dynamic
range recording instrument.
Note the distinction between potential recorded bit other hand, image quality is more likely to be maintained
depth and file bit depth. Like a bag of potato chips, the bit or maximized through the image processing workflow if
depth supported by the file is not necessarily maximized higher bit depth information is preserved throughout. This
by the data stored within it. A bag of chips might be twice is particularly true any time that tone changes or remap-
the size relative to the volume of chips, a disappointing but pings are introduced.
necessary design that gives the fragile food an air buffer for Testing dynamic range is challenging. In a laboratory
transport. File bit depth simply tells us the supported range setting, a calibrated step wedge such as the ISO 12233
of tones that could be stored if they exist in the scene. standard test chart shown in Figure 11-7 can be used. It
Otherwise, that extra space is padded. Unlike a bag of features a set of neutral tone patches spanning a wide
potato chips, we have the power to fill our image file when tonal range. A measured system dynamic range is specific
needed. to a single lens, sensor and ISO setting combination and
Raw image files use between 12 and 16 bits to store the dynamic range achieved in a laboratory is nearly always
capture data (purely brightness information, not yet larger than what is achieved in the real world. Real-world
demosaiced into color channels). It’s impossible to distin- testing can involve a wide dynamic range scene and a
guish between a 12-bit or 16-bit file by eye, so we rely on spot meter to calculate the range (in stops) of light levels
software to tell us exactly what data we’ve got. An image successfully recorded with detail.
with 12–16 bits of information equates to somewhere
between 68.68 billion and 281 trillion possible colors. It is
estimated that humans can see approximately 10 million Identifying Non-Uniformities
colors.2 The additional headroom is critical for editing and
processing decisions to prevent artifacts like posterization There are multiple types of non-uniformity in an imaging
and banding. A numeric representation of brightness system and all have potential mitigations to minimize
and color, even if it’s too granular for our eyes to see, has degradations to image quality. Pixel response non-uniform-
implications for the mathematical operations applied to ity is a form of noise described earlier due to the variation
that information. The characteristic of bit depth by itself from photosite to photosite in its response to incoming
does not inherently indicate any measure of quality. On the photons. We ideally want all of the photosites to exhibit

239
 s ecti o n 2: P roc ess

near-identical responses and recordings of light across the

sensor array. Focus non-uniformity is any inconsistency in
how the optics form an image on the image plane (look
back to Figure 2-42 for one example). We typically desire
an optical system that provides a sharply resolved image
at the center as well as at the corners and everywhere in
between. Measuring a lens, described later in this chapter,
helps to identify focus non-uniformity. Luminance non-uni-
formity is an inconsistent light level across the image plane
due to one or more types of vignetting. Here again, we
ideally want to have an even rendering of a flat field. Finally,
color non-uniformity is an inconsistency in color recording
across the image area and is most commonly the result of
filtering and oblique angle light.
Optical vignetting is a result of light bending around the
aperture in the lens. The result is less light reaching the far
edges of the image plane. The degree of optical vignetting
is lessened when the aperture is wide open relative to
when it is stopped down. The smaller aperture opening
causes the light to bend more severely, introducing more
light fall-off. Mechanical vignetting is the result of a phys-
ical obstruction in the light path such as a lens hood or a
filter that is not large enough for the lens. Pixel vignetting is Figure 11-8 Lens vignetting with a 50mm prime lens at f/1.2 (top).
A simulation of color shading (bottom).
caused by the physical depth of each photosite. It is more
difficult for the light to reach the photosites at the edge of
the sensor. This image quality artifact is often compensated
for by the ISP by boosting the signal in local areas of the in post-processing software assuming it has a calibration
frame. database or can dynamically detect the artifact. An image
Both optical and pixel vignetting can be modeled by made with a white translucent material over the lens
imaging a flat field (a subject with uniform reflectance). produces an image that can be used as a reference for
This provides software with a roadmap of brightness correction.
non-uniformity from which to counteract. Once the effect A related artifact can be introduced by inaccurate or
is modeled, it can be digitally corrected. This must be done overcorrection by the ISP. This can show color shading,
for every lens used with a given camera and at every aper- either non-uniform or radial from the center of the frame
ture setting. outward. Such an artifact is effectively an image processing
Non-uniformity in color is another potential image error attempting to resolve an optical artifact; it can be
quality problem in an imaging system. Color non-uniform- considered an inverse-color shading artifact.
ity is also called lens cast or color shading (see Figure 11-8).
Lens cast appears as green in one corner of the image
and ends as a magenta shade in the opposite corner.
This results from light entering the lens and/or the color
filters atop the photosites at sharp angles. This can be
corrected in the same manner as vignetting in the ISP or

240
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

Evaluating Lens Quality: performance data at both 10 lines/mm and 30 lines/mm.

Sharpness and Detail Published manufacturer information also includes sagittal
and meridional MTF data. Measuring along these planes of
Photographic optics are a complex network of engineer- an optical system identifies if it suffers from astigmatism,
ing, tradeoffs and use-case considerations. Photographers an asymmetrical blurring due to misshapen lens elements.
may not find themselves designing a camera lens for The term is familiar to those of us with prescription glasses
themselves but they are likely to be critical consumers and as our eyes can also be astigmatic. The further apart the
craftsmen in search of the best tools. The following sec- sagittal and meridional MTF curves are, the more astigma-
tions outline how to evaluate lenses coupled with image tism present in the lens.
sensors for their image-making capabilities. The obvious question remains: what is a good mod-
ulation transfer function for a digital camera and lens
combination? The MTF at 10 lines/mm is usually better
Modulation Transfer Function than the 30 lines/mm, meaning that a lens exhibits better
The most common method of evaluating lens quality is contrast reproduction than sharpness performance.
measuring its modulation transfer function or MTF. The first Contrast measurements are generally broken up into cat-
time MTF was explained to us, the professor said that you egories of excellent (greater than 0.9), very good (0.7–0.9)
measure MTF to determine if you paid too much for your and average (0.5–0.7). A value less than 0.5 indicates a soft
lens! MTF measures the performance of a lens by its sharp- lens. The scale for resolution is subjective and based on
ness and contrast reproduction ability. In other words, it how the lens is used.
measures the amount of detail reproduced by a lens. Let’s
unpack exactly what that means in optical terms.
Optical systems including our cameras use spatial fre- Interpreting an MTF Plot
quency to evaluate and report quality, telling us how many Once targets are photographed, the results of an MTF test
cycles per millimeter a lens or the system can reproduce. are commonly presented in the form of modulation transfer
These cycles are alternating white and black lines whereby function curves in which the input spatial frequency is plot-
an image of such a pattern theoretically cycles between ted as cycles per millimeter on the horizontal axis against
maximum and minimum brightnesses (i.e. white and output contrast as percent modulation on the vertical axis.
black). A pattern of 10 lines per millimeter is used to judge A perfect reproduction would have image contrast equal
contrast performance and 30 lines per millimeter is used to to target contrast at all frequencies. This would look like a
judge resolution. straight horizontal line at 1.0 on an MTF plot. In practice,
We start by passing a perfect pattern through the the lines always slope downward to the right, since image
system and record the output. If the recorded pattern is contrast decreases as the spatial frequency increases.
an exact replica, the system exhibits a perfect 1:1 contrast Eventually the lines reach the baseline, representing zero
reproduction. If you ever buy a lens that performs this well, contrast, when the imaging system is no longer able to
any amount spent is worth it. No lens is perfect in the real detect the luminance variations in the test target. The
world, though. The reproduced pattern image has a differ- advantage of modulation transfer functions is that they
ent modulation and can be shifted. provide information about image quality over a range of
Seems simple enough, however there are many fre- frequencies rather than just at the limiting frequency as
quencies present in an image. Low frequency information, does resolving power; there are also disadvantages. For
the details in a scene, may show very little or no degrada- example, photographers cannot easily prepare their own
tion. High frequency information, the fine details, may be modulation transfer functions and the curves are more
severely impacted or not reproduced at all. To further com- difficult to interpret than a single resolving-power number.
plicate the issue, camera manufacturers often provide MTF MTFs are most useful for comparison purposes.

241
 s ecti o n 2: P roc ess

Figure 11-9 An MTF plot of a lens at f/16. Figure 11-10 A United States Air Force (USAF) Resolution Target.

Representative curves for a single lens and camera at of the two sometimes appears less sharp. In defense of
ISO 100 are shown in Figure 11-9. These MTF curves show resolving power, studies show that observers can be quite
that the lens tests exhibit some astigmatism as the sagittal consistent even if they don’t agree with other observers.
and meridional curves diverge away from the center of Consistency makes their judgments valid on a comparative
the lens. Recall that the 10 lines/millimeter data represents basis such as between on-axis and off-axis images, images
contrast performance and 30 lines/millimeter data repre- at two different f-numbers and images formed with two
sents resolution. different lenses. It’s appropriate to make comparisons
Furthermore, resolving power is not a reliable indicator between lenses on the basis of resolving power as long as
of the appearance of sharpness of photographic images. it’s understood that resolving power relates to the ability of
Resolving power is a widely used yet controversial method the lens to image fine detail rather than the overall quality
of checking image definition. The testing procedure is of the image.
simple. The lens is focused on a row of a resolution target
placed at a specified distance from the lens featuring alter-
nating light and dark stripes as shown in Figure 11-10. The Shooting to Evaluate Image Definition
separate targets are arranged so that the images fall on a Testing image definition requires the following:
diagonal line on the sensor plane with the center target
on the lens axis and oriented so that the mutually perpen- 1. The subject must conform to a flat surface that is perpen-
dicular sets of stripes constitute radial and tangential lines. dicular to the lens axis and parallel to the sensor plane.
Resolving power is measured as the maximum number of 2. The subject must exhibit good detail with local contrast
light-dark line pairs per millimeter that can be resolved. as high as is likely to be encountered and contain a
Critics of the resolving power metric note that different variety of colors.
observers may not agree on which is the smallest set of 3. The subject must be large enough to cover the angle of
stripes that can be resolved and that in comparing two view of the camera and lens combination at an average
lenses, photographs made with the higher resolving power object distance.

242
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

4. The subject must also be photographed at the closest lens, ISO and light source combination. In other words, a
object distance. camera profile is needed for every unique configuration
5. Photographs must be made at the maximum, minimum and shooting scenario. This is a realistic option for studio
and at least one intermediate aperture opening. photographers with a highly consistent setup—shooting
6. Care must be taken to place the optimal image focus at product shots with the same lens and lights everyday—
the sensor plane, which may require bracketing focus. and lens profiling software exists to serve this niche. Lens
7. Care must be taken to avoid camera movement, noting profiling can mitigate color inaccuracies as well as non-uni-
that mirror action in SLR cameras can cause blurring formities and optical distortions.
especially with small object distances or long focal
length lenses.
Calculating Color Difference
If the same subject is used to test for image shape, it Those of us with a fleet of lenses, multiple camera bodies,
must have straight parallel lines imaged near the edges of varying subject matter and organic lighting scenarios
the sensor to reveal pincushion or barrel distortion. If it’s need another option. We lean on our trusty ColorChecker
also to be used to test for uniformity of illumination at the to evaluate color differences between a subject and its
sensor plane, it must contain areas of uniform luminance reproduction. Setting up the ColorChecker with uniform
from center to edge. Tests for flare and ghost images illumination, setting a custom white balance and shooting
require other subject attributes. A practical test cannot a perfectly exposed image in both raw and JPEG formats
be overly simple if it’s going to provide useful informa- provides us the opportunity to evaluate how a camera
tion about the lens. Evaluation of the results is easier and system reproduces color. The ColorChecker patches have
more meaningful if a parallel test is conducted on a lens of known, published CIELAB values though we can also
known quality for comparison. measure them ourselves with a spectrophotometer. Using
Practical tests such as these are not truly lens tests. these and reading the color patch CIELAB pixel values
Instead they are systems tests that include subject, lens, from the raw image file allow us to calculate Delta Eab (or
camera, sensor, exposure, onboard processing and the ∆Eab ) differences. Delta Eab is a measurement of the differ-
output medium (display or print). The advantage that such ence between two samples (see Equation 11.2). A value of
tests have of being realistic must be weighed against the approximately 2.3 is considered a just noticeable difference
disadvantage that tests of a system reduce the effect of var- when using this metric.3 We can do the same with the
iations of one component, such as the lens, even if all the captured JPEG to observe the color reproduction choices
other factors remain exactly the same. made by the camera ISP’s JPEG processing. From this
exercise we can work to correct major color inaccuracies if
present.
Assessing Color Accuracy
∆Eab = √ (LR – LO )2 + (aR – aO )2 + (bR – bO )2 (Eq. 11.2)
Reproducing color accurately is a complex issue. A display
and a printer can be profiled and calibrated to ensure that Where
the system is doing the best it can do to match what we LO, aO and bO are the CIELAB values of the original target
see on the screen to what we see on the print. So what patches
part does the camera play? Veteran shooters often talk LR, aR and bR are the CIELAB values of the image
about one brand or model rendering colors to their tastes reproduction
more than a competitor, though this is highly subjective The ∆Eab metric presented here is just one possible ∆E
and largely anecdotal. Cameras are not typically profiled, as measurement to consider depending on the intent of your
we would need a profile specific to every camera sensor, color accuracy assessment. Delta Eab is a simple distance

243
 s ecti o n 2: P roc ess

measurement in the three-dimensional CIELAB color space. with the same signal strength in all three. Figure 11-11
The CIE recommends a ∆E00 (Delta E 2000) which is much illustrates this approach in action, starting with an image
more computationally rigorous and places heavier weight with an inaccurate white balance. Taking an average of
on luminance over chroma and hue values. all pixels shows a decidedly blue color cast that is cor-
rected to a neutral gray by changing the color channel’s
tone reproduction.4 Making this same adjustment to the
White Balance Accuracy image results in a properly white balanced scene. Try this
Chapter 4’s overview of photographic exposure param- with your own images to learn that this works in some
eters included the role of white balance. Consciously instances but does not work in many others.
setting the white balance before capture can help in Another strategy is to identify common patterns like the
accurately recording and representing color. Doing so distribution of tones in a classic landscape composition
means estimating or correctly determining the dominant and use an internal database to find the closest match.
light source type in the scene, a degree of human insight, Building this database can be the difference between a
logic and context that camera software alone cannot good AWB algorithm and one that regularly fails. Machine
reproduce. Our visual system also benefits from color learning and object recognition software is increasingly
constancy where colors appear correct no matter the leveraged to make decisions on white balance. For exam-
light source. Automatic white balance (AWB) algorithms ple, face detection can send automatic image processing
leverage the incoming light information detected by the and color reproduction down a path of prioritizing the
sensor photosites. One software strategy employs a gray accurate reproduction of skin tones. Consider how such
world assumption, the premise that all scenes and sub- artificial intelligence insights that mimic those conscious
jects are neutral and should reflect or absorb light by the and contextual decisions made by photographers might
same amount. If this premise is true, then there should improve AWB performance. A photograph featuring a
be equal signal received across all three channels of color close-up of red autumn leaves causes an over-compen-
image data (red, green and blue). If one of these channels sation in cyan when employing a gray world assumption
records less light, signal gain is introduced to boost it up algorithm. A system that interprets pixel patterns and
to the level of the other two. This behavior is true for all recognizes the shapes and colors as leaves or recognizes
three channels; the goal of the algorithm is to end up the broad strokes of an outdoor landscape, and not simply

Figure 11-11 Averaging all pixels in this image reveals a blue cast. Correcting this averaged version to be neutral provides a channel
adjustment applicable to the original image.

244
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

a collection of mostly red pixels, might instead choose to

settle on a warmer white balance. Shooting ColorChecker
target reference images is always an excellent backup plan
for challenging lighting environments to ensure accurate
color reproduction.

Calibrating Autofocus

Front-focusing or back-focusing are problematic behaviors

where the camera’s autofocus puts the focused image
just in front of or behind the sensor plane. If this behav-
ior is observed across different lenses, it’s likely a camera
problem that is fixable with a simple calibration process.
If it’s exclusive to one particular lens in your loadout, it
might indicate a problem with optical element alignment
(maybe from that time you dropped the lens and thought
everything was fine) which could require professional
repair.
Recall that the primary method of automatic focus on
the traditional SLR camera involves a dedicated, phase
detection sensor receiving a reflected, secondary image
from the primary mirror. Micro-adjustments can be nec-
essary if the phase detection sensor is receiving a slightly
different version of the image that reaches the image
sensor during capture; they tell the algorithms to bias their Figure 11-12 A lens alignment chart.

findings by a user-determined offset. Mirrorless systems

or using contrast autofocus methods are not as likely to
require such calibration. auto focus position lands. Reshooting the target after an
Autofocus micro-adjustments are the solution to adjustment can confirm that the automatic focus is hitting
the first problem and luckily, professional cameras offer its desired focus distance every time. The steps to identify
calibration settings in their firmware. Using an alignment and correct focusing errors are as follows:
tool such as LensAlign can assist in determining if a lens
and camera combination is properly focusing. A focus 1. Setup camera on a tripod
calibration chart features a ruler at a 45o angle relative to 2. Set the camera AF mode to single point
the camera; instructions for do-it-yourself, inkjet-printed 3. Focus on an evenly illuminated test target positioned at
ones can be found online. We use the one shown in Figure 45° relative to the camera
11-12. Capturing an image of such a target using your 4. Capture at the widest available aperture
camera’s single-AF point mode can identify any front-fo- 5. Evaluate the image at 100% zoom. Check if focus is in
cusing or back-focusing behavior and the ruler offers useful front of, behind or at the center mark on the target
visual landmarks. The camera firmware offers manually-set 6. Tweak AF micro-adjustment in camera firmware menu
micro-adjustments in the forwards or backwards direction (plus or minus)
to compensate for any observed inaccuracy in where the 7. Repeat steps 3–6 until focus discrepancy is resolved

245
 s ecti o n 2: P roc ess

Similar to this method but perhaps a bit more difficult Proof of Concept: Comparing Noise
to execute involves using a resolution target mounted per- Characteristics of Two Cameras
pendicular to the camera. The image of the target can then
be examined for sharpness and any focus non-uniformity. Evaluating noise characteristics in a camera is a great way
This method requires some guesswork to determine in to make objective comparisons across sensors while also
which direction the system is mis-focusing. offering an opportunity for subjective analysis. The follow-
ing exercise can reveal the upper limits of sensor sensitivity
for your personal tastes. Your camera might have a super
Considering Your Imaging Tools high ISO setting that isn’t worth considering when adjust-
ing exposure variables if it produces a distracting amount
Sensor specifications, lens performance and everything of noise. Photographers ought to feel comfortable knowing
else in this chapter are interesting topics in isolation. There what sort of quality they can expect from their camera
are also additional image quality areas that extend beyond hardware when working in limited light environments:
what we’ve offered here. At the end of the day, what mat-
ters is the combination of equipment, settings and image 1. Set the camera to collect raw images. Turn off all
processing that come together in your photographic in-camera noise reduction. Do not assume that the
practice. Most photographers have multiple cameras and default settings exclude noise reduction.
knowing which one to use in which situation is key to 2. Place the camera on a tripod. Arrange a scene with a
creating the perfect image in every situation. ColorChecker target and any props that are relevant to
Testing tools like Adobe Camera Calibration uses a your shooting practices.
checkerboard target and accompanying Lens Profiling 3. Capture a properly exposed image at each available
software can make quick work of identifying image whole stop increment of ISO. We exclude “low” or “high”
quality shortcomings of your system (focus, distortion, settings in our example. Adjust shutter speed while var-
light fall-off, chromatic aberration). This is a great option ying ISO to create equivalent exposures to keep image
for the photographer that lacks the resources or time to brightness the same throughout. Figure 11-13 shows
dive into more exacting measurements. Resources like close crops from our sequence.
DxOMark are valuable when looking for new equipment 4. Ingest the captured set and open the raw files into
or to generally understand how camera systems can be Adobe Photoshop. Navigate to Window > Histogram. In
tested in-depth.5 the upper right-hand corner of the histogram window
Characterizing your camera system means knowing its is a pull down; select Expand View. This shows mean and
low light limits, its focusing accuracy, its detail resolving standard deviation readouts.
power and the numerous other image quality aspects 5. Use the rectangular marquee selection tool to select a
described in this chapter. No system is perfect and our take square area within the gray patch of the ColorChecker.
on camera hardware prioritizes facilitating good image 6. Measure and record mean and standard deviation
makers rather than any purely objective, on-paper perfor- values from the Histogram window. Repeat for the black
mance. We use our cameras every day and always keep patch on the ColorChecker and for all raw captures in
perspective that these are artistic tools with deep engineer- the set.
ing and technological foundations. 7. Calculate the signal to noise ratios for both patches:
SNR = 20*log(mean/standard deviation).

We calculated SNR values for our Camera A and Camera

B across a range of ISO settings. For example, at ISO 200,
Camera A shows a gray patch luminance value mean of

246
 c h a p t e r 11 : C a m e r a C h a r a c t e r i z at i o n a n d I m a g e Q u a l i t y

Figure 11-13 A lineup of gray patches from ISO 100–6400 captured by two cameras.

to Camera A and noise performance is a clear area where it

serves as a disadvantage.
Visually examine the gray and black patches in your
photographs at each ISO and for each camera. Consider
the following questions:

• At which ISO does the noise become objectionable? Is

luminance or chroma noise more distracting or degrad-
ing to image content?
• Do the numbers make sense when comparing the two
cameras given their hardware differences?
• Is the relationship between ISO and noise in midtones
and shadows linear?
• Are the noise levels the same in the gray patch com-
pared to the black patch?
Figure 11-14 Plotting SNR values of Camera A and Camera B for gray • How might you capture images to discover fixed pattern
and black patches across a range of ISO settings.
noise in a camera?

176.64 with a standard deviation of 5.82. This calculates to

a SNR of 29.64. Setting up a spreadsheet to calculate the Notes
set of recorded values considerably speeds up this exercise. 1 Phillips, Jonathan B. and Henrik Eliasson. Camera Image
Plotting this data with an x-y scatter is visually intuitive (see Quality Benchmarking. John Wiley & Sons, 2018, p. 29.
Figure 11-14). 2 “Color in business, science, and industry.” Color in Business,
We observe that Camera A records the neutral patches Science, and Industry, by Deane Brewster, Judd and Günter
Wyszecki, Wiley, 1975, p. 388.
at consistently higher signal to noise ratios at every ISO
3 Sharma, Gaurav. Digital Color Imaging Handbook. CRC Press,
compared to Camera B. For example, Camera B at ISO 200
2017.
has roughly the same amount of shadow noise as Camera
4 Abell, Mark. “White balance, part 2: The gray world
A at ISO 1600. This is consistent with our understanding
assumption and the Retinex Theory.” The Refracted Light,
that larger pixels are generally better at gathering light: Sept. 15, 2011, therefractedlight.blogspot.com/2011/09/
Camera A is a full frame, high-end DSLR while Camera B white-balance-part-2-gray-world.html.
is an older mirrorless crop sensor. Camera B has smaller 5 “The reference for image quality.” DxOMark, www.dxomark.
photosites packed into a small sensor size footprint relative com/.

247
12 Image
Compression

Photograph by Rochester Institute of Technology photography student Caroline Brodt

The growth and maturation of a visual medium happens when creatives are able to create,
imagine, make, capture and share new work. Though purely creative types may not think
of their creations as boiling down to digital data, the reality is that photographs (along
with graphic design, video and audio) are built on the foundational building blocks of
ones and zeroes. When we create millions of new pixels, dozens of composite layers or
hours of high-definition video, the challenge of storing all of this digital information can’t
be ignored. This textbook has so far highlighted the hardware used to house this data (see
Chapter 8 on the topic of File Storage) and approaches to managing it. Still, we’d drown
in data if it weren’t for compression techniques that help to keep files reasonably sized.
Applying compression to photographic data is a key technological necessity in the digital

249
 s ecti o n 2: P roc ess

realm. Here, we explore the details of image compression

mechanisms by reviewing the concepts of information
redundancy, the primary algorithmic approaches to com-
pressing data and the potential image quality artifacts that Figure 12-1 A selection of pixels from the sky in a landscape
crop up. photograph.

Image Compression and wrinkled because we assume an iron is available at our

Information Theory destination hotel room. Image compression is possible
because software algorithms can maximize the amount
Image compression is a little like physical compression. of data stored by finding redundancies and predictable
You pile on shirts, socks and other clothing when packing patterns. It similarly assumes that software on the other
a suitcase for vacation. Sometimes you resort to sitting on end can understand some instructions or coding short-
the luggage to get it to close: you’re compressing your hand to make sense of the stored information. Consider a
clothes to fit into a container with limited space. When you landscape photograph with a blue sky above the horizon.
arrive at your destination and open up the suitcase, all of Zooming in close on the image file, a section of the sky
your clothes are there and return to their original size when might contain a row of pixels like the ones shown in Figure
you take them out (wrinkled, but an iron can fix that). You 12-1.
just experienced lossless compression! Now imagine the Each pixel consists of a color and brightness that
same scenario with smaller, carry-on luggage. No amount defines its appearance. Storing the brightness and color
of sitting on your outfits compresses them enough to fit, information of each pixel can be accomplished in three
so you start sacrificing pairs of socks and extra shirts. This ways: without compression, using lossless compression
is lossy compression: you’re excluding some previously and using lossy compression. Imagine being tasked with
available material to fulfill a stricter storage requirement. describing these six pixels over the phone. You can use one
The consequence is that you won’t have those abandoned of these three approaches:
items available when you get to your destination.
This example is a simplified explanation of compression 1. Uncompressed: “Pixel 1 is light blue. Pixel 2 is light blue.
but it begins to convey the challenge and the compro- Pixel 3 is light blue. Pixel 4 is medium blue. Pixel 5 is dark
mises of the process. Image compression is based on blue. Pixel 6 is dark blue.”
information theory which the Encyclopedia Britannica 2. Lossless Compressed: “Pixels 1 through 3 are light blue.
defines as “a mathematical representation of the conditions Pixel 4 is medium blue. Pixels 5 and 6 are dark blue.”
and parameters affecting the transmission and processing 3. Lossy Compressed: “Pixels 1 through 6 are light blue.”
of information.”1 It’s important to clarify the difference
between data and information. Data are the binary rep- The uncompressed approach explicitly conveys every
resentations of the pixel values in an image. Information is pixel with absolute fidelity. The lossless compressed version
how the pixel values are arranged in the image frame to does the same but with fewer words thanks to the layer of
represent photographic content. The goal of image com- logic applied. The person on the other end of the phone
pression, then, is to reduce the information by reducing the knows that “pixels 1 through 3” describes a redundancy or
amount of redundancy in the data used to convey it. repetition of light blue. The lossy compressed description
The example of packing a suitcase works because simplifies the sky’s subtle differences in shades of blue in
it’s possible to maximize the use of a physical space by the interest of brevity. It effectively compresses redundant
minimizing the amount of air between objects. We also information as long as we don’t mind our sky depicted
accept the reversible process of our clothes getting with the same light blue throughout. It cannot, however,

250
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

be used to accurately reconstruct the original, varied pixel redundancy reduction processes; DCT is lossy. Map-
values if desired. ping RGB data to luminance and chrominance channels
Image compression algorithms are much more sophis- commonly occurs during this first step. When lossy com-
ticated than the previous example yet the basic concept pression is desired, the next step is the quantization stage.
holds true. The lossless description mimics a common data This is where information is removed: chrominance infor-
storage technique called run-length encoding that stores a mation and psychovisual redundancy are often reduced,
piece of information once and a count of how many times driven by user input. The encoding stage is the last step
in a row that information is repeated (i.e. light blue pixel, for both lossless and lossy compression. Variable-length
three instances). The brightness values of two neighbor- encoding, such as Huffman coding described later, occurs
ing pixels are more often than not the same or similar, so at this stage.
saving a version of an image that rounds some of these
similar, neighboring pixels (called interpixel redundancy)
reduces the file size with minimal impact to image detail. Finding Redundancies in Image Data
Additionally, common image compression algorithms
separate the luminance and chroma information and Lossless compression is appealing because there is no loss
approach each with different strategies that have the best of information between compression and decompres-
chances of minimizing noticeable quality degradation. The sion. It is limited in effectiveness, though, because there
delicate balance of compression is in reducing file sizes is only so much that a lossless compression algorithm
without introducing noticeable, distracting image quality can do to lessen the file size while preserving the original
artifacts. information.

The Basic Stages of Image Spatial or Interpixel Redundancy

Compression The digital counts or brightness values of adjacent pixels
are more often than not the same or similar. If we have to
Image data compression exploits redundancies and lim- guess a missing pixel value, we could do so based on the
itations of the human visual system. Redundancy means surrounding pixel values. Knowing this, it’s not necessary to
excess content or information that can be discarded keep all of the pixel data to represent the image. Dupli-
without it resulting in a noticeable difference to the final cate elements in a structure, such as the same pixel values
product. All digital images contain redundancy at the pixel repeated in a row of an image array, are the definition
level. We compress image files without noticeable loss of spatial or interpixel redundancy. Greater spatial redun-
of quality when taking advantage of these redundancies. dancy offers increased compression effectiveness. Graphic
The limitations of compression are defined by our ability designs tend to have many redundancies since they
to notice shortcomings or unfaithful reproductions. We are simplified tones and colors. Photographs of the real
could heavily compress every image and have the benefit world are much more random and varied in pixel content,
of small file sizes but at a certain point we’re not willing to meaning that taking advantage of interpixel redundancies
trade reduced visual fidelity or quality for those gains. happens on relatively local areas rather than across large
Image compression is a two or three step process. The parts of an image file.
first step is often referred to as pre-processing. Interpixel Spatial redundancy is also helpful for removing dead
redundancy reduction occurs during the pre-processing pixels or small dust spots. These outlier pixels are sur-
step. This is accomplished through run-length encoding, rounded by valid image data and as such our algorithms
differential encoding or using discrete cosine transform (DCT). borrow or interpolate from their neighbors. This is a highly
Run-length encoding and differential encoding are lossless effective process since there is a good likelihood that the

251
 s ecti o n 2: P roc ess

Table 12-1 The bits used in an uncompressed image compared to a a 1 or a 0, when describing its brightness value. Table 12-1
compressed version of the same image.
shows the number of bits used to record different ranges
Pixel Value Range Bits Used Bits Needed of pixel values. An 8-bit file uses 8 bits for each value
(uncompressed (compressed whether all of those bits are needed or not. For example:
image) image)
the binary form of a pixel value of 2 is 00000010. This uses
0–1 8 1
8-bits to describe the value yet the same thing can be
2–3 8 2 stored using just 2 bits (coded as 10 in binary). Both coding
4–7 8 3 methods describe a pixel value of 2; the latter does so
8–15 8 4 using less disk space. Lossless compression uses this coding
16–31 8 5
redundancy to reduce file size.

32–63 8 6

64–127 8 7

128–256 8 8 Psychovisual Redundancy

Taking advantage of the human visual system’s short-
comings is the strategy used by psychovisual redundancy
compression. Our visual system relies more on spatial
adjacent pixels are the same color and brightness. The junk information than color information and it is more sensitive
pixels are replaced with the interpolated values. to lower frequency information than high frequency infor-
mation. Subtle changes in luminance are more noticeable
than the same amount of change in chroma. Some color
Coding Redundancy and brightness differences fall below the threshold of
An uncompressed 8-bit image has a fixed length in which our perception. Compression takes advantage of these
to record each pixel value. A pixel can have a value from scientifically proven insights by treating color informa-
0 (black) to 255 (white) in each color channel. Computers tion independently from brightness to reduce file sizes.
use binary codes to represent all characters. In other words, Our greater sensitivity to luminance changes means that
every pixel in an image array uses 8 bits, each set to either the eye is quicker to detect artifacts and image quality

Figure 12-2 An example of psychovisual redundancy. The high frequency image on the left is
represented by 8-bits per channel. The image on the right is a 6-bit per channel image. The images
appear the same.

252
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

degradations if compression algorithms are too aggressive

with brightness data. Figure 12-2 is an example of psycho-
visual redundancy.

Temporal Redundancy
Videos and short animations are comprised of images
shown in sequence over a specified duration. Content
from one frame to the next is often highly redundant: this
is temporal redundancy. A video of a model walking down
a runway shows frame after frame of the runway and the
surrounding audience. The only changing content is the
model’s movement. The background pixels exhibit tempo-
ral redundancy; they are highly similar from one moment
to the next and potentially over many dozens of frames or
seconds. Compression algorithms take advantage of this
repetition across sequential video frames. We explored
temporal redundancy under the guise of intraframe com-
pression in Chapter 6.

Compression and Scene Content

Consider two images: one with large areas of high fre-

quency information (like the lion cub image in Figure Figure 12-3 A photograph with low frequency areas is easier for a
12-2) and another with large areas of low frequency compression algorithm to efficiently compress.
information (like the studio still life image in Figure 12-3).
Both images are taken with the same camera, saved in the
camera’s native, uncompressed raw format and are the
same approximate file size. The images are then ingested Lossless Compression Approaches
and converted to DNGs with lossless compression. Lossy for Removing Spatial Redundancy
compressed JPEG versions are also generated. The resulting
files from these two scenes end up being very different There are many methods to achieve lossless image com-
file sizes. The second scene includes large areas of simi- pression without losing any original data. Three methods
lar pixel values which both DNG and JPEG compression are described here: run-length encoding, differential
strategies take advantage of to minimize data storage. The coding and bit-plane encoding.
lion cub image does not have significant data redundancy
and therefore does not compress well. These examples
highlight that the effectiveness of compression is image or Run-Length Encoding
scene content dependent: some images are easier to com- We started to explain run-length encoding (RLE) earlier in
press than others. Both scenes can be lossy compressed to this chapter when describing Figure 12-1. RLE is sup-
the same file size if desired, but the lion cub image is more ported by most bitmap file formats and operates under
likely to show image quality degradation from that exercise. the principle that adjacent image pixels exhibit spatial

253
 s ecti o n 2: P roc ess

Bit-Plane Encoding
Another approach is bit-plane encoding which looks at an
image as a stack of eight 1-bit images rather than a single
layer comprised of 8-bit data. Recall that every pixel in an
image uses 1 byte of memory. One byte is made up of
8-bits or a combination of 8 ones and zeros. Separating
each bit into its own plane creates eight binary image
Figure 12-4 The pixel values of the blue area through the letter “m” planes, planes 0 through 7, as illustrated in Figure 12-5. So
are represented by the row of black and white squares. This string of
ones and zeroes is simplified to the description below it using run- how does that help to compress the data?
length encoding. We need to look back at how information is stored to
answer that question. The first digit of our 1-byte number
is considered the most significant bit as it contains most
of the data. The first bit is 1 for all pixel values above 128
redundancy. This is especially true for images with graph- and 0 for all pixels below 128. Table 12-2 shows that the
ics or text in them. This example contains large areas or amount of information each bit represents gets smaller
runs of white (pixel value 1) with small instances of black as the pixel values decrease. The last bit is considered the
(pixel value 0) that represent the letters. The size of the file least significant bit as it often represents the digital noise in
is greatly reduced by replacing a run, or repeated string, an imaging system.
of the same pixel value with a 2-byte representation. The When each level is reduced to a binary image, the
first bit is the run count or the number of times that a pixel probability that an adjacent pixel is equal translates to
value occurs. The second byte is the run value, the value higher spatial redundancy. Higher spatial redundancy
of the pixel. Figure 12-4 shows how long runs of the pixel translates to higher compression rates. Figure 12-5
value 1 are losslessly replaced using a run-length encod- illustrates a portion of an image separated into eight
ing strategy. separate bit planes. The first two pixels have values of

Table 12-2 Binary representation of pixel values.

Decimal Binary

128 10000000

64 01000000

32 00100000

16 00010000

8 00001000

4 00000100

2 00000010

1 00000001

Figure 12-5 An image pixel array separated into eight bit planes. Bit Plane 7
exhibits spatial redundancy across the entire plane.2

254
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

127 and 95 and they are not similar in any way. The spa- Lossless Compression Approaches
tial redundancy becomes apparent when viewed in bit for Removing Coding Redundancy
planes: planes 0, 1, 2, 3, 4 and 6 for both pixel values are
the same. There are many techniques for removing coding
redundancy. The two techniques commonly used with
photographic content are Huffman coding and Lempel–
Differential Encoding Ziv–Welch (LZW) compression.
The last method of reducing spatial redundancy is called
differential redundancy. It’s unlikely to find long runs
of identical pixel values even when looking at low fre- Huffman Coding
quency areas like a clear blue sky. That said, low frequency Huffman coding is a lossless technique that looks for statis-
areas have minimal variation in those values. Differential tical patterns and assigns symbols to each pattern element.
encoding takes advantage of this low variance. The first This approach is not limited to image data and can be
pixel value is recorded and only differences from that applied to video, audio, text documents and spreadsheets.
pixel are recorded after it. This is conducted in small pixel Static Huffman coding can be applied in a single pass if the
groups segmented from the entire pixel array. Figure content of the data is already known. Dynamic Huffman
12-6 is an example of differential encoding applied to a coding is a two-pass process. A histogram of pixel values
row of six pixels starting with a value of 60. These six pixel is created on the first pass. Each bin in the histogram is
values require 48-bits to store them uncompressed. Using turned into a probability of occurrence called a leaf node.
differential encoding reduces this to 13-bits, a substantial Table 12-3 is a frequency count of letters found in alphabet
space saving. High variance makes this approach much soup and its corresponding leaf nodes.
less effective since the differential might be just as big or The second pass encodes the data. Both static and
bigger than the native value. dynamic methods use the same encoding technique to
build a Huffman Tree like the one shown in Figure 12-7
(overleaf). The algorithm locates the two leaf nodes with

Table 12-3 A frequency tally of letters found in alphabet soup and

the corresponding leaf node values calculated for dynamic Huffman
coding.

Letter Tally Leaf Node Value

a 8 0.08
Figure 12-6 Differential encoding stores the first value in a set and
b 10 0.01
then the relative difference of the values that follow. The fifth pixel
value, 58, is 2 away from 60. Compared to storing the value of 58, a c 12 0.12
differential value of 2 requires fewer bits.
d 14 0.14

e 24 0.24

f 32 0.32

Total 100 1

255
 s ecti o n 2: P roc ess

compress GIF, TIFF and PDF files. LZW compression creates

a code table that typically contains 4,096 entries. Com-
pressing an 8-bit image means that the first 256 entries
in the code table contain the pixel values 0 through 256
while the other entries are left blank. The algorithm steps
through the image data looking for unique strings and
adds them to the code table while concurrently compress-
ing it.
Let’s step through compressing a string of letters,
“ABADABA” in Figure 12-9. The original file size before apply-
ing LZW compression is 56-bits. We’ll assume that there are
only four possible letters in a string: A, B, C and D. The code
table for the example string is populated with those letters
as codes 0 through 3 and the remainder of the table is left
Figure 12-7 A Huffman tree for our alphabet soup data. empty. In Step 1, the first character in the original string
(“A”) is examined to determine if it is in the code table. It is
present, so Step 2 takes the next character from the original
the lowest occurrences or probabilities and combines string and adds it to the current string (making it “AB”). The
them into a new node. The leaf node with the lower string “AB” is not in the code table, so the LZW coding adds
probability prior to combing is assigned a 0. The higher leaf it as code value 4. The “A” is removed from the front of the
node is assigned a value of 1. The process is repeated until “AB” string and encodes it as 0, leaving “B” behind. Step 3
all nodes are combined to create a root with a probability examines “B” and finds it in the coding table. It then takes
of 1. A code look-up table is created to compress the data. the next character from the original string and adds it to
The same table is used in reverse to decipher these symbol the current string making it “BA.” In Step 4, “BA” is not found
assignments when an encoded file is decoded for display. in the code table, so it gets added as code value 5. The “B”
Figure 12-8 shows how the Huffman coding approach is then removed from “BA” and encoded as 1. This same
reduces the file storage of our alphabet soup data by 88%. process is followed for all of the remaining characters in the
original string. When it’s complete, the resulting LZW-en-
coded string is “010340” that can be stored with the coding
Lempel–Ziv–Welch Compression table in a file size of 10-bits. The code table is used in reverse
The Lempel–Ziv–Welch lossless compression algorithm was to decode the data. This method results in a full restoration
developed in 1984 by Abraham Lempel, Jacob Ziv and of the original information with nothing lost or abbreviated
Terry Welch. The algorithm is employed to effectively while offering a notable storage efficiency improvement.

Figure 12-8 Huffman coding compresses the alphabet soup data by 88%.

256
 C H A P T E R 12 : I M A G E C O M P R E S S I O N

Figure 12-9 Construction of an LZW code table.

the Ycbcr color space three values: luminance (Y), chroma blue (Cb) and chroma
red (Cr).
Every pixel in a color image is represented by a red, green Conversion to the YCbCr color space is a valuable prac-
and blue value. This is great for image display as moni- tice for compression algorithms. Human vision is more
tors expect these RGB values. It’s not the optimal space sensitive to changes in luminance than to small changes
in which to operate for image compression, however. in color. Separating the brightness information from the
Instead, RGB images are converted to a luma-chrominance color information allows algorithms to apply different,
color space where every pixel location is described by independent compression rates for luminance and color

Figure 12-10 The original RGB image (left) and the three YCbCr components after conversion to a luma-chrominance color space.

257
 s ecti o n 2: P roc ess

to achieve additional storage efficiency. The conversion compression can take advantage of the combined effects
from RGB to YCbCr is defined for standard definition tel- of psychovisual and temporal redundancies by subsam-
evision in the International Radiocommunication Sector pling color in image frames. Chrominance subsampling uses
(ITU-R) BT.601 standard. Equations 12.1–12.3 list the con- the same color values across multiple pixels in an effort to
version equations. reduce the amount of data; the relative insensitivity of the
human visual system suggests that this approach is possi-
Y = 0.299R + 0.587G + 0.114B (Eq. 12.1) ble with minimal observable degradation to the imagery.
Cb = 128 – 0.168736R – 0.331264G + 0.5B (Eq. 12.2) Chroma subsampling was originally invented for color
Cr = 128 +0.5R – 0.418688G – 0.081312B (Eq. 12.3) television broadcasting. If an uncompressed image is built
from color and luminance layers at full resolution, a chroma
Figure 12-10 illustrates an RGB color image file con- subsampled image has a full resolution luminance layer
verted to YCbCr. It’s possible that some information is lost and lower resolution chroma layers (see Figure 12-11).
when transforming from RGB to YCbCr due to rounding Chroma subsampling is described with the following
errors. notation:

A:B:C (Eq. 12.4)

Compressing Color where

Converting image data into the YCbCr space means that A = the number of pixels wide (by 2 pixels in height)
the color characteristics can be compressed independently that define the sample area
from luminance. Applying additional color compression is B = the chroma samples in the first row of Ax2
a strategy used in JPEG compression and it is also particu- C = the chroma samples in the chroma samples in the
larly effective in video encoding. This is because video second row of Ax2

Figure 12-11 Converting an RGB image to YCbCr and subsampling the chroma channels.

258
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

Figure 12-12 The array on the left represents a group of image pixels with full luma and chroma
sampling (4:4:4), the array on the right represents a 4:2:2 sampling schema.

An image without chroma subsampling is described

as 4:4:4 where a 4x2 pixel section of an image has unique
chroma samples for all pixels. Using the same chroma
values for every other pixel in a row is called 4:2:2 and
effectively halves the chroma resolution (see Figures 12-12
and 12-13). Further sharing the chroma samples from row
1 with row 2 yields 4:2:0 subsampling (also called 2x2).
Using the same chroma values for every fourth pixel (i.e.
1/4 sampling) is 4:1:1. The more that the chroma data is
subsampled, the greater the potential compression ratio
relative to the original, uncompressed image or video
frame. Video encoders offer this chroma subsampling
notation as a tunable parameter (4:2:0 is common, particu-
larly with cameras designed for still photography). JPEG
compression may use chroma subsampling at higher com-
pression levels (lower “quality” levels) to achieve smaller file
sizes, though the exact sampling method is decided by the
image editing software.

Using Discrete Cosine Transforms

to Compress

Discrete cosine transforms rely on representing all image

content using combinations of cosine waves at different
frequencies. Images are broken up into 8x8 pixel blocks Figure 12-13 An image compressed with 4:4:4 (top) and with 4:2:2
(bottom). Our vision is not as discerning when the color resolution is
or 64 pixel values in total. Any combination or arrange-
halved.
ment of pixels in the 8x8 pixel block can be represented

259
 s ecti o n 2: P roc ess

image is unlikely to introduce a noticeable difference in

appearance.
The user defines the desired quality level when com-
pressing an image using DCT. The quality level is used to
create a quantization table. Recall from Chapter 3 that we
quantize the continuous analog light signal into digital bits
that represent pixel values in a photograph. The quantiza-
tion table is used to weight the DCT coefficients which are
then rounded to the nearest integer. Many of the DCT coef-
ficients are set to zero when the compression level is set to
a low value, for example. This may result in noticeable data
loss or image quality degradation.
There are other options beyond DCT, namely using
discrete wavelet transform (DWT). DWT is used by JPEG 2000
for specialized imaging applications. For a detailed explana-
tion of this, we suggest Chapters 28 and 29 in The Manual
of Photography by Allen and Triantaphillidou.3
Figure 12-14 The DCT 64 base cosine waves used to replicate any
8x8 image pixel group.

JPEG Compression
by combining 64 cosine waves of different values. Figure
12-14 shows these combined cosine waves represented The JPEG is the go-to format for end-product image files to
in image pixels. The upper leftmost square represents send to clients or for embedding on websites. Recall that
low frequency flat areas. The frequency of the image data the JPEG name refers to both the file container format and
increases moving toward the lower rightmost square. the lossy compression standard it uses to store the image
DCT compression starts by segmenting an entire data. A color image undergoes the following steps with
image into 8x8 pixel blocks. Using the 64 base cosine JPEG compression algorithms:
waves shown in Figure 12-14 (also an 8x8 block), a coef-
ficient is assigned for each of the blocks. Each coefficient 1. The image is converted from the RGB space to the YCbCr
represents how much that cosine wave pattern contrib- color space.
utes to the block. For example, if the coefficient for the 2. The chrominance channels (CbCr) are subsampled,
upper left-hand cosine wave block of Figure 12-14 is taking advantage of psychovisual redundancy.
large, it indicates that there is a significant amount of low 3. The image is divided into 8x8 blocks. If the image rows
frequency information in the 8x8 pixel block. The result- and columns are not perfectly divisible by eight, the
ing 8x8 array of coefficients will typically have a common bottom row and right column of the image are dupli-
arrangement. The upper left square is the DC coefficient cated to fill the block.
and it typically stores a very large value. The DC coeffi- 4. A Discrete Cosine Transform is performed to determine
cient represents the low frequency information in the compression coefficients.
8x8 block. As you traverse the 8x8 block diagonally to the 5. The compression coefficients are quantized according
lower right, the coefficients typically become smaller and to a user-defined quality factor. The range of this factor is
smaller. In terms of image data, the lower right checker- defined by the software package used to create the JPEG.
board pattern in Figure 12-14 is the high frequency data. 6. The resulting quantized data is Huffman coded for addi-
We often find that removing high frequency data from an tional storage efficiency.

260
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

Figure 12-15 The JPEG compression quantization factor was set to an

aggressive value that degraded image quality but reduced the file size
by more than half. Photograph by Ruben Vargas

7. The compressed image data is saved to a JPEG container

file (see Figure 12-15).

The JPEG compression process has two lossy steps. Step

2 subsamples the chrominance channels (CbCr). Once
the chrominance channel data is subsampled, it cannot
be restored faithfully to the original image data. Step 5
introduces additional data loss; the lower the user-defined
quality factor, the more data is lost.
Figure 12-16 A JPEG-compressed image (top) and an image-
differenced rendering (bottom) showing pixel changes between the
original raw data and the JPEG.
Common Image Compression
Artifacts

In a perfect world, lossless compression is computationally Image degradations resulting from compression algorithms
cheap, fast and extremely efficient at keeping file sizes small. are called compression artifacts. Ultimately, choices made
The real world is one where tradeoffs and compromise are regarding compression can boil down to the final product:
necessary: the smaller the desired file, the more we must is the resulting image an acceptable representation of the
accept losses in image fidelity. There is always a balance original content? Navigating and balancing storage space,
between minimizing disk space used and image quality loss. data transmission and image fidelity hinges on minimizing

261
 s ecti o n 2: P roc ess

the presence of compression artifacts. The primary arti- the shared border where they meet will have a natural tran-
facts resulting from JPEG compression—highlighted here sition of color or tone. Our eyes start to suspect something
because if its ubiquitous use—include blocking, ringing and is amiss and a subtle, grid-like abruptness of pixel values
posterization. Artifacts are rarely exclusive and we encoun- becomes apparent along these block borders. Blocking arti-
ter combinations of blocking and ringing, for example, in a facts increase with the increase in target compression ratio.
heavily compressed image. Figure 12-16 shows a close crop
of a JPEG-compressed image next to a calculated difference
between it and the original raw capture; notice how the Ringing Artifacts
compression algorithms grouped similar background tones Edges are a critical component to any image as they define
into uniform blocks. This method of visualization starts to the boundaries of subjects and image details. Edge detail
highlight the compression operation and potentially prob- must be maintained between the source content and a
lematic areas of scene content that may show artifacts. compressed variant, leading compression algorithms to
handle edges differently from other pixels. JPEG compres-
sion can introduce halo-like outlines like pond ripples along
Blocking Artifacts contrast boundaries called ringing artifacts. Recall that JPEG
A commonly spotted image quality artifact resulting from compression segments an image into 8x8 blocks; ring-
excessive compression, particularly JPEG compression, is ing occurs when an edge spans across multiple adjacent
blocking. One might describe it as “pixelated” or “check- blocks. In addition, most compression algorithms sacrifice
erboard” because the blocks of local compression are high frequency data as we are less sensitive to it. Edges are
apparent (see Figure 12-17). Compression algorithms take high frequency information and are therefore nearly always
small subsets of an image frame and work to simplify their affected by compression algorithms. Ringing artifacts are
contents on a per-block basis. The problem arises when the illustrated in Figure 12-18.
compression of one block (for example, an 8x8 pixel array)
alters the pixels along that block’s edge. If the neighboring
block gets the same treatment, there’s no guarantee that

Figure 12-18 Ringing artifacts occur along edges and areas of high
contrast when JPEG and JPEG 2000 compression image quality is set
very low.

Posterization
Areas of continuous tone or subtle gradation can suffer
from posterization if they are too heavily compressed.
Areas that seem homogenous like skies often have small
changes in tone (say, from a darker blue to a lighter blue-
Figure 12-17 The 8x8 pixel blocks become apparent when they are cyan). Compression algorithms see a lot of pixels that are
heavily quantized. The abrupt changes in color and tone from one
block to the next is far from the continuous, seamless transitions we
similar to their neighbors and generalize them as the same
experience in real life. luminance and/or color. This strategy, when applied over

262
 c h a p t e r 12 : I m a g e C o m p r e s s i o n

Figure 12-19 Posterization seen in the sky as a result of excessive JPEG compression. Photograph
by Ruben Vargas

an area that transitions from one color to a slightly different from people downloading, saving and uploading content
one, introduces discrete changes. Posterization describes over and over again. Each time that image data is reup-
these rough, abrupt transitions where we expect to see a loaded or changed and re-saved, the resulting content
visually seamless, full color gradient. It is also called banding becomes one generation more distant from its parent
because of the distinct bands of color or tone that manifest source. Photographers can get away with resaving a JPEG
in areas of gradation (see Figure 12-19). a few times in a row without major artifacts appearing.
Content on social media, though, is typically hit with high
compression ratios to save the host companies money
Color Bleeding on bandwidth and storage. The effects of generation loss
The negative consequence of chroma subsampling is seen quickly manifest as these heavily compressed files are
as color bleeding artifacts where the subsampled colors shared, downloaded and reuploaded and compressed at
begin to exceed the boundaries defined by the luminance. every iteration.
Like the name indicates, think of this artifact like watercolor
paint: the color boundaries blur and do not necessarily stay
within the lines of subject edges in the image. Notes
1 Markowsky, George. “Information theory.” Encyclopedia
Britannica, Encyclopedia Britannica, Inc., June 16, 2017,
Generation Loss www.britannica.com/science/information-theory.
If you’ve spent enough time on social media platforms 2 Patil, Aditi Y., et al. “Security of grayscale images by VSS
with novel method of bit-plane slicing – Semantic Scholar.”
then you’ve undoubtedly seen photos, video or graphics
Semantic Scholar, Jan. 1, 1970, www.semanticscholar.org/
that, like leftovers from a once-fresh meal reheated one
paper/Security-of-grayscale-images-by-VSS-with-novel-of-
too many times, show serious signs of wear. Generation
Patil-Alexander/68d7294b51b105443fc71d956d1317dae18
loss is degradation due to repeated lossy compression
d114c.
and can reveal all of the previously listed artifacts com- 3 Allen, Elizabeth and Sophie Triantaphillidou. The Manual of
pounded on one another. On social media, this stems Photography. Elsevier/Focal Press, 2017.

263
Section 3
output
Output is an exciting stage in the life of a photograph simultaneously excel and fall short of satisfying our highly
because it promises to showcase all of the pre-planning, sensitive visual organs. Then we tackle the much-maligned
careful capture and exposure, digital asset management topic of color management that brings together everything
and post-processing editing work. Without delivering your we’ve explored regarding light, digital imaging technology,
photography, either via print or screen, it’s doomed to software and hardware. Once we’re comfortable with con-
reside as files on a hard drive. Photographic output is the cepts such as color profiles and rendering intents, we look
final leg of the race: just like the earlier stages of capture at the potential output technologies for our photographs:
and processing, it demands an attention to detail and a displays and printed media. Showing your photography to
working understanding of human vision, output technol- the world is like crossing the finish line after the marathon
ogy and technical photographic processes. of visual media production (and of reading this book): it
This final section of this textbook covers color vision and should be satisfying, impactful and the culmination of
systems, color management, displays and finally, printing everything you’ve worked toward. Photographic output is
and physical media. We start with color vision to appreci- the means by which the world experiences your creative
ate how our eyes work and how our photographic media endeavors.
13 Color Vision
and Systems

Photograph by Rochester Institute of Technology alumnus and Lecturer Nanette L. Salvaggio

The experience of perceiving color does not depend upon language or numbers, but
the communication of information about that experience does. It is important to use an
agreed-upon and established vocabulary to talk about color if we are to be successful at
communicating it. In the past, various disciplines including photography, painting and the
graphic arts have had difficulty communicating with each other due to the absence of a
common color language. The Munsell Color System and the CIE System of Color Specifica-
tion emerged as two of the more successful attempts to organize the variables of color into
a universally accepted system. In addition, the Inter-Society Color Council brought together
representatives of the various disciplines that use color in an effort to lower the barriers to
communication. The stated aims and purposes of the organization are to “stimulate and

265
 s ecti o n 3: O utp ut

coordinate the work being done by various societies and Sensing Light with the Human Eye
associations leading to the standardization, description
and specification of color and to promote the practical The human eye is a complex sensing organ worthy of its
application of these results to the color problems arising in own textbook (others have written them) detailing its anat-
science, art and industry.”1 omy and function. Figure 13-2 illustrates the basic anatomy
Before examining color systems, however, we have to of the eye. The cornea is a fixed lens and does the majority
establish a definition of color. Color is a visual experience. It of the light-focusing, followed by a flexible crystalline lens
is a perception, a human physiological and psychological suspended by tissue that provides us with some refocusing
response to light. Color perception requires three compo- abilities for close and far subject distances (called accom-
nents: an object, a light source and an eye or visual system. modation). The main cavity of the eye is not empty space
There is no perception of color if any of these components despite how it may appear in illustrations. It is occupied
are absent. Color is associated with words like red, green, by a clear fluid called the vitreous humor that functions to
blue and yellow in everyday language, yet these words maintain the shape of the eye.
refer to only one of three attributes of color: hue. The other
attributes are brightness (or lightness) and saturation.
We further clarify our language of describing color by
separating the terms referring to the visual perception of
an object and those referring to the physical character-
istics of that object. For example, one can argue that we
shouldn’t describe an apple as being red. We perceive it as
red because the surface of the apple has physical prop-
erties that reflect certain wavelengths and absorb other
wavelengths of the white light that illuminates it. When a
red object is illuminated with blue light, most of the light is
absorbed and the object is perceived as black (see Figure
13-1). When we identify an object as being red, we mean
that individuals with normal color vision generally perceive
it as being red when viewed under normal viewing condi-
tions and white light illumination.

Figure 13-2 Anatomy of the human eye. Our central vision is sensed
from the fovea, the illustrated here as the dark region on the retina.

Figure 13-1 A red apple photographed under white light (left) and
under blue light (right).

266
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

The retina is the area at the back of the eye that detects Photopic vision is quantified by measuring the sensitiv-
light entering from the opening at the front of the eye (the ity of cone receptors to white light. The result is a luminous
pupil). The retina contains both rods and cones, the names efficiency function, VM(λ), describing the average spectral
given to the types of light receptors residing in the retina. sensitivity of human vision. Figure 13-3 reveals that human
Both get their name from their unique physical shapes. vision is most sensitive in the green-yellow wavelengths
There are approximately 120 million rods in a human eye. around 555 nm. The maximum sensitivity shifts when
Rods provide scotopic vision or our ability to see at night in dark-adapted or using scotopic vision to approximately
very low light conditions. Rods are only active in the dark, 507 nm. This shift in peak sensitivity is referred to as the
so their use is also described as dark-adapted vision. The Purkinje shift.
chemical in the rods, rhodopsin, gives us this ability to see There is one more type of vision: mesopic vision. Mesopic
when light is limited. Rhodopsin bleaches and becomes vision is used in low light environments by employing both
inactive when exposed to bright light. It takes about 30 rods and cones. Such visual perception provides some abil-
minutes for the rods to reach maximum sensitivity in a dark ity to see colors (though they appear muted). The peak of
environment and due to their inability to distinguish color, the luminous efficiency function shifts with mesopic vision,
we are colorblind when completely dark-adapted. pushing our peak color sensitivity toward a slightly lower
There are approximately six million cones in the retina. wavelength.
Cones provide us with both the ability to see color and Rods and cones are not evenly distributed across the
with photopic vision, the ability to see in lighted conditions. retina. Instead, a high concentration of cones is found at
Your vision is considered to be light-adapted when the the fovea, the area responsible for our central vision. Lock-
eyes rely on the cone receptors. There are three types of ing your gaze at text on a page is a collaboration between
cones in the retina that are sensitive to different wave- the body and mind to align the fovea such that your best
length bands. You may have learned these in grade school visual acuity is put to good use. Anything that falls outside
as being red, green and blue cones. This is an oversimpli- of this center portion of our field of view is perceived at
fication and are correctly referred to as long-wave (red), a lesser degree of fidelity. We often like to compare the
mid-wave (green) and short-wave (blue) cones. Each cone camera sensor to the human eye, but this is one place
type senses one of those three wavelength ranges, respec- where the analogy is deceiving. Human vision is a combi-
tively, with an additional degree of overlap. nation of detection and interpretation and its detection
mechanisms are not uniform. An image sensor is evenly
1 distributed with photosites and color filters. The pixels at
V (6) Photopic the center of the frame detect and record light exactly as
M
0.9 V'(6) Scotopic
the photosites along the edges do. Our retina lacks this
0.8
Relative Luminous Efficiency

uniformity, making up for it with cognitive processing

0.7
that keeps us from noticing. Our eyes resolve a fraction of
0.6 the detail outside of a 20° field of view from our center-
0.5 most vision and with much less color information at that.
0.4 Figure 13-4 (overleaf) is an interpretive illustration of the
0.3 visual information detected by the retina. Our central
vision contains the most detail while the surrounding
0.2
periphery quickly decreases in fidelity and color informa-
0.1
tion moving away from the fovea. This raw input data
0
350 400 450 500 550 600 650 700 750 800 is masked as something much better thanks to color
Wavelength (nm)
memory, rapid eye movements and a stage of cognitive
Figure 13-3 V-Lambda curves for photopic and scotopic vision. processing.

267
 s ecti o n 3: O utp ut

of colors. This demonstration gave visual evidence that

white light can be separated into different colors. It was not
until 1704 that Newton put forth a hypothesis to explain
the process by which we see these colors. He speculated
that the retina contains innumerable light receptors, each
of which responds to a specific color stimulus. Thomas
Young rejected this idea nearly one hundred years later
in 1801. Young hypothesized that there are only three
different kinds of light receptors in the retina with each
kind responding to one color of light: red, green or blue.
His theory was rejected and ignored by his contemporaries.
Young’s theory was rediscovered by Maxwell and Helm-
Figure 13-4 An interpretive illustration of what the retina detects, given holtz in 1850.
the high concentration of cones at the fovea and low concentration Despite the many theories that attempt to explain our
of rods along the periphery. Adapted from an illustration found in the
response to color, there remain numerous unanswered
textbook Basic Vision2
questions regarding color vision and the human eye. Some
theories are more useful than others. The oldest, most
All of this established, we’ve so far neglected to point efficient and most persistent theory of color vision is the
out that we’ve got two of these incredible, complex sensing Young–Helmholtz theory of trichromatic color vision. This
organs in our heads. Binocular vision is the result of having theory postulates that there are three kinds of receptors in
two eyes, spaced apart, perceiving the same scene. Unlike the retina that selectively react. We can think of these as
certain animals like chameleons or hammerhead sharks, red or long-wave, green or mid-wave and blue or short-
our perceptual experience of these two eyes is that of a wave receptors that combine in our visual system and
single, fused image. We do not typically see two different produce the perception of all possible colors.
feeds or have the ability to control each eye independently Some unanswerable questions have been raised regard-
(something is medically awry if it does happen). Binocular ing this three-color theory for human vision. For example,
vision provides the brain with subconsciously interpreted the eye distinguishes four fundamental or primary colors—
data because we see the same scene from slightly different colors that are distinct and have no trace of other colors.
points of view. This gives us depth perception, an under- These colors are red, yellow, green and blue and are called
standing of relative object positions and distance, up to the psychological primaries. The Hering opponent color
about 20 feet in front of us. The brain uses other cues (see theory takes this into account.
the discussion on perspective later in this chapter) to infer Whereas the Young–Helmholtz theory is based on
depth and distance beyond 20 feet. That 20 feet distance of three-color stimuli, the Hering opponent color theory is
binocular depth perception could be extended if we could based on the response to pairs of color stimuli. It assumes
somehow move our eyes farther apart than 2.5 inches, the that there are six basic independent colors (red, yellow,
average person’s interpupillary distance. green, blue, white and black). Rather than postulating
special and separate receptors for the six colors, Hering
proposed that the light absorbed by the red, green and
Color Vision Theories blue sensitive receptors in the retina starts a flow of activity
As early as 1666, a 23-year-old instructor at Cambridge in the visual system. Somehow this flow is channeled into
University, Sir Isaac Newton, demonstrated that white three pairs of processes with the two components of each
sunlight is made up of all colors of light. Newton did so opposing one another. The opposing pairs are blue-yel-
by passing sunlight through a prism to reveal a spectrum low, green-red and white-black. For example, a color may

268
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

look bluish or yellowish, never both at the same time. Blue

opposes and cancels yellow and vice versa.
A variation on the Hering theory that provides for quan-
tification is the modern opponent-colors theory (also called a
stage theory). This theory assumes a two-stage process. The
first is the excitation stage, occurring in the cones of the
retina and consisting of four light-receiving cells that con-
tain combinations of three photochemicals. The second
stage is an associated response stage occurring beyond
the retina in the visual nerve center. It has three paired-op-
ponent processes: blue-yellow, green-red and white-black.
This opponent-response theory represents differences in
the neural response to the stimulation that originates when
the cones in the retina are excited. A single nerve transmits
two different messages by one of the pairs of opponent
colors, raising the neural firing rate above the normal rate Figure 13-5 An Ishihara test plate containing the number 12.
(excitation) and the other lowering the firing rate below
the normal rate (inhibition).

Testing Color Vision

Take a moment and think about how color plays a role in
your day-to-day experiences. Whether it’s selecting clothes
in the morning that match, picking a ripe banana for
breakfast or telling the difference between a red and green
traffic light on the drive to work, the functional role of color
is undeniable. It’s particularly important to photographers.
Color vision screening begins as early as 4 or 5 years old
when children learn to identify colors.
How do we test our color vision or the possible lack
thereof? One common method is the Ishihara Color Vision Figure 13-6 The Farnsworth–Munsell 100 Hue test trays.
Image courtesy of © X-Rite Incorporated | www.xrite.com
Test, also referred to as pseudoisochromatic plates. The test
is named after its creator, Dr. Shinobu Ishihara (1879–1963),
who published the test in 1917. The Ishihara Color Vision
Test screens for red-green color deficiencies. The test con- of the Ishihara test that use shapes instead of numbers so
tains between 14 and 38 color plates. Each plate is made that young children can trace the figures or identify the
up of many dots that vary in size, brightness and color shapes.
arranged in a circle, an example of which is shown in Figure The Farnsworth–Munsell 100 Hue Test (FM-100 Hue Test)
13-5. The dots are placed in such a way that a person with is a more extensive method of evaluating color vision
normal color vision is able to identify the digit number and helps to identify all types of color vision issues. The
within the circle. A person who is colorblind or has a color FM-100 Hue Test consists of four separate trays of color
deficiency sees either no number or a different number caps containing disks of varying hues as shown in Figure
than a person with normal color vision. There are versions 13-6. Each tray contains two disks that are fixed in place.

269
 s ecti o n 3: O utp ut

The test administrator removes the caps from the tray and The red, green and blue cone sensitivities suggest a
mixes their order. The test taker is tasked with placing the simple trichromatic theory of color vision, though much of
caps back into the tray in the proper order whereby the the experimental evidence supports the opponent theory
color gradually changes from one end of the box to the of color vision whereby information from the red-, green-
other. A quantitative score is calculated upon completion and blue-sensitive cones is thought to be transmitted in
of all four boxes that quantifies an individual’s level of color combinations—specifically red-green, blue-yellow and
discrimination. black-white—through three separate channels.
Aside from missing one or two of the three types of
cone photopigments, defective color vision can also be
Color Vision Deficiencies associated with reduced sensitivity of one or two of the
Not everyone has normal color vision. Persons with normal three cone types or with a shift in sensitivity along the
color vision are identified as normal trichromats based on spectrum for one or more of the cone types. People who
the three types of cones in the retina. Even then, not all have all three types of cone photopigments but who
normal trichromats respond to colors in exactly the same do not have normal color vision for either of the reasons
way. For this reason, scientific studies of color vision make cited are referred to as anomalous trichromats—as dis-
use of the average response of a number of people with tinct from the conventional trichromats, dichromats and
normal color vision. monochromats.
There are a number of types of defective color vision Accurate color identification requires normal color
despite the common catch-all term “colorblind.” A person vision, standard viewing conditions and the opportunity
who is missing one of the three types of cone pigments to make side-by-side comparisons with standard colors.
is known as a dichromat. Monochromats are missing two A person with normal color vision may have inaccurate
of the three cone pigments (or possibly have rhodopsin, perception of stimulus colors under certain conditions
the rod photopigment, in the cones). Dichromats have such as:
difficulty distinguishing between red and green, or more
rarely between blue and yellow. There are very few mono- • when the image is formed near the periphery of the
chromats, but their vision is the equivalent of black and retina;
white photography. Figure 13-7 simulates how each of • when the light level is very low or very high;
these color vision deficiencies renders the perception of • when the stimulus is small in area;
colors. • when the stimulus is presented for a very short time;
• when the stimulus is illuminated with something other
than white light;
• when the viewer is adapted to a different color.

Heredity is responsible for most cases of defective

color vision. It may also result from other causes such
as the use of certain drugs, excessive use of alcohol and
brain damage. About 8% of Caucasian males and 0.4% of
females across all ethnicities have some form of defective
color vision. There is no cure for congenital defective
color vision. Some people whose occupations require
being able to discriminate between certain colors have
Figure 13-7 Simulations of a) normal color vision, b) protanopia and
been helped by using color filters over the eyes. Pho-
c) deuteranopia. tographers with defective color vision are able to make

270
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

color prints that are acceptable to themselves, but the blue light present. Additive color mixing is demonstrated
prints often do not appear correct to people with normal by using three identical projectors in a darkened room,
color vision. each equipped with a different colored filter (red, green
A company called Enchroma® developed a pair of and blue) aimed at a projection screen. Equipping each
glasses that aid those with red-green color blindness. projector with a variable aperture allows the illuminance
Those with red-green color blindness have the mid-wave of each color to be adjusted to create brighter or darker
and long-wave cone responses overlap more than normal. colors. Overlapping the blue and the green lights and
Enchroma® glasses have a notch filter that helps the brain altering the illuminance of each of the lights produces
separate these two signals. This technique restores some a series of blue-green or cyan colors. Adding red light
color vision. We have seen several of our students try to the system and overlapping it with the blue light
these glasses and the results have been truly amazing and results in a series or red-blue or magenta colors. Likewise,
life-changing for them. overlapping the red light with the green light produces a
series of yellow colors.
Figure 13-8 shows the results of such a demonstration.
Methods of Creating Color That cyan is formed where the blue and green overlap
is not surprising, nor is the formation of magenta from
All color reproduction systems operate on the assumption the combination of blue and red light. Visually, the colors
that the human eye contains three different types of color produced are consistent with the contributions made by
receptors. Many experiments show that almost all colors the primary colors. However, for the mixture of red light
can be produced through the appropriate mixture of red, and green light to appear yellow is an amazing result when
green and blue light. Although many theories of color yellow does not seem to resemble the red or green light.
vision have been proposed, the trichromatic theory offers This phenomenon is a consequence of the three cone
the most satisfactory explanation of color perception as sensitivities. To perceive yellow, the M (green) and L (red)
it relates to color photography. For a color photographic cones must be excited. The sensitivity responses of these
system to be successful, it must be capable of recording two cones overlap each other as shown in Figure 13-9
and controlling the red, green and blue components of the (overleaf). White light results at the spot where all three
light being reflected, transmitted or emitted by the subject. colored light beams overlap.
There are two basic methods for controlling the color of
the light produced in a color reproduction system: additive
color mixture, which involves combining red, green and
blue light; and subtractive color mixture, which involves
colorants that selectively absorb red, green and blue light.
Reproducing color captured in photographs on displays
and in print is the key to creating breathtaking images.
Mastering color reproduction means exploring additive
and subtractive color mixing along with how partitive
mixing plays a role in both systems.

Additive Mixing
Computer screens and digital image sensors use addi-
tive systems to produce a wide variety of colors. They
do this by controlling the amounts of red, green and Figure 13-8 The additive color system: red, green and blue.

271
 s ecti o n 3: O utp ut

L Cone
1 M Cone
S Cone

0.8
Cone Sensitivity

0.6

0.4

0.2

0
400 450 500 550 600 650 700
Wavelength (nm)

Figure 13-9 Cone sensitivity functions. Figure 13-10 The Maxwell triangle.

The term additive implies that combinations of colored the three additive primaries at the corners of an equilateral
light produce additional colors. All additive color repro- triangle and proposed that all possible colors exist at points
duction systems require the use of the three colors within the triangle as a result of different primary color mix-
represented in this example: red light, green light and blue tures (see Figure 13-10). The Maxwell triangle set the stage
light. Consequently, these colors are referred to as the addi- for many color systems to follow.
tive primaries. Table 13-1 lists the color combinations that There is another approach that employs the principles
result from these additive color primaries. of additive color mixing to create a spectrum of colors. A
Practically any color is reproducible by controlling the single projector projecting red, green and blue images in
amount of light coming from each projector. Hue, lightness rapid succession is perceived as a full color image by our
and saturation can all be adjusted by varying the amounts eyes. The rapid switching between each red, green or blue
of red, green and blue light falling on the screen. The only version of the image is not distinguishable by the human
colors that can’t be produced are spectral colors. These are visual system and instead appear as one. This percep-
highly saturated colors that are not easily produced with tual phenomenon is called persistence of vision and is the
our projector example. result of temporal color mixing since the colors are created
Recall from Chapter 5 that James Maxwell proposed the with the help of lapsed time. Digital Light Processing (DLP)
concept of the Maxwell triangle in the 1870s. He placed projection systems, sometimes used in movie theaters, use
temporal color mixing to produce full color imagery.

Table 13-1 Additive color formation (the mixing of lights).

Lights Produce Subtractive Mixing

Blue light + green light Cyan
Printing photographs is unique from displaying them with
monitors or projectors because it requires subtractive color
Blue light + red light Magenta
mixing. Subtractive color mixing is characterized by the
Green light + red light Yellow
mixing of colorants such as inks, dyes, pigments and paints.
Blue light + green light + red light White Although subtractive mixing involves a fundamentally
No light Black different set of materials compared to additive mixing,

272
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

the approach is capable of producing nearly the same colorant absorbs the blue portion of the white light and
range of colors. Subtractive mixing colorants control the the cyan colorant absorbs the red portion of the white
red, green and blue portions of white light that reflect off light, leaving only the green to reflect back to our eyes.
of the image surface to render faithful reproductions of Similarly, where the yellow and magenta colorants overlap,
our photographs. There are three colorants that meet this the color red appears. The blue light is absorbed by the
requirement: cyan, magenta and yellow. Cyan absorbs red yellow colorant and the green light by the magenta color-
light and reflects or transmits blue and green light, creating ant, leaving only the red light to be reflected to our eyes.
its blue-green appearance. Magenta absorbs green light Finally, where the cyan and magenta colorants overlap, the
and reflects or transmits blue light and red light, causing its color blue appears. The red light absorption by the cyan
blue-red appearance. Yellow absorbs blue light and reflects colorant and the green light absorption by the magenta
or transmits red and green light. These subtractive primaries leave only the blue light component of the original white
can also be described in terms of what color they subtract light to be reflected to our eyes. A wide variety of hues
from the incident white light: are reproduced by mixing varying amounts of these three
colorants. If all three are mixed in equal amounts, a neutral
• cyan = minus red results. Table 13-2 provides the possible results of mixing
• magenta = minus green colorants.
• yellow = minus blue A fourth colorant is used in many subtractive printing
systems: black. These systems are referred to as CMYK with
Figure 13-11 illustrates the results of mixing of cyan, K representing the black ink (it stands for “Key” and comes
magenta and yellow colorants. We assume that the from printing press terminology). The use of a black col-
receiving substrate (typically paper) is illuminated by white orant provides two benefits. First, it is difficult to produce
light which contains nearly equal amounts of red, green a true black using cyan, magenta and yellow inks. When
and blue light. A white sheet of paper should reflect that mixed, the resulting color is often a muddy gray-brown
light and appear white before anything is printed onto it. instead of a rich, neutral black due to colorant impurities.
Where the yellow colorant is mixed with the cyan colorant, Second, these three colored inks are expensive, meaning
the area of overlap produces the color green. The yellow that the practice of using all three to make black is a costly

Table 13-2 Basic facts of subtractive color formation (mixing of

colorants).

Colorant Absorbs Produces/Reflects

Cyan Red Cyan (blue-green)

Magenta Green Magenta (blue-red)

Yellow Blue Yellow (green-red)

Cyan + magenta Red and green Blue

Cyan + yellow Red and blue Green

Magenta + yellow Green and blue Red

Cyan + magenta + Red, green and Black

yellow blue

Figure 13-11 The subtractive color system: cyan, magenta and yellow. No colorant Nothing White

273
 s ecti o n 3: O utp ut

one. Adding a separate, no-mixing-needed black ink solves shirt and pair of pants, are metamers or a metameric pair.
both problems. It ensures a pure black and is much less Despite being manufactured independently with differ-
expensive to produce. Some inkjet printers further subdi- ent fabrics or dyes, the two articles of clothing appeared
vide the primary colorants to include dedicated cartridges as the same color in the store. These differences become
like Light Magenta and Light Cyan. apparent only when the available light source changes
the color reproduction equation. Figure 13-12 shows
metameric fabrics next to one another under different
Metamerism’s Role in Color Reproduction illuminants.
Have you ever bought a shirt and a pair of pants thinking What makes two objects appear to be the same color
they are a perfect match, only to wear them together and despite being different materials? The colors we perceive
realize that they look completely different in the light of are governed by three components. The first is the object’s
day? This fashion embarrassment occurred because you spectral reflectance: the amount of energy reflected at all
viewed the clothing under the store’s lighting and then visible wavelengths. The second component is the spec-
later under different illumination. They matched in the tral power distribution of the light source under which
first environment and not the second due to metamer- we view the object. The third component is our vision.
ism. Metamerism is the phenomenon of perceiving two As discussed earlier in this chapter, our retinal anatomy
objects as the same color when viewing them under includes three cone types (long-wave, mid-wave and
an illuminant even though they have different spectral short-wave) with overlapping sensitivities. All observed
reflectances. These two objects, in this example the colors are therefore reduced to three sensory quantities,
one from each cone. These three quantities integrate
together when we perceive color. An object’s spectral
reflectance and our vision remain the same while the light
source can vary. When the spectral power distribution of
the dominant light changes and the color we perceive
changes with it, two seemingly similar colors may reveal
themselves to be otherwise. The example of the shirt and
pair of pants describes two materials that do not have
the same spectral reflectance yet the store’s lighting lead
you to perceive them as matching. The change in illumi-
nants when you brought the clothes home revealed the
underlying disparity. Measuring the spectral reflectance of
objects is accomplished using a spectrophotometer under
multiple illuminants with distinct spectral power distribu-
tions such as Standard Illuminant A and D65.
Consider how the appearances of Sample 1 and Sample
2 change when the lighting is changed (see Figure 13-13).
The spectral reflectance of the samples intertwines up
until approximately 620 nm. Beyond that, their reflec-
tance’s separate and end up roughly 40% different from
each other. When the samples are viewed under daylight
illumination (D65), the amount of energy above 620 nm
Figure 13-12 Three fabric samples that change appearance under
is insufficient to show the sample differences. A lack of
different light sources. red energy means that the samples cannot reflect it by

274
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

80 80 80 80

70 70
70 70

60 60
60 60
Spectral Response
Spectral Response

Spectral Response
Spectral Response
50 50
50 50

40 40

40 40
30 30

30 30
20 20

Sample
Sample
1 1 Sample
Sample
1 1
20 20 Sample
Sample 2 2 Sample 2 2
Sample
10 10
D65D65 Illuminant A A
Illuminant

10 10 0 0
400400 450450 500500 550550 600600 650650 700700 400400 450450 500500 550550 600600 650650 700700
Wavelength
Wavelength
(nm)
(nm) Wavelength
Wavelength
(nm)
(nm)

Figure 13-13 The spectral reflectance’s of a metameric pair compared under two different illuminants.

different amounts, thus rendering their appearances simi- Partitive Mixing

lar. The amount of energy present around 620 nm is much Additive and subtractive mixing rely on spatial color
greater when the samples are viewed under Illuminant mixing, also called partitive mixing. In partitive mixing, small
A (tungsten). Sample 1’s reflectance at this portion of amounts of red, green and blue lights or small amounts of
the spectrum means that it appears quite different from cyan, magenta and yellow pigments are placed side by side
Sample 2. to form an image. The individual points of light or ink drops
Although metamerism can cause us frustration when are so small that the eye cannot readily distinguish them
trying to match clothes, its role in color photographic as distinct colored spots at a normal viewing distance. A
output and color management is an essential one. We classic example of partitive mixing in art history is painter
reproduce recorded colors of our world and display them George Seurat’s pointillism works.
on monitors, prints, fabrics and various other media with Partitive mixing is employed in televisions and computer
a relative degree of success because metamerism allows monitors. These displays create images using small pixels
us to perceive the same colors through different means. In filtered to be red, green or blue (the additive primaries).
other words, without metamerism, the use of subtractive Their intensities are electronically controlled to reproduce a
mixing and ink colorants does not work to recreate the range of perceived colors and tones. The smaller the pixels
colors we see when taking photographs. We can pho- or the greater the viewing distance, the less likely we are to
tograph and print an image of green grass in a field that perceive them as independent points of color. Displays use
appears true to life in its color without needing the grass fixed pixel sizes as they cannot dynamically adjust in size or
itself and the sun shining on that print thanks to color shape.
mixing and metamerism. For subtractive color, small dots of inks are deposited on
paper or other substrates at varying densities. These dots
are printed in precise patterns and amounts that appear

275
 s ecti o n 3: O utp ut

Figure 13-14 A continuous color photograph (left) reproduced with a CMYK ink pattern (right).

as continuous, full color images when viewed from the method is defined by color mixing. This system provides
appropriate distance (magnified for illustration in Figure the observer with a discrete set of color sample chips to
13-14). Ink dot sizes can vary in radius depending on the choose from. When a selection is made, documented
capabilities of the printer and the pigment or dye used. information from that chip is used to create that color.
Printing on a substrate like white paper means that any That information includes the percentages of different inks,
areas not covered with ink influence the perception of pigments or paints that must be mixed to produce a color
color and tone; varying the distances and spaces between to match the selected chip.
dots adds an additional parameter in partitive mixing that The second system for object-color specification is a
is not available in fixed-pixel, additive color displays. Taking color-order system defined by human perception. Color
a close look at both monitors and printed materials reveals chips are spaced so that they appear to have uniform
these dense but distinct units of color that form our photo- intervals or differences in hue, saturation and lightness. This
graphic reproductions. requires judgment by a standard observer under a standard
set of viewing conditions. The Munsell Color System is a
color-order system that consists of hundreds of samples; it
Defining Color with Color continues to be useful for visual artists, agriculture and food
Specification Systems industries and even government standards.
The third system is defined by matching perceptions
Color specification systems define color spaces, color to a standardized set of lights. An example is the CIE Color
models with known references and scales. There are three System that specifies color in terms of three parameters:
basic systems for describing or specifying a color. The first Y, x and y. These parameters are calculated based on the

276
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

visual response of a standard observer, the spectral power Some samples are available on several different surfaces
distribution of the light source and the spectral reflectance and finishes.
of the sample. The CIE system is used extensively in color The Pantone system is used in printing, plastics, fash-
management systems and is discussed in greater detail ion, home and color evaluation industries. It’s great for
later in this chapter. identifying and using specific colors for designs, logos
and graphics made from scratch, though we don’t have
the luxury of hand-selecting colors in quite the same way
The Pantone System of Color when photographing real-world objects. One shortfall of
Specification the Pantone system is that not all colors are available. Only
those found in the Pantone Sample book can be used. As
A commonly encountered example of a color mixing a consequence, it’s not possible to make an exact color
system is The Pantone Matching System®. You may be match if we convert a photographic image to a Pantone
familiar with this approach if you’ve ever gone to a home color space.
improvement store and picked out a sample that is mixed
by an employee on the spot. It is a functional system
that uses prepared samples and precise recipes that help The Munsell System of Color
create the color you pick out every time. The Pantone Specification
Matching System® is widely used in the graphic arts and it
uses inks as the colorants. Using colorants similar to those Albert Henry Munsell was a painter and an art teacher. He
used for printing provides a closer match between the found communicating about color with his students to be
colors selected and specified by the client and the colors frustrating and sought a scientific approach to make things
produced by a printer. The Pantone Matching System® easier. The result of his explorations is the Munsell Color
includes a series of books containing thousands of printed System. Munsell went on to publish A Color Notation in 1905
color samples and their formulas as shown in Figure 13-15. and the Atlas of the Munsell Color System in 1915 based on
his explorations of organizing color and describing color.
The Munsell System for color identification is one of the
earliest attempts to create an accurate system to numeri-
cally describe color. One of the Munsell System’s strengths
is in providing a common notation for color in which a
person first identifies a color visually and then uses lan-
guage to describe it. To accomplish this, Munsell prepared
a large variety of painted color chips. Munsell described
hue as “the name of the color” or “the quality by which we
distinguish one color from another, as a red from a yellow,
a green, a blue or a purple.” He defined value as “the light
of color” or “the quality by which we distinguish a light
color from a dark one.” Lastly, he described chroma as “the
strength of a color” or “that quality by which we distinguish
a strong color from a weak one.”3
Munsell started with a collection of chips spanning a
wide range of colors. He first grouped the chips into ten
Figure 13-15 An example Pantone color sample book. Image courtesy
named major hues: red, red-yellow, yellow, yellow-green,
of © X-Rite Incorporated | www.xrite.com green, green-blue, blue, blue-purple, purple and

277
 s ecti o n 3: O utp ut

Figure 13-16 The Munsell Color System considers hue, value and Figure 13-17 The Munsell color tree. Image courtesy of © X-Rite
chroma along different axes. Adapted from Jacob Rus, released for Incorporated | www.xrite.com
free use under a Creative Commons license

red-purple as shown in Figure 13-16. He then arranged color tree (see Figure 13-17). He wrote: “The color tree . . .
the chips in each hue group by value from darkest (0) to carries a scale of value, for the trunk. The branches are at
lightest (10). The third and most difficult sorting stage was right angles to the trunk . . . carry the scale of chroma . . .
arranging the hue chips with equivalent lightness accord- branches tell their hue.”4 The hues change as one walks
ing to their color intensity or chroma. Chroma values in around the tree, the values increase as one climbs up the
the Munsell System range from 0–20 for normal reflective tree and the chromas increase as one moves out along
materials but can be as high as 30 for fluorescent materi- the branches.
als. A color chip that appears nearly neutral sits on one end The overall shape of the color volume that Munsell
while the opposite end has a vibrant or highly saturated creatively imagined as a tree is not symmetrical. This is
instance of a given hue. because the range of chromas is not the same for all hues.
Munsell also made one critical refinement from For example, it’s possible to have a yellow hue of high
there. He ordered the colored chips so that the inter- value and high chroma but not one of low value and
val between adjoining chips was visually equal in hue, high chroma. Concurrently, a blue hue of high value and
value and chroma—a monumental task. With this visual high chroma is not attainable. As a result, the blue and
structuring completed, he had to communicate it to yellow portions of the tree do not have the same shape.
others so they could also visualize the arrangement and The three-dimensional, asymmetrical color volume of the
adopt a common language to describe the colors. The modern Munsell color tree represents stable and repro-
three attributes of color as defined by Munsell (hue, ducible colorants available today, excluding fluorescent
value, chroma) occupy a three-dimensional space or colors. The Munsell space is expanded as new colorants
volume. Munsell described this volume in terms of a become available.

278
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

as R). Another example, R 5/12, translates to a strong red

because its value is 5 and its chroma is 12. At the center of
the Munsell tree are the neutrals. The notation for a neutral
(achromatic) color is written NV/; for a middle gray (18%
gray) the notation is N5/. Since the chroma is zero, it is
usually omitted.
Near-neutrals have chromas less than 0.3 and are usually
treated as neutrals. If more precision is needed, the format
N V/(H,C) is used where H is represented by one of the
ten select hues. For example, a light gray with a slightly
yellowish cast might be noted as N 8/Y,0.2 or, using the
regular H V/C form, Y 8/0.2. The form N V/0 can be reserved
for absolute neutrals.

The CIE System of Color

Figure 13-18 A top-down view of Munsell’s color tree highlights the
radial positions of different hues. Specification

The International Commission on Illumination (CIE, based on

There are 100 unique hue groups represented in the the French name Commission Internationale de l’Eclairage)
Munsell system though only a select sample of ten are Color Specification System became an international standard
actually shown in the color tree. The unseen hues exist for colorimetry in 1931. Color specification with the CIE
between those ten with the full 100 hues being equally system differs from Munsell’s in two important ways. First, it
spaced in terms of visual differences, illustrated in Figure uses mixtures of red, green and blue light to match a given
13-18. sample rather than color chips. Second, the CIE system is
Figure 13-18 shows how the ten major hue names are based on stimulus-synthesis (the additive mixture of three
used for qualitative notation. The numbers along the circle color primaries) whereas the Munsell system is based on
provide more precise notation. These numbers make it appearance. Both systems have advantages and specifying
easy to use the Munsell system for statistical work, cata- a color sample with one system allows conversion and
loging and computer programming. The combination of notation to the other. The obvious advantage of the Mun-
numerals with hue initials is considered the most descrip- sell system is its simplicity and its directness. The selection
tive form of designation. This combination is shown in the of one of the color chips provides a color and color notation
inner circle: 5R, 7.5R, 10R, 2.5YR, 5YR, 7.5YR, 10YR and so directly in terms of hue, value and chroma; it is a physical
on for a total of 40 named hues called constant hues. The match easily made and easily understood. The CIE system
Munsell Book of Color provides over 1,600 samples of these requires instrumentation and is based on mathematical
40 constant hues. conventions, making it abstract and more difficult to con-
The Munsell system specifies colors with the alphanu- ceptualize. It has the advantage, however, that any color can
meric sequence Hue Value/Chroma (H V/C). For example, be matched, numerically specified and positioned on a CIE
a 5R hue with a value of 8 and a chroma of 4—a light diagram or map. The CIE system provides psychophysical
pink—is designated as 5R 8/4 or simply R 8/4. Hues measurements of color whereas the Munsell system, based
positioned anywhere other than at the 5 position along on color appearance, provides psychological or perceptual
the hue circle must be explicitly noted, but those at 5 can measures. The CIE color specification system is used exclu-
leave the number off (as it’s assumed to be 5R if it’s written sively for color management applications.

279
 s ecti o n 3: O utp ut

The CIE color system specifies colors using three param- Δλ = measurement wavelength interval
eters: Y, x, y. The tristimulus value Y is a luminance value, ∑λ = summing of the products of the multiplications
representing brightness. The x and y terms are chromaticity over the wavelengths measured
coordinates. Chromaticity coordinates are plotted on a chro- k = normalizing constant
maticity diagram that offers insight into the chroma and hue
qualities of a color sample. Chromaticity coordinates are Chromaticity coordinates are calculated once the tristimu-
derived from the tristimulus values of the sample described. lus values are obtained using the formulas in Equation 13.2.
There are three tristimulus values: X, Y and Z. Tristimu- Note that x + y + z always evaluates to 1.
lus values are derived by integrating three components
X Y Z
needed to see a color (see Equation 13.1). The first com- x= y= z= (Eq. 13.2)
X+Y+Z X+Y+Z X+Y+Z
ponent is the spectral power distribution (S ) of the light
source used to illuminate the color sample. Spectral power
distributions are measured with a spectroradiometer. The
CIE defined several standard light sources that are typically The CIE Chromaticity Diagram
used in tristimulus calculations. The second component is Chromaticity describes two attributes of color: hue and
the spectral reflectance (R ) of the color sample. Spectral chroma. These two qualities, called dominant wavelength
reflectance is measured with a spectrophotometer. The and purity in the CIE system, are plotted on a horse-
third component is the observer. The CIE quantified the shoe-shaped map called a chromaticity diagram shown in
response of the standard observer through well-docu- Figure 13-19. This diagram represents all of the colors vis-
mented experiments to produce the CIE color-matching ible to the average observer. Around the periphery of the
functions used to calculate tristimulus values. curved section are wavelengths of visible light scaled from
It’s worth noting that the CIE provided the illuminant 380 to 770 nm. The x- and y-axes are used to plot the posi-
and color-matching function data in one nanometer incre- tion of the chromaticity of a particular color. Importantly,
ments from 360 to 830 nm. Ideally, we want our sample the color is always plotted relative to a specific light source
spectral reflectance in the same one nanometer intervals under which it is viewed. In this example, the daylight
over the same wavelength range. Most spectrophotome- source is located in the lower middle of the chromaticity
ters are not designed this way and instead measure from diagram. It is located on the Planckian Locus, a plotted line
400 to 700 nm in increments of 10 or 20 nm. The CIE uses representing all light sources that are incandescent black-
a standard method for altering their data to match for the body radiators. This serves as a neutral reference point.
calculation to work when using these spectrophotometers. The farther a color plots from the reference neutral, the
greater the purity (chroma). A green sample is identified
X = k∑λSλRλ xλ∆λ (Eq. 13.1) by the coordinates x = 0.22 and y = 0.52. Not shown is the
luminance (brightness) of the green. This value is obtained
Y = k∑λSλRλ yλ∆λ
separately from the data used to calculate the x and y
Z = k∑λSλRλ zλ∆λ coordinates. The dominant wavelength (hue) is deter-
100 mined by drawing a line from the reference point through
k=
∑λSλ yλ∆λ the plotted point to the curved wavelength line. The green
filter shown has a dominant wavelength of 520 nm.
Where: The chromaticity diagram is also valuable for visualiz-
X, Y and Z = tristimulus values ing color gamut. The gamut is the range of reproducible
Sλ = spectral power distribution of the illuminant colors for a given system or device and it’s most commonly
Rλ = object’s spectral reflectance a subset of the perceivable color space (CIE space). The
–x , y–, –z = CIE standard observer color-matching functions chromaticity diagram as a whole represents the gamut of

280
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

CIELAB Space
The CIE 1931 chromaticity diagram in Figure 13-19 provides
a convenient mapping space for displaying chromaticity
coordinates and color gamuts, however, color differences
displayed by these coordinates are not uniformly distrib-
uted. Similar color differences between two samples of a
reddish or bluish color are compressed, whereas the same
visual difference greenish colors appear expanded in the
diagram. This non-uniformity of plotted color differences
was reported in research done by David MacAdam in 1942.
MacAdam used a large number of color difference pairs
to determine how much of a color change, in any direc-
tion, could be made before an observer perceived a just
noticeable difference (JND). He found that the variabilities in
JNDs were different for each color and that plots of these
variability boundaries formed similar elliptical shapes that
varied greatly in size from one another. These experimental
results helped describe the perceptual non-uniformity of
Figure 13-19 A chromaticity diagram. The inset triangle represents a the 1931 CIEXYZ chromaticity diagram and are known as
color gamut and the dot in the center represents a light source. The
the MacAdam ellipses.
Planckian Locus is the thin line upon which the example light source
sits. The CIE recommended the use of a more uniform color
difference formula and related color space in 1976. They
designated CIE L*a*b* (or CIELAB) to provide better cor-
human vision. A greater number of colors can be shown relation between visual and measured color differences.
with a larger gamut while a small gamut limits the possi- The use of CIELAB provided a significant improvement
ble colors to a smaller number and range. Mapping the over the non-uniformity of the CIE tristimulus space by
primary colors used by an output device (like a display) on deriving metrics that correlate to perceptions of color.
the chromaticity diagram helps to identify its reproducible CIELAB, based on a simple color vision model, can adjust
colors. The inset triangle in Figure 13-19 maps the chro- for changes in the color and amount of illumination, signal
maticities of the red, green and blue pixels of a computer compression and opponent signal processing. The CIELAB
monitor. Drawing straight lines to connect the three addi- color space does not have an associated chromaticity dia-
tive primary chromaticity points reveals the monitor’s color gram because of its exponential non-linear derivation from
gamut as a two-dimensional area. Any color that sits inside CIE tristimulus space. However, a three-dimensional color
of this triangle can be accurately displayed. Photographs space can be constructed with L* as the vertical axis and a*
don’t always contain a wide range of colors and this gamut and b* occupying horizontal planes.
may be enough to reproduce all colors in the image. The L* value represents luminance and ranges from
However, you may find that a vibrant red dress doesn’t 0 (black) to 100 (white). The a* value represents the
look quite so saturated on screen compared to seeing it in redness-greenness of the sample. The b* value repre-
person. In this example, the image content includes reds sents the yellowness-blueness of the sample. There is no
that fall outside of the monitor’s reproducible gamut in the numerical limit on a* or b* values and they can be either
red portion of the color space. We’ll explore this further in positive or negative. It is customary to plot only the a*
Chapter 14’s look at managing these color reproduction and b* values, resulting in a two-dimensional graph. Color
challenges. gamut comparisons are easily constructed using the a*

281
 s ecti o n 3: O utp ut

distance on perspective in Chapter 2. Perspective gives us

a sense of relative sizes, distances, position and general
spatial awareness of three-dimensional scenes. There are
additional ways to convey perspective and affect the per-
ception of subject matter in photographs. Light adaptation,
linear perspective, color, shape, size, depth of field, overlap,
aerial haze and texture all play a role.

Light Adaptation
First, let’s explore how our eyes adapt to differences in
lighting in ways that affect our perception of objects and
potentially make our vision an unreliable instrument. Earlier
in this chapter we introduced the functions of photopic,
mesopic and scotopic vision for adapting to different light
levels. When taking light adaptation into account, the
Figure 13-20 An a* versus b* CIELAB plot with a red, green, blue,
cyan, magenta and yellow color gamut. ratio of luminances to which the visual system responds is
truly amazing. A white object in sunlight has a luminance
approximately 100 million times that of a white object in
(redness-greenness), and b* (yellowness-blueness) horizon- starlight, yet both objects are easily seen when a person
tal opponent measurement planes (see Figure 13-20). is fully adapted to each light level. A person can detect a
Specifying color in terms of CIELAB provides a means flash of light with a luminance 1/100th that of white paper
for various color specification systems used in photogra- in starlight under optimal adaptation conditions. A person
phy to communicate with one another across media can detect luminance differences up to 1,000 times that
components. Regardless of the input or output device of white paper in sunlight when adapted to bright light,
used—camera, monitor, scanner, printer—and how the though this approaches luminances that can damage the
colors are being specified, all can feed into a CIELAB-based retina. The total response range of the visual system is
device-independent reference color space. This allows for approximately 10,000,000,000,000:1 (ten trillion to one), or a
compatible conversions across all existing components log luminance difference of 13.
of a digital color reproduction system. CIELAB’s L* and a*, It’s important to distinguish between luminance, which
b* component performance as a simple color appearance is psychophysical and measurable with a light meter,
model makes it useful in color management systems. How- and the perception of brightness, which is influenced by
ever, due to its still-not-perfect uniformity, some software physiological and psychological factors and is not directly
vendors prefer to implement their own color space config- measurable. The eye is not a dependable instrument for
urations based around CIEXYZ tristimulus values. measuring luminance values. It’s difficult to judge whether
a black surface in direct sunlight is reflecting more or less
light than a white surface illuminated with dim, incandes-
The Roles of Light and Color on cent indoor light, for example. This difficulty stems from
Perception and Perspective two variables: reflectance of the two surfaces and the
amount of light falling on them. The adaptation level of
We round out our understanding of vision by outlining the visual system affects perception in that a surface with
how the visual system plays a part in perception and per- a fixed luminance appears lighter when the eye is dark
spective. We described the effect of focal length and object adapted than when it is light adapted. Also, a gray tone

282
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

match the brightness or lightness of two areas if they differ

in hue. If, for example, the inner circle in a visual densitom-
eter is red and the outer ring is blue, the viewer struggles
to identify when the two match in brightness. Matching
the brightness of differently hued samples requires the
use of a flicker photometer. This device presents the two
samples to the viewer alternately in rapid succession.
A flickering is observed when the two are not closely
matched in brightness; the flickering ceases when they
match.
Figure 13-21 Lateral adaptation causes the two boxes, both the
same shade of gray, to appear brighter or darker depending on their
surround.
Color Contrast
Contrast is defined as the difference in luminance between
appears lighter in front of a black background than in front two tones. Color contrast is the difference in luminance
of a white background, an effect known as lateral adap- between two adjacent colors. This applies to overlaid colors
tation or simultaneous contrast as shown in Figure 13-21. as well, such as text on a color background. Colors that are
Simultaneous contrast can impact the appearance of hue, complementary to each other, such as green and magenta
saturation and lightness. shown in Figure 13-23, have a high color contrast when
Conversely, the eye is excellent as a null instrument used together. Colors that are adjacent to each other, such
in detecting tiny luminance differences in side-by-side as magenta and red, have low contrast. Color contrast
comparisons. Figure 13-22 illustrates an old style of visual influences the use of graphics, computer interfaces, web
densitometer. These devices were quite accurate and took design and should be considered by photographers for
advantage of the side-by-side comparison strength of the composition and color emphasis when scene and subject
human visual system. The measured sample is seen in a content are controllable.
circle surrounded by a ring that’s adjusted to match the The Web Content Accessibility Guidelines (WCAG)
inner circle in brightness. 2.0 provides web designers with guidelines in color
This method of putting tones next to each other for contrast for accessible web content.5 For example, a
comparison works well for grayscale objects. It’s difficult to minimum contrast ratio of 4.5:1 is recommended for text

A B C
Figure 13-22 Simulated field of view in a visual densitometer. (A) The
dark small circle is the unknown sample density. The larger circle is the
matching field set at 0 density. (B) The density of the matching field
is increased but is insufficient to match the density of the sample. (C)
More density is added to the matching field and it now matches the
density of the unknown sample. The known density of the matching Figure 13-23 Looking to opposite ends of a color wheel finds
field is the density of the sample. complimentary colors with high color contrast, like green and magenta.

283
 s ecti o n 3: O utp ut

on a background. Photography doesn’t have such clearly we can ignore the visual importance of linear perspective.
defined guidelines but knowledge of color contrast can be Making careful use of parallel and converging sight lines
used to create more effective images. in a composition goes a long way toward emphasizing or
indicating scene depth. The composition of Figure 13-24
places the vanishing point near the center of the image
Linear Perspective frame, making that a strong focal point. The eye is drawn
Linear perspective is used by artists whereby the relative to the center and the instinct is reinforced by the bright
size, shape and position of objects are determined by patch far off in the distance.
drawn or imagined lines converging at a point on the Linear perspective is bolstered by the size of objects
horizon. It’s a technique of representing a three-dimen- rendered on our retina. The same object, like the trees or
sional object in a two-dimensional medium. There are the fence posts in Figure 13-24, appears smaller on the
three components to creating linear perspective: a horizon retina when it is farther away. We discuss this further in the
line, a vanishing point and orthogonal lines. The horizon section on size a little later.
line runs across the scene. This is often where the sky If we asked you to precisely describe an object’s
appears to meet the ground in an outdoor landscape. appearance over the phone, you’d likely break it down
The vanishing point is where all parallel lines in the scene in terms of shape, color and size. Different authorities on
appear to converge and sits along the horizon line. The visual perception often disagree about the critical stimulus
orthogonal lines are visual rays that assist the viewer in attributes for visual perception, the relative importance of
connecting all of the edge points in the scene to the each and even the definitions of the terms. For example,
vanishing point. Figure 13-24 illustrates linear perspective the word form generally implies different object qualities
by identifying these three components. Photographers to artists and photographers. Six attributes are particularly
create photographs with linear perspective, sometimes important to photographers: color, shape, depth, size,
without realizing it, simply because the medium records a sharpness and motion.
two-dimensional version of a three-dimensional scene. We
don’t need to worry about drawing any lines or following a
vanishing point—thankfully, as some of us aren’t the most
talented with paper and pencil—but that doesn’t mean

Figure 13-24 This landscape photograph exhibits visual cues of linear perspective (shown in the overlay on the left). The photographer chose
a composition that takes advantage of linear perspective to convey depth and to guide the viewer’s gaze through the scene.

284
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

Color and Perceived Distance analysis of the parts. Experienced photographers have little
We’ve established systems for communicating and organiz- difficulty separating an object from the background (figure
ing color, but what happens when our perception of color from ground) in a photograph by means of background
plays tricks on our understanding of scenes and objects? choice, lighting and depth of field. In contrast, military
Warm colors (reds, yellows and oranges) are approaching camouflage conceals the shapes of objects to make them
colors. Red objects appear closer to the foreground in an appear to be part of the background and therefore escape
image compared to a cool color such as blue, for example. detection.
Cool colors are receding colors in how our brain interprets It’s not always necessary to see the entire outline shape
and perceives them. Using both types, approaching and of a familiar object to identify it and visualize its entire
receding, adds exaggerated depth or visual emphasis. shape. Most people perceive the moon as round when
Recall that we defined chromatic aberration in lenses in there is a full moon and when there is a half-moon or quar-
Chapter 2. Your eye also suffers from chromatic aberration. ter-moon and the shadow side cannot be separated from
The red wavelengths come into focus behind the retina the background (see Figure 13-25). One can conduct a
such that the eye’s lens changes shape to bring the image simple experiment by looking at objects outdoors through
forward. This causes the red object to appear closer. The Venetian blinds, starting with the slats in the fully open
opposite happens for blue wavelengths that come into position and then gradually closing them to determine
focus in front of the retina and have to be refocused back- how small the openings can be before encountering dif-
wards. Red is particularly effective at drawing attention and ficulty in identifying the objects. Similarly, a small number
creating a sense of excitement.6 of dots arranged in a circular pattern can easily be seen as
representing a circle. This effect is known as the principle of

Shape and Dimension

An object’s shape is a description of its outline. Silhouettes
emphasize shape and eliminate or deemphasize other
attributes such as color, form and texture. We depend
heavily upon the attribute of shape for the identification
of many objects. In fact, shape is often the only attribute
needed for a viewer to recognize an object in a drawing or
photograph.
Three-dimensional objects have many shapes because
they are viewable from many different angles. The view-
point chosen by a photographer to provide the best shape
for an object is important even when the object is lit to
provide detail. This choice is particularly critical in a silhou-
ette. Silhouettes of people are commonly recognizable
only in a full left or right profile view; profile views are used
for the images of famous people on coins.
Photographers control the emphasis on object shapes
by controlling the separation or contrast between object
and background. The term figure-ground refers to the sub-
ject of a picture and the surrounding area. Figure-ground
Figure 13-25 View of the moon’s north pole assembled from 18
is an important concept in Gestalt psychology, where the images taken by Galileo’s imaging system through a green filter on
emphasis is on the perception of the whole rather than an December 7, 1992. Courtesy of NASA

285
 s ecti o n 3: O utp ut

closure in Gestalt psychology whereby the viewer mentally fixations per line. Eye motion studies reveal that viewers
fills in the spaces between picture elements. A distinc- rarely scan photographs as thoroughly as they do a printed
tion exists between closure and fusion, where the latter page; they fixate on a few points while the mind fills in
describes when the optical-retinal system in the eye cannot shapes and details between these fixation points.
resolve the small discrete elements as in a halftone repro- The accuracy with which we perceive shapes is
duction or a photographic image at small to moderate important in the study of visual perception, but it is not
magnifications. considered as critical in our daily lives as normal color
Reading printed text is an excellent example of how vision and good acuity—which are usually tested before a
the mind fills in spaces between fixation points. The area person obtains a driver’s license or is permitted to per-
of sharpest vision represented by the fovea of the retina is form certain occupational tasks. Studies demonstrate that
very small such that when a person fixates one letter in a it is easy to deceive a viewer about shape under certain
line of type, only a few letters on each side are clearly seen. conditions. Figure 13-26 shows how a straight line can be
It’s possible to read rapidly with only two or three fixations made to appear curved. Under normal conditions, how-
per line because the reader recognizes groups of letters as ever, we are best able to detect changes in images with
familiar words without examining each letter and under- simple geometrical shapes such as straight lines, squares,
stands the meaning of a sentence without examining each circles and triangles (see Figure 13-27). We are also better
word. Printed material containing unfamiliar words and a at making comparisons with superimposed or side-by-side
high concentration of factual information requires more images than with images separated in time or space that
rely on our memory.
Our perception of shapes is complicated by the fact that
image shape changes with the angle and distance of the
object relative to the eye or camera lens. Parallel subject
lines are imaged as converging lines except when viewed
or photographed perpendicularly; tilted circles are imaged
as ellipses. We learn through experience that the paral-
lel lines and circles do not change shape with a change

Figure 13-26 Illusory figures. The long straight Figure 13-27 Which is the perfect square?
lines appear curved because of the influence of
the diagonal lines.

286
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

in viewing angle, so we mentally compensate for linear Size and Relative Size
perspective effects. Shape constancy refers to this stability The perceived size of an object has little relationship to the
of the perceived shape of objects as the viewing or camera size of the image on the retina (or the size of the image in
angle changes. Shape generalization is the tendency to a photograph of the object). A car is judged to be approx-
perceive an irregular shape as a simpler shape—obtuse imately the same size when viewed over a wide range of
and acute angles seen in perspective may be perceived as distances as a result of size constancy. Experiments demon-
right angles, and an ellipse seen in perspective may be per- strate that the accuracy of judging the size of an abstract
ceived as a circle. Memories of perceived shapes can also shape, such as a circle, depends greatly upon being able to
change with time. The simplification of irregular shapes estimate the distance. Accuracy decreases as distance cues
because of memory is called leveling. The exaggeration of are systematically eliminated.
a distinctive feature, such as a small gap in an otherwise A ruler or other measuring instrument is used for a
continuous line, is called sharpening. direct side-by-side comparison when the precise size of

Figure 13-28 Size and distance constancy. For familiar objects, size and distance are
inseparable. For an object to be the same actual size at a farther distance it must be
larger and therefore it appears larger.

287
 s ecti o n 3: O utp ut

an object must be known. As with the perception of other an important role. Arranging a scene so that a nearby
object attributes, the eye is most precise when used as a object obscures part of a distant object provides the
null instrument in making comparisons between adjacent viewer with powerful clues as to the relative distances of
stimuli. It is sometimes necessary to include a ruler beside the objects.
an object when size determination is critical, as is often the
case in forensic photography. In pictorial photographs it is
usually sufficient to include an object of known size with Aerial Haze
the unfamiliar object and to provide good distance cues. There is no controlling the weather, however a hazy day
It’s easy to deceive viewers about the size of objects can actually be useful. The scattering of light that occurs
represented in photographs. Use of a short focal length in the atmosphere caused by haze makes distant objects
lens tends to make foreground objects appear larger appear lighter and less contrasty than nearby objects. Cre-
than normal and background objects appear smaller than ating thick fog and smoke can create the illusion of depth
normal when the photograph is viewed at a comfortable with relatively small differences in distance.
distance. Long focal length lenses have the reverse effect.
Line drawings can also cause viewers to misjudge the rela-
tive length of lines or size of images. Figure 13-28 illustrates Texture
that the Muller–Lyer arrow illusion and the Ponzo railway Texture refers to the small-scale depth characteristics of
lines illusion both contain lines of equal length that are a type that might be felt with the fingertips, such as the
perceived as being unequal. roughness of a wood file or the smoothness of window
glass. Effectively representing texture in two-dimensional
photographs depends largely upon using an appropriate
Depth of Field scale of reproduction, as well as lighting that produces
Depth of field is controlled by the aperture setting and shadows in the recessed areas and highlights in the raised
allows photographers to have all the scene in sharp focus areas. Photographs made through optical and electron
or just a portion of it. Use of a limited depth of field, where microscopes reveal that many seemingly smooth surfaces,
objects in front of and behind the point of focus in the such as writing paper, have a rough texture or heteroge-
image are blurred, creates a stronger appearance of depth neous form when magnified. Conversely, the craters on
or perspective than when the entire scene appears sharp. the moon appear to have a finger-touching type of texture
when photographed from a distance with a small scale of
reproduction.
Lighting
Controlling light is key to successful photography and can
affect perception in images. Depth is emphasized with Proof of Concept: Looking Closely at
lighting that produces a gradation of tones on curved Additive Color Mixing on Displays
surfaces, that produces a separation of tones between the
planes of box-shaped objects and between objects and This is one of our students’ favorite exercises. It ties our
backgrounds, and that casts shadows of objects on the knowledge of partitive mixing and additive color mixing
foreground or background. together. Figure 13-29 shows a set of color patches with
pure red on one end and pure blue on the other. The
patches are assigned to an RGB color space: their three
Overlap channel (red, green and blue) brightness values exist as
Although photographers cannot always control where numeric, digital representations of color interpreted by a
objects within our image fall, object placement can play display. Of course, you’re likely reading this on a printed

288
 c h a p t e r 13 : C o lo r V i s i o n a n d S y st e m s

Figure 13-29 Color patches from different combinations of RGB values to create red, blue and the gradient of colors transitioning between them.

Figure 13-30 Low power microscope views of display showing a red patch (far left), a blue patch (far right) and patches transitioning from one to
the other (middle).

page where that information was translated to a subtrac-

tive CMY space!
The red patch is made using the brightest possible
pixel values in the red channel (255) and the darkest values
in the green and blue channel (0). This recipe means
that there is truly only red data present in the patch. On
the other end of the illustration, the blue patch uses the
brightest possible pixel values in the blue channel and Figure 13-31 A smartphone screen up close, revealing how a white
the darkest values in the green and red channels. The on black user interface icon is rendered on an LCD display. Photograph
by Rochester Institute of Technology alumnus Andy Kempchinsky
patches sequenced between red and blue are created by
simultaneously increasing the blue channel brightness and
decreasing the red channel brightness. The exact middle that the partitive mixing effect is lost. The black gaps and
point in this color mixing exercise creates purple: an even individual pixel elements are not visible at the proper
mix of red channel brightness (128) and blue channel viewing distance. Try squinting or backing away from the
brightness (128). printed page to confirm that the combinations do create
Figure 13-30 illustrates how an LCD monitor takes these the desired colors.
RGB pixel value recipes and controls color-filtered liquid Recreate our illustration using the illustration software
crystal element light transmission to display them. On the of your choice or make a unique sequence of color patch
left is a close-up view of our pure red patch, displayed by transitions by modifying R, G and B channel values. Get
turning on all red pixels to maximum brightness while a close look at your computer monitor using a magnify-
leaving the blue and green pixels off. On the far right is ing glass, a camera with macro lens or your smartphone
the pure blue patch using a similar strategy. In between camera (though it may not be able to focus at such a
them are a few of the color patches transitioning between short working distance). Even with the naked eye, you can
the red and blue: this is accomplished by mixing different discern the separate red, green and blue illuminated pixel
amounts of both. These images are so highly magnified elements. Similar to the three color channel digital counts

289
 s ecti o n 3: O utp ut

in the illustration patches, the monitor creates every color

by varying the brightness of its three additive primaries.
Lastly, Figure 13-31 reveals how a smartphone display
creates white: the red, green and blue pixels are turned on
at equal brightnesses. The presence of these primary colors
is evident in this image but appears as a neutral tone at the
proper viewing distance. We encourage you to look at a
variety of graphics and photographs rendered on different
output media to better understand how color is repro-
duced and perceived by the human visual system.

Notes
1 “Inter-Society Color Council aims and purposes.” ISCC
Aims and Purposes, Apr. 23, 2013, www.iscc-archive.org/
organization/aims.php.
2 Snowden, Robert, et al. Basic Vision: An Introduction to Visual
Perception. Oxford University Press, 2012.
3 Munsell, Albert H. “A color notation by A. H. Munsell.” Free
e-book Project Gutenberg, July 14, 2008, www.gutenberg.
org/ebooks/26054, pp. 14–15.
4 Munsell, Albert H. “A color Notation by A. H. Munsell.” Free
e-book Project Gutenberg, July 14, 2008, www.gutenberg.
org/ebooks/26054, p. 19.
5 “Web Content Accessibility Guidelines (WCAG) Overview.”
Edited by Shawn Lawton Henry, Same Origin Policy - Web
Security, Reuters Limited, June 22, 2018, www.w3.org/WAI/
standards-guidelines/wcag/.
6 Gerald, Gorn, et al. “Effects of color as an executional cue
in advertising: They’re in the shade.” Operations Research,
June 1996, pubsonline.informs.org/doi/abs/10.1287/
mnsc.43.10.1387.

290
14 Color
Management

Photograph by Rochester Institute of Technology photography alumnus Kevin Young

Color management is a term that strikes fear and anxiety in photographers’ hearts and
minds. Like taking vitamins or getting a flu shot, many of us understand its importance but
don’t look forward to the follow through. Color management’s reputation is one involving
frustration and complexity: extra equipment, extra time, extra settings and configura-
tions to keep straight. The good news is that excellent color management systems are
increasingly accessible and streamlined for photographers that don’t want to grow old
attempting to understand a properly executed, color-managed workflow. A color-man-
aged workflow is dictated by your specific needs in the context of a gallery exhibition,
print publication or soft-copy client delivery and only requires a degree of diligence and
attention to detail once established. This chapter is an overview of the concepts and

291
 s ecti o n 3: O utp ut

components to make sense of the concepts, terminology magazines and online journalistic institutions: a client or
and software settings used for maintaining accurate, pre- publisher requires that a file be ready the first time, not
dictable and desired color in photographic reproduction. after eight or nine iterations.
By the end of this chapter, the instances of being surprised
or frustrated by your photographic output should dramat-
ically decrease. Who Governs Color Management?

Before going further, it helps to understand the origin of

Why Color Management Matters color management and who governs its standards. The
International Color Consortium (ICC) is the governing body
The previous chapter laid out the many aspects of the in the field of color management. Founded in 1993, the
human visual system that result in color vision. We’ve group oversees the field of color management and main-
looked at ways of making color (mixing additive and sub- tains the ICC profile specification (which we’ll dive into later
tractive primaries), metamerism and the CIE chromaticity in this chapter). The ICC’s mission is to “promote the use
diagram all in the service of creating a robust vocabulary and the adoption of open, vendor-neutral, cross-platform
and framework for working with color. Color management color management systems”.2
takes these concepts and provides a scaffolding to bolster The five founding members of the ICC were Adobe
our ability to retain and reproduce accurate, consistent Systems Incorporated, Agfa-Gevaert N.V., Apple Computer,
color in our image-making endeavors. Now we need to Inc., Eastman Kodak Company and Microsoft Corporation.
establish new ideas like calibration, characterization, ICC There are approximately 40 members today across the
profiles and rendering intents to take advantage of that photography and printing industries. There are also 16
scaffolding. honorary members including Rochester Institute of Tech-
The cold, hard reality is this: we don’t have true nology, home to the Munsell Color Science Laboratory.3
color matching from input (capture) to output (print or These industry giants realized the need for all devices and
display).1 The technologies used are varied and often cut- software created by different companies to communicate
ting-edge, yet our tools lack the ability to reproduce every with each other. To accomplish this, they developed the
naturally occurring (or synthetically concocted) color framework for a color management system that uses ICC
and tone. Defining the boundaries and limitations of our profiles developed to a standard format. The current ver-
materials and mitigating their shortcomings is the best sion of the standard is Specification ICC.1:2010-12 and can
available strategy. Sometimes this means altering image be found at www.color.org. Without this cross-company
data to fit the constraints while other times it means collaboration, color management could easily devolve into
accepting that one output is never visually equivalent to manufacturer-specific walled-gardens of workflow solu-
another. Color management is named appropriately; we tions and limited compatibility.
don’t yet have magical, infinite, perfectly true-to-life color
solutions. Color management helps ensure predictable
color reproduction. Reviewing Color Data Concepts
Even the wealthiest photographer with unlimited ink
cartridge funds appreciates minimizing wasted time. Let’s take a quick look back at color concepts relevant
Printing and reprinting with incremental changes each to color management. Most tools used to examine the
time until an optimal print is made is an option, but it’s contents of a color profile use a chromaticity diagram
time-consuming and wholly avoidable. Time is perhaps a to display a color gamut like the one in Figure 14-1. A
more critical restraint than materials or monetary resources color gamut describes all of the colors formed by mixing
for highly scheduled productions like newspapers, different amounts of the output primaries (typically RGB or

292
 c h a p t e r 14 : C o lo r M a n a g e m e n t

Human vision’s color gamut is the largest gamut that we

consider. Unfortunately, there is no device available that
can reproduce every visible color. The chromaticity coordi-
nates of the pure red, green and blue primaries of a device
are calculated and plotted on a chromaticity diagram. A
device gamut is the triangle formed by connecting these
three points. Any color that falls in the interior of the trian-
gle is reproducible. Any color outside the triangle cannot
be reproduced and is referred to as an out-of-gamut color.
If a system, such as a printer, uses more primaries beyond
cyan, magenta and yellow inks, the color gamut may be
differently shaped.
Mismatched gamuts is one of the problems addressed
by a color management workflow. For example, when the
color gamut of the monitor is larger than the color gamut
of the printer, how are the out-of-gamut colors going
to print? This is handled using rendering intents, a set of
behaviors or instructions that dictate the reproduction
Figure 14-1 The CIE chromaticity diagram. strategy.

Color Space Types

CMYK) and it’s used to define the color reproduction capa- Chapter 13 introduced the CIELAB color space as a
bilities of a device. In 1931, the CIE defined the CIEXYZ color solution to modelling and describing color. CIELAB is
space. Working with tristimulus values is difficult because a device-independent color space encompassing all of
they are not easily visualized. Most of us find it difficult if human vision. It does not describe the color reproduction
not impossible to identify a color represented by X, Y and Z capabilities of any one output device. In color manage-
tristimulus values. ment, there are two device-independent color spaces
Realizing this, the CIE also defined a method to trans- that use numeric values to model human vision: CIEXYZ
form tristimulus values into chromaticity coordinates. These and CIELAB. CIELAB is valuable as a universal translator
values are xyY. The Y is a luminance value and is not plot- sitting between the input media or device and the output
ted. The chromaticity coordinates x and y are calculated device.
according to Equations 14.1 and 14.2 and can be plotted A device-dependent color space is based on the charac-
on a chromaticity diagram. teristics of the inks, in the case of a printer, or the filters, in
X the case of a display with color-filtered pixels. Every device
x= (Eq. 14.1)
X+Y+Z defines its own device-dependent color space. The color
space of a monitor, for example, is defined by a finite set of
Y
y= (Eq. 14.2) RGB values used to control its output. The same RGB values
X+Y+Z
displayed on two unique monitors may appear different
These chromaticity coordinates allow for plotting and have different measured tristimulus values, making the
any color on a chromaticity diagram independent of RGB values device-dependent.
luminance. The chromaticity diagram is used in color man-
agement to illustrate color gamuts and white points.

293
 s ecti o n 3: O utp ut

sRGB, Adobe RGB and ProPhoto RGB by far the largest color space. There is no printer available
There are three standard color spaces that photogra- today that can print a gamut this large, however, it offers
phers are likely to encounter when capturing images: the most amount of data headroom for post-processing
sRGB, Adobe RGB and ProPhoto RGB. These color spaces adjustments.
describe predefined color gamuts and are plotted
together on a chromaticity diagram in Figure 14-2. Adobe
RGB and sRGB share the same red and blue primaries. Pointer’s Gamut
They differ in the green primary with Adobe RGB exhibit- Pointer’s Gamut was developed by Michael R. Pointer in
ing the larger gamut. The sRGB color space (the “s” is for 1980 by measuring over 4,000 color samples.4 The gamut
“standard”) is the appropriate choice if a photograph is represents all of the real surface colors visible to the human
viewed on a projector or monitor. Although sRGB is the eye. These are colors reflected by the object’s surface using
smallest gamut of the three, you do not risk the colors subtractive color mixing. This color gamut is irregular in
being poorly reproduced by delivering images with colors shape as shown in Figure 14-2 and is not defined with
and tones beyond what the viewer’s display is capable three primaries like all other color gamuts. It covers 47.9%
of reproducing. The Adobe RGB space provides greater of human vision.
headroom for editing and includes ranges of colors than
can be reproduced in print. Many cameras offer settings to
capture in either the Adobe RGB or sRGB space. We sug- Revisiting the ColorChecker
gest setting your camera capture to Adobe RGB as you can We introduced the X-Rite ColorChecker in Chapter 9
always map to a smaller gamut later on. ProPhoto RGB is because its patches are valuable for color balancing and
evaluating tone reproduction during post-processing
edits. The ColorChecker target is also incredibly useful
for color management and output reproduction assess-
ment. The ColorChecker target patches are grouped into
four sections: Memory Colors, Hard to Reproduce Colors,
Gamut Colors and a Neutral Grayscale as shown in Figure
14-3.
Patches 1 through 4 represent dark skin, light skin, blue
sky and foliage. These are colors that we all remember
and are consequently critical when judging photo repro-
duction. Patches 5 through 11 are colors that film did not
always adequately reproduce. Patch 5 represents a prized
morning glory flower nicknamed “Heavenly Blue.” The
orange (patch 6), moderate red (patch 8) and orange-yel-
low (patch 11) represent oranges, grapefruits and lemons
respectively. The third row of patches are the additive
and subtractive primaries used to evaluate color gamuts.
The last row are Munsell neutral patches that covered the
dynamic range of film in the 1970s.
Each of these patches helps when assessing color and
tone reproduction. The X-Rite ColorChecker is produced
Figure 14-2 ProPhoto RGB, Adobe RGB and sRGB color spaces
plotted on a chromaticity diagram. Pointer’s Gamut is also plotted for with published CIELAB values. The data can be evaluated
reference. against measured data from a reproduction to access

294
 c h a p t e r 14 : C o lo r M a n a g e m e n t

device. In the context of color management, character-

izing is determining the color gamut that a printer can
reproduce, a monitor can display or an imaging device can
capture.
Calibration is a process of returning a device to a known
standard or its proper and optimal operating state. If you
have a weight scale that no longer reads zero pounds
when there’s nothing on it, the scale is adjustable back to
zero as a form of calibration. Similarly, if a printer, monitor
or scanner can be adjusted back to a known state, it can be
calibrated.
All the parts of a digital system—the camera, monitor,
scanner, printer and any hardware used in the color man-
agement process—are describable in mathematical terms
that a computer understands. This allows us to characterize
all of these devices. The term used for this characterization
Figure 14-3 The X-Rite ColorChecker Color Rendition Chart. is a color profile. When a device’s characteristics change or
drift, a new standard for calibration in the form of a color
profile may be necessary. The need to recharacterize a
accuracy objectively. For example, the ∆E or color differ- device usually becomes apparent when the device can no
ence can be calculated for all 24 patches. This provides longer be calibrated, often indicating that its performance
insight into how well our profile performed and can assist characteristics have changed.
in identifying any present color cast. The gamut patches Characterization builds a profile of a device. A specific
are used to create a basic gamut plot. The neutral patches ICC profile identifies how a device reproduces color. It
are used to create a tone reproduction plot to determine is only valid if the device exhibits consistent behavior.
if the print brightness levels are matched to the original. If something changes with the device between profile
Finally, a subjective visual assessment is performed using creation and producing the output product, the profile
all patches. We outline all of these practices later in this becomes invalid. A device that doesn’t produce predict-
chapter. able results from one use to the next is a poor choice for
a color-managed workflow; if it can’t be trusted to do the
same job over and over, it’s nearly impossible to anticipate
Defining Characterization and and accommodate its output capabilities. As long as a
Calibration device can be calibrated back to the operating condition
in which it was characterized, there is no need to create a
The terms characterization and calibration are often new profile.
used interchangeably in color management. They do
not mean the same thing. Characterization is the process
of determining how a device performs. This includes its Color Management Systems
capabilities and its limitations. We can characterize the
performance of a camera by testing it for a linear response, A color management system, or CMS, is software that con-
measuring its spectral response, measuring its signal-to- trols the accurate reproduction and appearance of colors
noise ratios and the range of illuminances it can record. across different devices. The CMS’s main responsibilities are
Characterization, then, describes the behavior of the twofold: to assign RGB or CMYK values to a specific color

295
 s ecti o n 3: O utp ut

and to control the appearance of that color as it goes from a destination profile. For example, an image is displayed
one device to another so that it appears the same. on a monitor with its own source profile. The intent is to
There are three components in a color management print this image. The CMM maps the monitor source profile
system: the color management module or color matching to the destination printer profile that defines the output
model (CMM), the profile connection space and the color device such that the printer receives and understands the
profiles. The CMM is the overarching software that takes an image data.
input or source profile into the profile connection space
and creates the output or destination profile. There are
often several CMMs to choose from, however, the Adobe The ICC Profile
Color Engine (ACE) performs well and is often the default The International Color Consortium released the first speci-
CMM. fication document in 1994. The consortium established the
The profile connection space (PCS) is a perceptually eponymous ICC profile: a cross-platform, vendor-neutral file
based, device-independent color space. The PCS uses format used to describe the color rendering behavior of a
either the CIELAB or CIEXYZ color space to function as the device.5 ICC profiles are used in color management systems
universal translator between devices. Other than appreciat- to formalize the translations and capabilities of our displays
ing that the PCS is where the color control magic happens, and printers.
we never have to work directly with this part of the CMS. An ICC profile allows an input image to be displayed
Color profiles describe a device’s behavior, be it a or printed with consistent color reproduction. The profile
camera, printer, scanner or display device. These are real, defines the relationship between a color space, such as the
physical devices. A profile can also be virtual, for example RGB primaries of a monitor, to the corresponding values in
the sRGB or Adobe RGB spaces. The profile tells us the a color space like CIELAB. Profiles can be device-depend-
range of colors that a device can capture, display or print. ent or device-independent. A device-dependent profile
All profiles conform to ICC specifications which ensures depends on the profiled equipment. A device-independ-
functionality with any encountered CMS. ent profile is one where the color designation is absolute
An image’s journey through the CMS is diagrammed in in that it defines one color no matter where it is applied.
Figure 14-4. An input device passes its source profile into The CIELAB and CIEXYZ color spaces are used to produce
the CMM. The CMM converts this profile into the PCS. The device-independent profiles.
PCS maps the input device to the output device through We categorize profiles into three classes: input, display
and output. Each plays a unique role in color management.
Input profiles describe image-creation devices like scan-
ners and cameras. These are one-way profiles; you cannot
output image data to a scanner or camera. A one-way
profile only maps the RGB pixel values of the device to its
corresponding CIELAB values used in the PCS. Input profiles
tell us how these image-creation devices record color (or
“see” color, if you prefer to think of it that way).
Display profiles describe monitors and transmissive
display devices. These are two-way profiles, as a display can
act as both an input device and an output device. When
editing an image viewed on your computer display, it’s
effectively an input device; it’s contributing to the forma-
Figure 14-4 The color management module workflow that takes input
tion or interpretation of the image data. When displaying
image data and translates it to an output device. an image as a means of photographic output, the display

296
 c h a p t e r 14 : C o lo r M a n a g e m e n t

acts as the output medium or device. A two-way profile The additional functionality that iccMAX offers include:
must communicate data to the PCS and understand data
coming from the PCS. • the ability to use actual illuminant and observer data
Finally, output profiles describe printers and printing • incorporating a new gamut boundary description
press technologies. These are one-way profiles because • measurement data using the CxF format (Color
they are used to move image data from the PCS to the Exchange format)6
output device. The output information is in a form that the • a new encoding of named color to support tints
output device requires. For printers, that form is typically • a Material Connection Space (MCS) to support identifi-
CMYK. cation and visualization of material amounts in addition
to color
• support for bidirectional reflectance distribution func-
Canned Profiles tions (BRDF).
Cameras and printers come with generic or canned profiles
provided by the manufacturer (you may need to navigate Although most of these go beyond the realm of
their websites to download them). These profiles approx- photography, the ability to incorporate actual illuminant
imate the performance of the device, however, they are data may prove useful to account for the lighting used for
created by averaging the performance of a sample set of viewing and displaying photographic work.
devices from the production line. A canned profile charac-
terizes your make and model printer in a generic way, not
your printer in particular. CMYK Profiles
In many cases, a generic profile works well in support- Output to print media requires, more often than not, that
ing a color-managed workflow. It’s certainly better than your image data be translated into a language of cyan,
not having any profile at all to guide your color output. A magenta, yellow and black (CMYK): the subtractive prima-
generic profile offers a sense of the output device’s limi- ries. A CMYK profile that matches the destination printing
tations. However, there are times when a unique, custom condition is required when sending images to a commer-
profile is necessary. For example, Epson creates its canned cial offset printer. There are unique CMYK profiles for North
printer profiles using Epson inks. The canned profile may America, Europe and Japan as each country uses different
not produce satisfactory results if we switch to inks made printing standards.
by another manufacturer. It’s also possible that your spe- Printing in North America offers three options for CMYK
cific Epson printer is an operational outlier compared to the profiles. Sheet-fed lithographic printing uses the General
larger production line of that model and its performance Requirements for Applications in Commercial Offset Lithog-
differs from that of the canned profile. raphy (GRACoL) setup for printing on coated paper. The
Coated GRACoL 2006 profile is used for Grade 1 paper.
When using web offset production presses, the Specifica-
The iccMAX Profile tions for Web Offset Publications, or SWOP, setup is used for
The iccMAX profile was first introduced in 2017. The printing on lightweight papers. There are two options: Web
current version is ICC.2:2018 (iccMAX). The ICC created Coated SWOP 2006 for Grade 3 and Web Coated SWOP
iccMAX to support a more flexible and extensible system 2006 Grade 5.
than the current profile system provides to meet the Most printers in Europe and the UK use the ISO
evolving needs of the industry. The ICC approved a private 12647/2-2004 standard developed by FOGRA, a German
tag in V5 profiles that accommodate the ability to embed printing research organization. They developed the
an iccMAX profile, making them backward compatible FOGRA39 dataset which consists of two CMYK profiles: ISO
with current software. coated v2 (ECI) for sheet-fed offset on coated papers and

297
 s ecti o n 3: O utp ut

ISO coated v2 300% (ECI) for web offset printing on whiter

web offset papers.

Converting or Assigning Profiles

There are two options for using profiles with image files:
converting an image to a profile color space or assign-
ing the profile and embedding it in the file’s metadata.
Throughout this book, we’ve stressed the importance of
maintaining original image data and editing images in a Figure 14-5 Spectrophotometers and colorimeters are used for
non-destructive manner. Profiles can be handled in the measuring output color on displays and print.
same way. When we convert an image to an output profile,
the actual RGB pixel values are permanently altered. It
bakes the changes into the pixel data and it cannot be Spectrophotometers measure color across a large set (>30)
undone. This method is the best option when you are of discrete wavelengths and provide additional informa-
delivering an image to a client or printer service to ensure tion from which colorimeter data can be calculated along
that the image appears the way you intended. with density measurements. A spectrophotometer can
The alternative is to assign a profile and embed it in the gather the same information as a colorimeter but not the
metadata. Assigning a profile to an image file is non-de- other way around. Of course, the more capable device
structive and offers a translation guide to tell the image is more expensive. If basic display measurements are all
display software what the RGB values mean. The software that is needed, a cheaper colorimeter will suffice. Figure
applies the profile and adjusts the pixel values for display 14-5 shows examples of both types of color measurement
or for printing, however the original pixel values are left devices.
unaltered. This allows for needed flexibility. For example,
perhaps you need to make a print for a gallery show that
is 40x60 inches and also make 4x6 prints for giving out Input Profiles for Cameras or Scanners
during the show. The two print types are likely printed on Building a camera profile is useful primarily for studio
different printers and paper combinations. Assigning a pro- product photographers. An input profile helps characterize
file at the time of printing allows you to use the same file the camera’s color reproduction under a highly specific
for both sizes. Some web browsers read and correctly use and rigid set of conditions; if those conditions change, the
embedded profiles but it’s not a universal behavior of web profile is invalid. If you use the same two strobes and the
browsers across desktop and mobile software today. same lens to shoot different products against the same
white seamless day in and day out, then creating an input
profile for your setup makes sense. Otherwise, the moment
Creating Profiles with Color you change your lights or lens, your profile describes a
Measurement Tools situation different from the one you’re photographing.
X-Rite’s i1Profiler software can be used for building input
We introduced some measurement tools used to quantify profiles although there are many options, both commercial
light in Chapter 1. Here, we’re interested in measuring color and open source.
and density in our output media. Colorimeters measure We photograph or scan a calibrated target like the
emitted, reflected or transmitted light through a set of Image Target 8 (or IT8 for short) shown in Figure 14-6 to
(approximately seven) color filters. They are primarily create an input profile. The IT8 target is a ubiquitous name
used for taking color readings off of computer displays. for calibration targets just as Kleenex refers to any brand

298
 c h a p t e r 14 : C o lo r M a n a g e m e n t

A scanner profile, or any input profile, is not used to

make the images look good. It does not correct color casts,
improper exposure or poor contrast. Rather, the input pro-
file allows software to display the input image file on your
monitor in a way that matches the original.

Display Device Profiles

We promised at the start of this chapter that proper
color management is increasingly user-friendly for pho-
tographers. Profiling a display requires a colorimeter or
Figure 14-6 An IT8 target used for creating input profiles.
spectrophotometer to measure it. Unlike the calibrated
target used for a camera or printer, the software sends
of tissue. The IT8 is the reference color set supplied by the color patches directly to the display that’s measured by
manufacturer. Each color sample is accompanied by meas- the colorimeter or spectrophotometer in turn. The profil-
ured CIELAB and CIEXYZ values. Profiling software takes the ing software requires the user to set certain parameters
input RGB values from the camera or scanner and maps such as device type, gamma and white point. The bright-
them to the target’s known values to create the profile. The ness value is also adjusted. The profiling software then
generated profile is specific to the camera or scanner used. runs through neutral, red, green and blue color patches

Figure 14-7 A display color gamut as measured by profiling software and stored in an
ICC profile file.

299
 s ecti o n 3: O utp ut

that vary in brightness from 0 to the maximum digital There are often multiple test targets to select from. The
count of the display. The CIEXYZ values are measured and adage “go big or go home” applies here. Selecting the
CIELAB values calculated. The profile is created such that largest test target with the most color patches results in
the PCS understands what color every RGB pixel combi- a robust printer profile. A print should dry for 24 hours to
nation represents. The operator is prompted to name the ensure that there is no further shift in the color of the inks.
resulting ICC profile at the end of the process. We sug- Once dry, we measure the printed patches with a spec-
gest including the monitor name, luminance value and trophotometer or colorimeter and feed the values to the
calendar date for easy identification. Figure 14-7 illustrates profiling software. The resulting ICC profile is specific to the
the measured color gamut of the projector in the audito- paper and printer with which it was created and should be
rium we use to lecture on photography topics. We’re able named accordingly for easy identification.
to view the measured gamut in the operating system’s
profiler viewer utility.
Display characteristics change or drift as they age and Gamut Handling and the Profile
require profiling on a regular basis. Most profiling soft- Connection Space
ware includes a 30-day reminder to make a new profile.
The profile is also affected by the brightness and con- As introduced earlier, the profile connection space or PCS is
trast settings on the display. These settings are typically the place where the input and output profiles are bridged
controlled through software for built-in displays like the using a universal color space. A device profile defines the
ones found on laptops and as hardware controls for relationship between a device and the PCS. For example,
external monitors. Once a profile is created, these controls a camera profile defines the relationship between the RGB
cannot be changed or the profile becomes invalid. This digital counts from the camera to the PCS; a printer profile
detail is notable because we often like to vary our laptop defines the relationship between the PCS and the CMYK
screen brightness to conserve energy or boost visibility in values required by the printer.
brightly lit environments. This action invalidates the pro- The ICC based the PCS on the ISO 13655 standard
file, meaning that we have to change this habit or avoid Graphic technology – Spectral measurement and color-
color-management-critical image editing work on our imetric computations for graphic arts images. The PCS
laptop computers. uses the 1931 CIE standard observer, D50 illuminant and
either 0°/45° or 45°/0° color measurement geometries. It
also requires a black backing behind prints when making
Printer Profiles reflectance measurements. The ISO 3664 standard is used
Printers often come with a canned profile and paper man- to properly view prints created with ICC profiles. The view-
ufacturers create profiles for their papers. Most times they ing setup requires the D50 illumination at 500 lux on a 20%
work well. When they do not, or if the profile you need is reflecting background.
not available, you may need to create a custom printer pro- Addressing the difference between input gamut and
file. Starting the process of profiling a printer begins with output gamut, especially when the output gamut is
confirming that the printer is properly working. A printer smaller, is the responsibility of rendering intent decisions
with clogged or dirty nozzles causes banding in the print. discussed in the next section. First, though, we need clear
Printing software utilities can run test prints to exercise the strategies for handling out-of-gamut color information.
nozzles and confirm they are operating as expected. Print- One available strategy is gamut clipping. Every out-of-
ing profile software steps through the process of printing gamut color is mapped to the closest color that is in-gamut
a test print needed to create a profile. This may include when using this strategy. Several different input colors may
turning off color management, selecting the printer, the be mapped to the same output color. This can cause a loss
output media and the print quality. of tonal variation in highly color-saturated areas.

300
 c h a p t e r 14 : C o lo r M a n a g e m e n t

The second strategy is gamut compression. The larger The perceptual rendering intent employs a gamut
input gamut is compressed to fit inside of the smaller compression strategy that works to maintain visual detail
output gamut. Since this impacts all of the image’s colors, and luminance over hue and saturation. We’ve mentioned
including those which would have otherwise natively fit many times that our vision is more sensitive to lumi-
in the available gamut, some colors that were perfectly nance than it is to color, a characteristic that’s exploited
reproduced prior to the compression are changed or at many stages of image-making. In color management,
shifted after compression. Gamut compression can also we consider the perceptual rendering intent to be one of
lead to loss of some tonal variation in highly saturated the most useful strategies because it plays to our visual
areas. system sensitivities. The perceptual rendering intent shifts
in-gamut data to ensure that all out-of-gamut data is
shown. This shift is a compromise to maintain the overall
Rendering Intents relationship of all the colors in the image. It may mean that
Using output profiles and an optimal strategy for handling in-gamut colors are desaturated.
out-of-gamut information for that output involves the The saturation rendering intent employs a gamut clipping
use of rendering intents. These are the specific rules that strategy, preserving saturation over luminance. It’s primarily
dictate what the color management module does with, used for graphics such as company logos or illustrations.
say, the vibrant shades of green in a rainforest scene that The rationale here is that it’s preferred to maximize saturated
your printer cannot faithfully reproduce. Without an exact, colors to grab attention and that graphical elements are not
reproducible match for those colors, we turn to rendering necessarily based on real-world color from which to compare.
intents for a treatment that yields a good looking print The saturation rendering intent moves out-of-gamut colors
with vibrant greens—even if they are adjusted to a slightly to the edge of the destination color gamut and may move
different shade. The ICC specifies four rendering intents: in-gamut color out to the edge as well. This rendering intent
perceptual, saturation, ICC-absolute colorimetric and is not concerned with the accuracy of color reproduction.
media-relative colorimetric. Figure 14-8 illustrates these There are two colorimetric rendering intents: ICC-ab-
available approaches with two example color samples solute colorimetric and media-relative colorimetric. Both
measured from an image. Consider how the out-of-gamut rendering intents attempt to leave original colors unaltered
and in-gamut samples are treated by each approach as we when possible. The difference between them is how the
describe them in greater detail. color white is treated.

Figure 14-8 Example treatments for sampled image areas using different rendering intents.

301
 s ecti o n 3: O utp ut

The ICC-absolute colorimetric rendering intent recognizes

that white in the image is a function of the substrate’s
white appearance. In other words, white is never pure and
instead it takes on the base color of the paper on which
the image is printed. For example, an image printed in
a magazine must be proofed for reproduction on the
magazine paper. We normally consider print output using
photo-quality printers and white paper substrates. Mag-
azines, though, tend to use cheaper paper with a darker,
sometimes off-white white point. The absolute rendering
intent maps the white in the image to the white of the
print material. The remaining colors are left unaltered when
possible and adjust those that require it.
The last rendering intent is media-relative colorimetric.
In this rendering intent, the white in the image or source is
mapped to the brightest white possible in the destination Figure 14-9 Example of a gamut comparison between an original and
color gamut. This approach is referred to as white point a reproduction. The reproduction is less capable of rendering some
colors in the magenta region which may or may not be relevant to our
compensation. All of the other colors in the image file are photographic content.
shifted accordingly. The result may be slightly brighter or
darker than the original. Out-of-gamut colors are set to
their nearest, in-gamut likeness. This intent often works well chroma ratio. The hue of the original and the reproduction
with photographic content. can also be calculated and compared. A tone reproduction
curve can be generated to determine if the brightness is
faithfully reproduced.
Evaluating Color Reproduction We quickly evaluate the relative color gamuts of an orig-
inal versus its reproduction using a two-dimensional color
At the end of the day, there are two ways to evaluate gamut plot like the one shown in Figure 14-9. This plot
the color reproduction of your photographs. The first uses the a* and b* values of the gamut colors (red, green,
is a simple visual inspection: does the output medium blue, cyan, magenta and yellow). Gamuts are defined as
show the content as you intended it to look? The only three-dimensional spaces, however, this simplified rep-
tools required here are your artistic vision, your eyes and resentation provides insight into areas where the color may
your subjective opinion. If objective color reproduction is be inaccurate.
deemed valuable or necessary, a quantitative approach is The most common metric used to evaluate the accuracy
required. of a color match is Delta E (or ∆E). Delta E is a measure of
the difference between two samples. Refer back to Chapter
11, Calculating Color Difference, for a review of this metric.
Measuring and Comparing Color There are two ∆E options: ∆Eab and ∆E00. ∆Eab is easily calcu-
There are several established methods for examining color lated by hand or with spreadsheet software. However, the
differences. An overall color difference can be exam- CIE recommends a ∆E00 (Delta E 2000) calculation which is
ined. The first step might be comparing the color gamut much more computationally rigorous and places heavier
differences between the source and the reproduction. If weight on luminance over chroma and hue values. Either
a noticeable difference is identified, further analysis can metric provides valuable feedback on the accuracy of color
determine if a sampled color’s saturation changed using a reproduction.

302
 c h a p t e r 14 : C o lo r M a n a g e m e n t

Figure 14-10 Calculating chroma and chroma ratio for a sampled color and its reproduction. We can conclude that the reproduction is less
chromatic or saturated than the source.

Chroma (C* ) is a measure of the degree of saturation hue angle (hab ) calculation identifies the sample’s color or
(see Figure 14-10). A chroma ratio provides insight into hue. Determining if there is a shift in hue means calculating
a saturation change in the reproduction. The formulas and comparing the hue angles for the original and repro-
for calculating chroma and chroma ratio are provided in duction (see Equation 14.5). The change in hue angle is
Equations 14.3 and 14.4. A chroma ratio equal to 1 indi- indicated in Figure 14-10.
cates a perfect match between the reproduction and the
hab = tan–1 ( a*(
b*
original. A value greater than 1 indicates that the reproduc- (Eq. 14.5)
tion is more chromatic than the original and a value less
than 1 indicates that the reproduction is less chromatic. A tone reproduction plot is made using neutral patches
Figure 14-10 shows these calculations applied to a sample included in a test print. L* values provide luminance or
color patch and its print reproduction. While a slight visual brightness information, therefore a plot of the original neu-
difference is apparent between the two red samples, calcu- tral patch L* values compared to the reproduction provides
lating chroma and chroma ratio put objective numbers to an easily interpreted visual to evaluate. The plot shown in
support that observation. Figure 14-11 (overleaf) shows that the original is lighter
than the reproduction, signaling a change to the overall
c*ab = √ a*2 + b*2 (Eq. 14.3) brightness in our reproduction that we can consider when
making final image adjustments for the desired print.
c*ab Reproduction
Chroma Ratio = (Eq. 14.4)
c*ab Original
Color Correction
Recall that a color sample representation in CIELAB Color corrections are made when we believe that our
consists of three elements: L* represents the brightness of workflow is properly color managed and predictable.
the color while a* and b* describe the color information. A When all components produce a predictable output and

303
 s ecti o n 3: O utp ut

at your local hardware store. The alternative is a daylight

balanced light with a brightness output between 30 and 65
lux. Figure 14-12 shows a great workspace environment for
color-critical photo editing and output.
Lux levels are not described on light bulb packaging.
Instead, they tend to advertise light output in lumens. For
a 10x10’ room, 30–65 lux is between approximately 278
and 603 lumens. As a point of comparison: a dinner candle
gives off approximately 12 lumens and a single 60-watt soft
white incandescent light bulb produces approximately 840
lumens. Choosing a desktop editing station environment
that lacks windows ensures that the room illumination
stays constant and is not subject to mixed source illumi-
nation. While we encourage photographers to get some
Figure 14-11 A tone reproduction plot. The reference line makes it sunshine, we want to avoid late-night editing sessions
easy to observe if the measured patches from the reproduction are
lighter or darker than the original.
getting revised the following day due to ambient light
changes from nearby windows. As a final detail, care must
be taken to avoid direct reflections or glare due to lights
positioned at complementary angles relative to those used
the result is still, say, too yellow, that’s the time to go back when viewing the display.
and introduce more blue to the file. If the system is not The display used for image editing should be set to an
properly color managed, for example, color corrections output luminance between 80 and 120 cd/m2. This is set by
applied to your image can have an unforeseen outcome the software used to create the display profile and should
on your print. Color correction techniques are covered in not be modified between profiling sessions. Running
Chapter 9. through the profiling process for the first time, photogra-
phers are often surprised at how dim the room lighting
and their screens appear when set to the appropriate light
Establishing an Appropriate levels. Minutes later, their vision adapts to the lower light
Workspace Setup levels and the confusion and concern fades.
Finally, the software interface and the computer desktop
Control and consistency are key when optimizing a background should be set to a neutral midtone gray to
workspace for a color-managed workflow. The ISO 3664 avoid simultaneous contrast effects. Image editing software
standard specifies standards for viewing color-managed
images, some of which we can affordably implement while
others we cannot.
Photography students tend to edit images wherever
they can find a wall outlet to power their laptops. We do
not suggest that anyone should do this with any expec-
tations for accurate color reproduction in an editing
workflow. The optimal workspace environment starts with
a windowless room with neutral walls (ideally a midtone
gray). The room lighting should be D50 according to the Figure 14-12 An example workspace ideal for editing and print or
ISO standard. However, D50 illuminants are not found display output.

304
 c h a p t e r 14 : C o lo r M a n a g e m e n t

user interface settings often allow for alternating between anticipation of the destination color space before commit-
different neutral tones (white, black and shades of gray in ting to a conversion in the CMM. We don’t ever want to
between). convert to a profile until all image editing is complete, so
soft proofing is a preemptive step toward getting a desira-
ble output result.
Soft Proofing with Profiles We open our image file in Adobe Photoshop Lightroom,
for example, and switch to its soft proofing mode through
Proofing is the practice of testing or evaluating how the the Develop module. This introduces new settings shown
final product is going to look upon final output. When for identifying the destination profile (from the ICC profiles
performing this on the computer display, it’s considered available on your computer) and the preferred rendering
soft proofing. Making a physical print on a substrate is hard intent strategy. In this case, the editing software offers
proofing. Both are valuable exercises and are worth using the option to create a derivative version or copy of the
hand in hand when considering photographic output image from which to make edits specific to your intended
adjustments as you work toward the best possible repro- output. This is helpful when considering work that may see
duction. Soft proofing minimizes printing costs when multiple avenues of reproduction; multiple copies can be
preparing media for physical media output and saves valu- kept separate with custom adjustments made to each as
able time when prepping for display output as well. necessary for optimized output.
Soft proofing offers an expectation of your image’s The software can indicate if there are any areas contain-
appearance with a specified output, including print, ing out-of-gamut colors, visualized with a colored pixel
even though it’s conducted exclusively with a computer overlay toggled through the histogram panel and shown
display. A major limitation to soft proofing is the difference in Figure 14-13 (overleaf). There are many editing actions
between the additive, emissive color mixing of the monitor that can help correct this gamut issue. One option is to
compared to the subtractive, reflective color mixing of the globally desaturate all colors. This is likely the least desir-
physical substrate and colorants. Even set to a low bright- able approach, as it’s guaranteed to negatively impact
ness, a display is brighter with a higher white point than color reproduction relative to the original. We might
any photo paper. To compensate, we must use the display instead desaturate or shift the hues of out-of-gamut colors
output profile plus the print output profile to approximate selectively using the targeted adjustment tool or similar
an accurate appearance. eyedropper style controls. Adobe Photoshop Lightroom
offers a panel for adjusting hue, saturation and lightness: a
combination of these three characteristics results in some
A Soft Proofing Workflow of the photograph’s pixels to be outside of the destination
Once the creative edits are finalized, we work to prepare profile’s color gamut. Working with one or a combination
the image for reproduction. The steps in a soft proofing of all three can reign in the problematic pixels. Once the
workflow vary depending on the software used, though out-of-gamut warnings are minimized, we can assume that
the ingredients are always the same: a properly charac- there won’t be any surprising delta between the source
terized and profiled monitor, an environment optimized and destination renderings of our image. Indeed, with a
for image editing and an ICC profile for the printer/paper thoughtfully selected rendering intent and minor tweaks to
combination (or alternate display) intended for output. hue, saturation or lightness, we can often create a prepared
Adobe Photoshop and Adobe Photoshop Lightroom file that yields excellent output results with minimal visual
both offer soft proofing tools. By using destination pro- changes or differences.
files for output devices or color spaces, we can see a
simulated, visual approximation of how a photograph will
render upon output. This allows for making adjustments in

305
 s ecti o n 3: O utp ut

Figure 14-13 Simulating a reproduction using an Epson printer and luster photo paper profile. The out-of-gamut
warning colors pixels that cannot be directly reproduced using the destination profile (left). After some edits to
saturation and hue, the out-of-gamut pixels are minimized (right).

Notes
1 Murphy, Chris. “Top ten color management pros.” Color 5 The latest specification as of this publication is Specification
Remedies: Library, Nov. 2001, colorremedies.com/library. ICC.1:2010.
html. 6 “CxF – Color Exchange Format.” X-Rite, www.xrite.com/
2 “International Color Consortium,” color.org/. page/cxf-color-exchange-format.
3 “Program of color science.” Reasons Students Plagiarize or
Cheat | Academic Integrity | RIT, www.rit.edu/science/pocs/.
4 Pointer, M.R. “The Gamut of real surface colours.” Color
Research & Application, vol. 5, no. 3, 1980, pp. 145–155.,
doi:10.1002/col.5080050308.

306
15 Display
Technology

Photograph by Rochester Institute of Technology photography alumnae Jenée Langlois

Displays are incredibly important to photographers. We take for granted the number and
variety of displays we encounter in day-to-day life, from watching television to using our
phones and desktop computers. Photography and video content are experienced on all of
these yet the most critical display is the one we use for reviewing and editing our work. A
good display for a photographic workflow takes into consideration the specific display tech-
nology, resolution, dynamic range and color gamut to best serve our images. It must also be
coupled with proper calibration and color management. Understanding displays requires a
familiarity with the underlying technologies, industry standards, principles of color mixing
and a sprinkling of color measurement. Here, we review the technologies and characteristics
of displays as they relate to photographic reproduction. At the end of the chapter, we look
at emerging trends and how mobile phone display capabilities have grown in recent years.

307
 s ecti o n 3: O utp ut

A cheap display can undermine a whole host of expen- that phosphorescence is a type of light energy emission
sive equipment before and after its place in the processing that persists even after excitation. Using materials that have
pipeline. We consider computer hardware upgrades every long afterglow emission times can cause image ghosting;
few years and yet the desktop monitor’s lifespan is often CRT displays with phosphor emission times below 5–10 ms
longer. Additionally, photographers rely on quality displays are desired for motion media.1
beyond their desktop computer setup. The camera preview CRT technology started as a monochromatic method of
screen, electronic viewfinders, laptops, tablets and tele- rendering images and evolved to use three electron guns
vision sets all play a role in a photographic workflow and and three phosphor-coated color screens (red, green and
output. Tethered capture and remote viewing are common blue emitting) to create color images in the mid-1950s.
methods of reviewing images during and after shooting A thin metal grid called a shadow mask is sandwiched
sessions, either due to inconvenient camera positions, between the screen and electron gun to help create pre-
having a viewing audience or to preview captured frames cise excitation elements for resolving image points.
with instant post-processing adjustments. These secondary The cathode ray tube was a ubiquitous display technol-
or tertiary viewing displays may not be as keenly calibrated ogy used for televisions, computer monitors and a variety
or optimized and yet as visual media practitioners we lean of other uses. The technology offers very good viewing
on them heavily for feedback. Their use means that one angles and fast response times. The CRT’s peak brightness
photographer may encounter or take advantage of a vari- is 100 cd/m2. The Rec.709 specification for color, discussed
ety of display types, configurations and capabilities in their later in this chapter, is based on the display capabilities of
endeavors. CRT technology. CRTs are increasingly rare for personal or
professional use, though some are found in specialized
industries such as military, medicine and industrial engi-
Relevant Display Technologies for neering (such as oscilloscopes). Their large size and high
Photographers power consumption are notable downsides compared to
newer options.
Let’s go in chronological order of mass adoption as primary
display technologies: cathode ray tube, liquid crystal
display, organic light-emitting diode and quantum dot. Liquid Crystal Display
All displays require an input signal sent from a computer’s The CRT’s popularity was eclipsed by liquid crystal dis-
graphics processing unit (GPU) that is first held in a frame play (LCD) technology that offers better picture quality
buffer in random access memory (RAM). The frame data in a smaller form factor. An LCD consists of an array of
is sent to the display which houses its own signal inter- sub-pixels positioned in front of a light source. A layer of
pretation hardware to translate and command its display molecules aligned between two transparent electrodes
elements to drive brightness and color. and two polarizing filters comprise the display screen.
The polarizing filters are aligned perpendicular to each
other, blocking all light from passing through. The elec-
Cathode Ray Tube trodes apply an electrical field to the liquid crystals which
The cathode ray tube (CRT) display uses an evacuated glass align to control the amount of light passing through the
tube with a phosphorescent screen and an electron gun. system like small, light-blocking shutters. This sandwiched
Light is emitted when the electron gun fires electrons (the layering of components is shown in Figure 15-1. A color
cathode ray) at the screen and excites its phosphors. The filter layer sits atop the liquid crystal elements such that
electron beam is modulated to display images or graphics individual sub-pixels are filtered to be red, green or blue
and scans from the top of the screen to the bottom in as seen under the microscope in Figure 15-2. This filtering
rapid sequences, one line at a time. Recall from Chapter 1 strategy takes advantage of partitive mixing to appear as a

308
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

Figure 15-1 The fundamental layers of an LCD (top) and an OLED display (bottom).
Illustration adapted from LG.com

wide variety of colors (including white) when viewed at a

standard working distance as the pixels vary the amount of
light transmitted.
Importantly, LCDs do not produce their own light. A few
applications of LCD displays work just fine without light
emission of any kind. The readout of a basic calculator, for
example, instead relies on reflected light from the environ-
ment. Most applications, though, whether it’s a microwave
oven timer or a living room television, require backlight-
ing to create an emissive light display. The backlight of
a display used for photographic content provides even,
bright white illumination that is diffused, polarized, filtered Figure 15-2 The surface of an LCD screen of a portable music player
and transmitted through the layers stacked on top of it. photographed through a reflected light microscope. Photograph by
Rochester Institute of Technology photography alumnus Griffin Kettler
Early LCD monitors used a cold cathode fluorescent lamp
(CCFL) as a backlight source. Though cheap, CCFL is heavy
and uses more power than newer light-emitting diode (LED)
backlighting solutions. LED backlights are most commonly
comprised of blue LEDs coated with yellow phosphors
to produce a neutral, white light. These are power effi-
cient but don’t offer the full spectral distribution range
that RGB LED backlights can; a limited spectral emission

309
 s ecti o n 3: O utp ut

Figure 15-3 An LCD monitor displaying a pure black image and

viewed in a dark room reveals backlight non-uniformity and an
elevated black point.

range means poorer color reproduction. The use of LED

backlights spawned confusing marketing terminology
where displays are sometimes branded as “LED displays”
even though the driving technology is LCD. Simpler LED
backlights use arrays of lights along the outside borders of
the active display area that get diffused with light-guiding
sheets. This can lead to backlighting non-uniformity that Figure 15-4 An illustration of an LED backlight panel with
becomes most apparent when displaying dark, uniform independently controlled zones. The brightness pattern acts like a
low-resolution luminance interpretation of the on-screen content to
tones on screen (like the end credits to a film). An exam- enhance dynamic range.
ple of this backlight non-uniformity is seen in Figure 15-3
where an entirely black image is displayed in a dark room.
More complex LED backlight solutions use a full array of
controllable diodes behind the LCD. LCD screens additionally employ different hardware
A traditional backlight is only capable of global dim- designs for handling the liquid crystals: in-plane switch-
ming. LED backlight panels with segmented, local zones ing (IPS), twisted nematic (TN) and vertical alignment (VA).
whose luminance are independently controlled improve Although IPS is the most expensive to manufacture, it
the dynamic range of LCDs. This feature is called localized offers the best color reproduction and viewing angles
dimming. The more finely divided (and thus more numer- among the three. IPS technology’s recent evolutions
ous) the zones, the greater the potential improvement over include Super IPS (S-IPS), Advanced Super IPS (AS-IPS) and
global dimming. The ability to drive backlight intensity in IPS Pro. IPS technology allows each pixel to display a
specific parts of the frame means that a landscape shot value between 0 and 255 which results in a better color
with a bright sky and dark foreground can be backlit sep- rendition over the TN and VA alternatives. LCDs with IPS
arately to great effect. Figure 15-4 illustrates this concept are often not as bright, however, and the liquid crystals
using a simple grid array representing backlight lumi- cannot change state as quickly as in an LCD with TN.
nances. In reality, the localized backlight zone boundaries Viewing angle is a critical characteristic when considering
are slightly blurred or softened by a layer of diffusion before the increasing frequency of viewing displayed content
the liquid crystal layer. on portable devices and those in our work or social

310
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

environments; displays viewed head-on from a desktop or without any backing at all to produce translucent displays
living room couch are far from their only use case. (of questionable utility aside from the major science fiction
High-quality LCD panels are made by manufactures wow-factor). The manufacturing process can additionally
including Eizo, BenQ, LG, Samsung and NEC. Professional integrate a touch-sensing layer that makes a separately
photographers relying on an editing workstation for a color manufactured and assembled layer of touch screen
managed workflow must consider the appropriate display technology unnecessary. LG Electronics, Inc. and Samsung
hardware offering. Some monitors, specifically marketed to Electronics Co., Ltd are the primary manufacturers of OLED
photo and video editors, feature integrated, self-calibrating panels today.
hardware and software. OLED technology’s manufacturing complexity kept
prices high and yields low until recent years. This fact also
explains why they proliferated quickly for small panel
Organic Light-Emitting Diode applications like phones before becoming viable options
Organic light-emitting diode (OLED) displays offer addi- for television panels. However, larger panels are becoming
tional improvements over traditional LCD technology. The increasingly price-competitive with LCD options. OLEDs
emissive electroluminescent layer in an OLED is made of can be susceptible to image burn-in and early iterations
a solid-state semiconductor film of organic compounds. saw short lifespans for the blue-emitting phosphorescent
The layer is typically a polymer substance that allows the material.3,4 Many mobile devices incorporate software
organic compound to be deposited in rows and columns solutions that keep any single pixel from displaying the
on a flat carrier that become individually addressable. The exact same thing for too long in an effort to avoid burn-in
compound, a fluorescent or phosphorescent material, degradation.
emits varying amounts of light when current from a cath-
ode is applied. OLED displays do not require a backlight
(the pixels are self-emitting) and they use less energy than Quantum Dot
an LCD display.2 The organic compound layer is covered LCDs are one of the most commonly encountered display
by a single polarizer and is typically sandwiched between a technologies today. Without technological advances,
transparent cathode and an anode layer. Figure 15-1 shows they may be feature-eclipsed by OLED. However, quan-
the relatively fewer components present in an OLED panel tum dot (QD) technology is keeping LCDs in the running
design relative to LCD. as contenders arise. Quantum dots are semiconductor
A consequence of OLED’s self-emitting pixels is truly nanoparticles that emit different colors when excited by
deep, dark black tones. Displaying black is accomplished incident blue light. The color emitted is dictated by the
by not sending any electric current through the organic size of the crystals which range from 2–6 nm in diameter.
material and thus the only brightness at that location The smaller the dots, the shorter the emitted wavelength
comes from reflected light off of the screen surface. LCD of visible light: the largest dots emit red, the smallest, blue
cannot accomplish this because the backlight panel stays within narrow bands. The blue backlight itself is often used
on to illuminate other pixels and the cross polarization is as the third primary alongside red and green emitting dots.
not 100% effective at blocking light. Building a display using a variety of quantum dot sizes
OLED panels started showing up in camera hardware for means that exciting all of them with blue backlight can
the back-of-camera preview screens. They are increasingly produce any desired color. The nanoparticles absorb and
found on smartphones and televisions. The lack of a back- reemit light energy with a high degree of quantum effi-
light layer means that the panels can be thinner than their ciency (nearing 100%). Green QDs fade or fail to emit earlier
LCD counterparts. OLED technology can also be mounted than red ones, meaning that a full color display may see
on flexible substrates (like plastic) to produce curved or a shortened lifespan, not unlike early OLED displays with
semi-bendable form factors. They can be manufactured fading blue emission.5

311
 s ecti o n 3: O utp ut

Quantum dot technology produces highly saturated distance. Recall that pixels are a useful building block for
colors which allows QD displays to render wide color images yet LCD displays are comprised of red, green and
gamuts (potentially wider, even, than OLED). The lack of blue sub-pixels; pixel density considers a group of RGB
filters means that they are more efficient in their use of sub-pixels as a single pixel. The standard pixel density
the backlight to create an image. The primary method of used to be 72 ppi. This was generally the case for desk-
creating quantum dots involves the use of cadmium, a top displays for a number of years. Today, most desktop
potentially toxic metal. The undesirable risk of its use in and laptop displays are 100+ ppi in pixel density. Smaller
consumer electronics spurred the development of cadmi- screens like those found on the smartphones and the
um-free quantum dot technologies. These variants exhibit backs of cameras often offer even higher pixel densities of
reduced quantum efficiency and lessened color gamuts.6 200–300 ppi or more. A combination of function and the
QD technology is additionally being leveraged for realities of manufacturing challenges, small screens pack
lighting solutions with high Color Rendering Index (CRI) high spatial resolution into small spaces and hold up to
characteristics. Instead of driving a full color display, quan- close viewing distances. While a desktop monitor is likely
tum dots’ light emission can make for spectrum-tunable, to be viewed at a single distance, a phone or tablet invites
flat-panel light fixtures. variable viewing distances and needs to look good across
all of them. High pixel density displays are also critical for
wearable or head mounted display applications with very
Display Connection Standards short viewing distances.
External monitors for desktop computing use High-Defi- The higher the resolution, the more information can
nition Multimedia Interface (HDMI), Digital Visual Interface be shown at a time. Using a display that’s lower resolution
(DVI) or DisplayPort connection protocols. Specialized than the content, such as editing 4K video on a sub-4K
products might offer some functionality via USB, WiFi or monitor, makes it mathematically impossible to view the
Bluetooth, though these are not common for continuous content pixels 1:1 on screen while showing the entire con-
video display due to bandwidth and power limitations. tent frame. Photographers have long since understood that
HDMI and DisplayPort come in different flavors with they have more pixels than they’re viewing when zoomed
newer versions offering higher throughput, power and out to fit images to the screen.
daisy-chaining.

Aspect Ratio and Screen Size

Display Resolution We listed the standard resolutions and aspect ratios for
video formats in Chapter 6. Displays vary greatly in aspect
The display resolution describes how many pixels comprise ratio depending on their application, though we generally
the screen in width and height. The more pixels in an array, see a trend toward wider formats in mobile and desktop
the more information can be displayed. Similar to image applications. The closer a display aspect ratio comes to a
sensors, displays have one set, native resolution when film and video standard like 16:9, the more of the screen is
they’re manufactured. Options to change resolution in active during playback of such content. Desktop displays in
operating system settings are actually just modifying the particular are used for much more than watching movies,
scale of rendered graphics and user interfaces. Describing however: extra screen real estate in vertical or horizontal
display resolution as a function of area uses the measure- dimensions can be valuable for additional user interface
ment of pixels per inch (ppi). elements, panels or side-by-side application windows. It’s
The higher the pixel density, or number of pixels per also common for professionals to arrange multiple mon-
given area, the more likely it is that your eyes cannot see itors side by side (supported in any modern operating
the separate, discrete pixel boundaries at a normal working system). This is an effective way to display an image as

312
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

Table 15-1 Common desktop display resolutions and aspect ratios.

Pixel Resolution (Width x Height) Aspect Ratio Category

1024x768 4:3 Full screen

1600x1200 4:3 Full screen

1920x1080 16:9 Widescreen

2560x1440 16:9 Widescreen

3840x2160 16:9 Widescreen

5120x2880 16:9 Widescreen

2560x1080 21:9 Ultra Widescreen

3440x1440 21:9 Ultra Widescreen

3840x1080 32:9 Ultra Widescreen

large as one display allows while relegating software win- Refresh Rates and Response Time
dows and controls to another display.
CRTs were, by and large, 4:3 in aspect ratio. This is A display’s refresh rate defines the number of times per
considered full screen. A 16:9 aspect ratio, perhaps the most second that the display refreshes or draws successive
common today with LCDs, is called widescreen. Displays images. Refresh rate is a measure of frequency in hertz
featuring even wider dimensions are categorized as ultra (Hz). Motion in video is shown as differences between
widescreen and feature aspect ratios such as 18:9 or 21:9. frames, therefore the refresh rate impacts the perception of
Table 15-1 lists common screen resolutions and their smooth motion. The terms refresh rate and frame rate are
aspect ratios. sometimes confused: refresh rate is a characteristic of the
Large displays are often described as a measurement monitor whereas frame rate is defined by the amount of
of their diagonal dimension, from one top corner to the information being sent to the monitor.
opposing bottom corner. Aspect ratio and pixel den- We perceive a slideshow-like display of sequential
sity are worth considering in the context of a display’s frames when displayed at frequencies less than 10 Hz.
diagonal dimension since size only tells part of the story. Starting at about 10 Hz, we begin to perceptually inte-
Larger displays do not necessarily come with increased grate the frames. A refresh rate of 24 Hz is the motion
pixel densities. A smartphone and living room television picture standard originally established in the 1920s. A
can feature similar display resolutions despite being common refresh rate for computer monitors, tablets and
dramatically different physical dimensions. The diagonal other mobile device screens is 60 Hz. Increasing things
dimension is measured on the active display area (not up to 90 Hz becomes important for fast motion with
including the bezel). The thinner a bezel, however, the minimal flicker, a necessary behavior for head mounted
more seamless the appearance becomes when plac- displays (more on this later). Refresh rates of 120 Hz are
ing multiple panels next to each other. Typical desktop offered on higher-end gaming monitors. Some televi-
display sizes include 24, 27, 32 and 34 inch diagonals. sions tout high refresh rates but don’t necessarily have
Televisions used as large review monitors in studios or the temporally sampled content to match. Perceiving
editing suites can be 32, 43, 55, 65 and 75 inches in size the apparent refresh rate of still frames results in flicker.
(as well as sizes in between). It is more easily noticed in bright light and dim light
environments and in one’s peripheral vision. It becomes
less noticeable with fatigue. Flicker is a function of the

313
 s ecti o n 3: O utp ut

display’s refresh behavior, not the capture method or Display Dynamic Range, Brightness
frame rate. and Bit Depth
High refresh rates are beneficial when rendering
graphics, graphic user interfaces and video game content We introduced the measurement of luminance in terms
because graphics processing units are able to generate of candelas per unit area in Chapter 1. It is increasingly
content at high rates. Televisions that attempt to reduce common in the context of display brightness to use the
motion blur with artificial frame interpolation to achieve term nit to describe this same measurement. One nit is
a 120 Hz signal can look unnatural or what many describe equal to one candela per square meter. Nits are convenient
as the “soap opera effect.” It is common for televisions (and less of a mouthful when discussing out loud) because
to default to such artificial motion smoothing despite its displays are two-dimensional areas from which light is
unpleasant appearance for the sake of motion smoothing. emitted. As discussed in Chapter 1, light measurement gets
High refresh rates mean more content per second. Espe- more complicated when dealing with bulbs that emit in all
cially at high resolutions (4K and above), high refresh rates directions. This chapter uses nits and cd/m2 interchange-
demand considerable bandwidth. Such bandwidth is only ably as they refer to the same light measurement. You
supported on the latest connection protocols of HDMI and may encounter both when shopping for, calibrating and
DisplayPort. evaluating professional displays.
The most accurate term to describe the fastest refresh The full range of scene brightnesses faces multiple
speed of LCD and OLED displays is pixel response time. stages of technological bottlenecks before a recording
Response times are the duration (in fractions of seconds) of that scene is viewed on a display. A high dynamic
that it takes addressable display pixels to switch from one range (HDR) scene exposure necessitates a camera sensor
value to another. LCD technology offers pixel response capable of operating across a potentially huge spread of
times between 1–5 milliseconds on average, limited by luminances. Sensors continue to improve, yet the next
how quickly the liquid crystal can change orientation and bottleneck shows up in how the recorded data is stored.
therefore light transmission. OLED can potentially refresh Image file formats have maximum bit depths that define
ten times as fast (0.01 ms). LCD and OLED displays use a the number of discrete tones they can represent. Finally,
sample-and-hold behavior that shows one image or video the output display produces a limited range of luminances.
frame until the next is available. The display is continuously Until recently, computer monitors and televisions sup-
backlit for the first frame, the next frame when refreshed ported standard dynamic range (retroactively referred to
and any time in between. This contributes to a perception as SDR-capable). Due to display technology and standard
of motion blur or ghosting when viewing motion media or definitions for broadcast television content, this limited the
when playing real-time rendered video games. Minimizing upper end of brightnesses to 100 cd/m2.8 OLED and QD
this time in between can help, but other strategies are technologies are pushing that upper brightness capabil-
used to help further such as black frame insertion. A black, ity to 1,000 cd/m2 and beyond. A benefit to such a bright
blank frame is displayed between two content frames at luminance display is its increased visibility in brightly lit
high refresh rates in an effort to lessen the appearance environments such as outdoors or in a living room with
of ghost-like images with fast-moving content. Another many windows. You do not want to edit photos in a dimly
approach is to rapidly strobe the backlight on and off. This lit studio with the display shining at the full 1,000 cd/m2,
can be done albeit with reduced luminance output rela- however.
tive to a display’s non-strobing mode.7 Dominant in the Additionally, their ability to render darker black tones
realm of gaming, such technologies are not as critical for relative to older displays means that the total range of
normal photo and video editing or viewing. Additionally, displayable brightnesses is significantly expanded. Mar-
they may introduce color shifts which is undesirable for keting-speak sometimes describes this characteristic of
color-critical work. a television as the contrast ratio. It’s based on the lowest

314
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

possible display luminance (Lmin) and the maximum possible requirements for resolution, high dynamic range, color
display luminance (Lmax). The greater the ratio between and other video and audio attributes with an eye toward
these values, the wider the scope between bright and dark. giving consumers the best possible UHD with HDR
LCDs, for example, tout contrast ratios of 1,000:1. Of course, experience.”10
taking advantage of such a capability means having high
dynamic range source content to output. High dynamic
range photo content has been achievable for a number Display Color Gamut
of years while video recording in HDR is a more recent
development. Recall that the gamut defines the range of addressable
HDR display standard adoption is in a state of flux as colors; the color gamut of a display defines the colors it can
of this writing. The primary contenders are Dolby Vision (a accurately create and show from input media. There are
proprietary solution requiring licensing), HDR10 (an open several additional named color gamuts when discussing
source solution) and Hybrid Log Gamma (royalty free, displays as shown in Figure 15-5.
co-developed by the British Broadcasting Corporation and The NTSC color gamut is perhaps the grandad of color
the Japan Broadcasting Corporation). Dolby Vision sup- gamuts, introduced in 1953 by the Federal Communi-
ports up to 10,000 nits even though no available display cations Commission. Developed by its namesake, the
is capable of reproducing that brightness and requires a National Television Standards Committee, this is the stand-
specific hardware decoder in the display. HDR10, created ard color gamut for analog broadcast television. The NTSC
by the Consumer Technology Association, supports up to is a wide gamut very similar to Adobe RGB, though they
1,000 nits and can be decoded in display software. HDR10 differ slightly in their red and blue primaries.
was adopted by Netflix, LG, Samsung, Sony and Vizio, to
name a few major players.9 Many television manufacturers
offer compatibility with both, at least until one becomes
the dominant choice for mastering, distribution and
broadcast.
It’s important to separate maximum brightness or con-
trast ratios from bit depth. A display can show an extremely
bright white and an inky black but its bit depth capabilities
define how smooth the gradation is between these two
extremes. Low bit depth translates to posterization, high bit
depth means natural tone transitions and a wider range of
displayable colors. Displays are 8-, 10- or 12-bit with devel-
oping technologies promising 14-bits and higher.
One of the challenges is deciding on a transfer func-
tion to convert the digital signal to optical energy given
the variety of brightness capabilities of different display
technologies. At the same time, some amount of back-
wards-compatibility with SDR displays is required at least
for broadcast television.
The UHD Alliance is an organization of over 40 film
and television studios, leading consumer manufactur-
ers, content distributors and technology companies. Figure 15-5 Comparison of the coverage of common color gamut
The organization is working to establish “performance standards.

315
 s ecti o n 3: O utp ut

Recommendation 709, also known as Rec. 709, was

introduced in 1990 to provide a standard for HDTV.11 Many
use the terms sRGB and Rec. 709 interchangeably as they
share the same RGB primaries and white point. Each uses a
different gamma, however: 2.2 for sRGB and typically 2.4 for
Rec. 709 (though the standard does not define it). Rec. 709
functions as the broadcast standard for SD and HD content
and is used in the majority of broadcast content today.
DCI-P3 was developed for digital cinema projection in
2007 by the Digital Cinema Initiative and published by the
Society of Motion Picture and Television Engineers.12 The
DCI-P3 gamut shares the same blue primary as sRGB and
Adobe RGB. The red primary extends to the spectrum locus
using a monochromatic light emission at 615 nm. DCI-P3
Figure 15-6 Measured spectral power distributions for the red, green
covers 45.5% of the chromaticity diagram (representing all and blue emissions of LCD and OLED displays.
of human vision), completely encompasses the sRGB color
space and covers 90.1% of NTSC and 93.6% of Adobe RGB
color spaces. color gamut. Figure 15-6 shows measured spectral power
Rec. 2020 (ITU-R Recommendation BT.2020) was pub- distributions for the additive primaries, red, green and blue,
lished in 2012 by the International Telecommunication for three common displays. The measurements show that
Union (ITU).13 It defines ultra-high-definition television the OLED display emits narrow band wavelengths at each
(UHD TV) with wider dynamic range and wide color gamut primary hue; the LED desktop and laptop measurements
(WCG). The standard defines expansions in color gamut, show that a shared display type may still show perfor-
frame rates, bit depth (10 or 12), resolution, luma-chroma mance differences.
color representations and subsampling relative to previous
broadcast standards. Rec. 2020 covers 75.8% of human
vision, 52.1% of Adobe RGB and 99.9% of Pointer’s gamut. Defining the Display Viewing
We mentioned WCG in Chapter 6 when considering trends Conditions
in motion video; film and television content is increasingly
captured with extended color gamuts, mastered and Chapter 14 covered soft proofing and how to best manage
streamed to viewers that are able to experience the differ- a workspace environment for an optimal color manage-
ence on newer displays. This expansion of data throughput ment workflow and how to calibrate a display. Here, we
and output can require additional connection bandwidth address the physical size of a monitor in that space. Select-
in the form of network speeds and/or display interface ing the proper size display for the environment can be a
connections and cables. topic of debate in many homes, particularly when it comes
A display’s ability to reproduce a range of colors or a to a home theater or living room setup. Personal opinions
percentage of a defined color space is directly related to aside, there are some objective guidelines that can settle
the color primaries of its sub-pixel elements. The higher the the debate.
saturation, color purity and range of luminances, the larger
the pool of possible colors that the display can repro-
duce when needed. It’s always a balancing act as highly Viewing Angle
saturated filters absorb more light and cause reduced We first consider viewing angle, a characteristic defining
brightness output while less saturated filters yield a smaller the range at which the display can be viewed off-axis while

316
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

retaining color and tone fidelity. A limited viewing angle

means that colors and brightness may shift when viewing
the display at any position other than head-on. Off-axis
viewing can reveal a much dimmer and color-inaccurate
view of displayed content. This is an important character-
istic for displays in environments where viewer position is
either varied or uncontrolled, such as movie theaters or in
open spaces. It is less critical for the photographer sitting at
Low Pixel Density High Pixel Density
a desk, well-centered and aligned with the monitor.
High Pixel Pitch Low Pixel Pitch
Viewing angle is dependent on the type of display
panel technology. Twisted nematic panels exhibit the most Figure 15-7 Keeping space between pixels minimal, larger pixels have
low pixel density and high pixel pitch (left) while smaller pixels have a
restricted view angle. Vertical alignment display panels higher pixel density and lower pixel pitch (right).
are considered to have moderate to good view angles.
In-plane switching panels exhibit the best performance
with many manufactures claiming a 178° view angle in
both the vertical and horizontal directions.14 It can also be Digital Systems and only requires the pixel pitch of the
a function of display shape: for a brief time, curved OLED display.15 Depending on the manufacturer, the pixel pitch is
panels were pushed by manufacturers for televisions and multiplied by a constant of 8, 10 or 15. This equation does
computer monitors. The former failed to catch on, perhaps not take into account human vision and is an estimation of
because making a display that curves inwards on either distance.
side makes for a worse experience when sitting anywhere
other than the center of the couch. Curved panels offer Optimal viewing distance (in feet) = Pixel pitch x 8 (Eq. 15.1)
marginally improved utility on a desktop, on the other
hand, as the curved sides provide a more consistent view- Optimal viewing distance depends on human vision
ing distance in a user’s periphery. which can resolve detail that is approximately 1 arc minute
or 0.01670°. When determining optimal viewing distance,
the pixel pitch must fit within this angle in the field of view.
Viewing Distance Individual pixels are visible if the pixel pitch exceeds 1 arc
Viewing distance is also worth evaluating and is ideally con- minute as illustrated in Figure 15-8 (overleaf).
sistent across sessions for both desktop workstations and Equation 15.2 is used to determine the proper viewing
home theater setups. There are many different methods distance for a display with a known pixel pitch or to deter-
for calculating optimal viewing distance, the distance where mine the required pixel pitch for a predetermined viewing
individual pixels cannot be seen and resolve to form a distance. The goal is the same in either case: to have the
clear image. This distance depends on the pixel pitch and proper combination of pixel pitch and viewing distance
pixel density. Pixel pitch is the distance from the center of that results in good image quality.
a pixel to the center of the next closest pixel. Pixel density
1 arc minutes
defines how many pixels there are in a given area (an inch, Pixel pitch = Viewing distance x tan (Eq. 15.2)
60 arc minutes
for example) as illustrated in Figure 15-7. A smaller pixel per degree

pitch and a higher pixel density results in a shorter viewing

distance just as a larger pixel pitch and low pixel density Viewing angle, viewing distance and environmen-
results in a larger viewing distance. tal lighting all play a role in selecting and optimizing
The simplest method is provided by display manufac- the performance and utility of your workspace displays.
turers in Equation 15.1. This method is put forth by Christie Many photographers have a desktop environment where

317
 s ecti o n 3: O utp ut

regular basis to counteract any drift or aging in the display

hardware’s performance.
More recently, color management software is inte-
grated into mobile operating systems and display profiling
solutions exist where they previously did not. This may
seem excessive for the average user of a phone or tablet,
but having color-critical settings and calibration is valuable
when using these convenient displays for remote tethering
or previewing of captured content.
The process of profiling a display involves some addi-
tional choices or options in addition to the measurement
of color patches. Let’s breakdown these other settings that
impact the behavior and appearance of displayed content.

Figure 15-8 Individual pixels are not visible if the pixel pitch forms an Recommended Settings and Conditions
angle with the eye of less than 1 arc second. Illustration adapted from
Creating an accurate display profile requires a proper
Extron.com
working environment. The recommended brightness
setting for a color managed, photographic workflow is a
luminance no higher than 120 cd/m2.16 This may look dim
they perform all of their image reviewing and editing. when working by the window at your favorite coffee shop
This allows for the selection of a display that best fits the or if you frequently keep your phone and laptop screens
environment, both in size, technology and capabilities. at maximum brightness. However, 120 cd/m2 appears
Laptops lack static or predictable viewing environments in perfectly bright in a properly lit editing environment with
exchange for portability—a tradeoff we’re happy to accept neutral surrounding tones, no windows in your line of
for on-location shoots and when keeping up with the sight and the help of a monitor hood to further block stray
workload during travel. light. Monitor hoods are purchasable accessories or can
be constructed with foam board and flocking material to
shield the monitor’s top and side edges. Additionally, any
Calibration and Configuration for objects in the room, including the table that the display is
Color Management sitting on, should be neutral in color. Non-neutral colored
objects may reflect a color cast onto the screen altering its
Display profile creation is inherently tied to color man- appearance.
agement and best practices for a photographic workflow.
Calibration involves measuring and adjusting the display
of brightness and color information from digital media Display Profile Characteristics
files. Thankfully, the process of calibrating and character- Creating a display profile is driven by software. There
izing our workstation monitors is aided by easy-to-follow are a few options to set prior to the calibration process.
software steps (assuming the measurement hardware is A common starting point is setting the white and black
on-hand). Desktop operating systems support setting the luminance levels of the monitor. This dictates the dynamic
custom ICC profiles created by this exercise as the default range available. This can be done prior to the calibration
for your computer and monitor combination. The hard- process or as the first step in the profiling software’s step-
est part is often the discipline in revisiting the topic on a by-step workflow. The contrast setting may also need

318
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

adjustment before calibration. This is evaluated by deter- Emerging Display Trends and
mining if there is a difference between a white and slightly Applications
non-white when displayed by the monitor. The software
determines if an adjustment is needed. Some display man- If you’ve ever purchased a television and then enviously
ufacturers do not provide the ability to adjust the contrast eyed the Black Friday sales the following year, you’re
setting. already familiar with the pace of consumer display technol-
White point selection for a display profile sets the color ogy development. Of course, most year-to-year changes
temperature or the appearance of white. Most monitors or offerings are marketing gimmicks or largely incremen-
range between 5000K and 6500K. The graphics industry tal upgrades. Still, display capabilities, form factors and
traditionally uses 5000K while video and web publishing applications never stop evolving. Manufacturers are always
industries use 6500K. A white point of 5000K appears too looking to improve performance with regard to resolution,
warm for photographic applications while 6500K is too frame rate, dynamic range and color gamut. Here are a few
blue. Neither of these is ideal or recommended for pho- additional categories to watch.
tographic editing and display. A white point of 5500K is
a good compromise between these two industry stand-
ards and should be selected if the native white point is Night Modes and Reduced Blue Light
unknown. Selecting the monitor’s native white point Exposure
provides the best dynamic range possible. It didn’t take long before we all became accustomed—
Additionally, the desired gamma or tone response curve possibly addicted—to staring at our smartphone screens
(TRC) must be selected. This defines the behavior of the and tablets late at night before bed. Scientists and restless
display between white (the brightest luminance) and Twitter users alike started wondering if exposure to bright
black. In other words, gamma controls display contrast. computer screens late into the night influences sleep
A value of 2.2 is used for photography, video and web. A quality for the worse. Indeed, studies show that “watch-
gamma of 1.8 is traditionally used by the graphic design ing a display at night can not only make it harder to fall
industry. asleep (up to 1.5 hours longer) but then the quality of
The last step of creating a display profile is measuring sleep during the night is also reduced.”17 Furthermore,
a series of color patches. The profile created from these exposure to blue light (460–490 nm) in particular seems
measurements defines the relationship between the to disrupt the body’s circadian rhythm by suppressing the
RGB digital counts driving the monitor and the CIEXYZ or production of melanin. Thus began the influx of mobile
CIELAB values measured. The more patches measured, the operating system software Band-Aids: adjusting the
more robust the profile created. Some high-end monitors color profile to skew very warm during late night hours is
designed for color critical work offer built-in color measure- thought to at least somewhat mitigate the problem. This
ment hardware that moves a probe into position over the is commonly done by limiting the use of the blue diodes
display and runs automatically at scheduled intervals. Such in the display. While this works for comfort and possibly
products may additionally use special lookup tables (LUTs) sleep quality, it is not ideal from a color management
internally, stored in dedicated memory, to further control standpoint. We expect to be staring at displays long into
and optimize the translation from the computer’s signal to the night for those endless photo editing sessions; future
displayable pixel values. displays may include designs that minimize emission
at the blue end of the spectrum for our health and
well-being.

319
 s ecti o n 3: O utp ut

Electronic Viewfinders of head mounted hardware, though this is quickly

One of the key features of the SLR design is its optical changing.
viewfinder. Putting an eye up to the viewfinder reveals a Many of us can detect a light source flickering below
projected image of the scene as it gets recorded on the 60 Hz. This is especially true with our peripheral vision.
sensor at the time of exposure. Electronic viewfinders The context of head mounted displays for virtual reality
(EVFs) display a live video feed of the image along with raises the minimum bar higher because flickering quickly
information overlays like exposure settings, focus points leads to user motion sickness. Common HMDs today have
and more. The display is quite small, as it resides inside the refresh rates of at least 90 Hz to avoid this uncomfortable
camera body and is only viewable when positioning an eye experience.
up to the viewfinder window. These small screens must
have high pixel densities to deliver an experience worth
abandoning that of our visual system viewing an optical Proof of Concept: Documenting the
image of the scene directly. Found in smaller camera form Growth in Mobile Device Display
factors, EVFs are a solution to the fact that the newer, Capabilities
non-SLR camera designs lack the functional design of the
pentaprism viewfinder. The pentaprism design allows for We can use our color management tools to look at display
the photographer to see exactly what the sensor sees— gamut capabilities in a simple way that quickly reveals
critical for composing and framing. EVFs are an alternative useful information. We’ve run an exercise in our classes for
that get back to this what-you-see-is-what-you-get visual a number of years now that asks students to measure pri-
feedback when shooting. mary color patches (R, G, B) on their smartphone screens
using photometers. We began this experiment as phone
screens were getting bigger, more colorful and higher in
Head Mounted Displays resolution in the early 2010s. This logically coincided with
Computers continue to shrink, their ubiquity and integra- a surge in visual media on mobile platforms and apps,
tion into our personal and work lives means that wearable even though the idea of mass media consumption on
solutions are the next frontier. Virtual reality (VR) uses head these devices just a few years earlier seemed far-fetched.
mounted displays (HMDs) that completely envelop the As soon as phone screens were considered good enough
user’s vision. Augmented reality (AR) is similar but requires to view and share photographs and motion media, their
the integration of the real world by introducing see- popularity exploded (as did the hunger for greater data
through displays. Both VR and AR demand considerable bandwidth).
and sometimes yet unseen display solutions. As photographers and avid media consumers, we hope
It’s too early to say if head mounted displays used that our smartphone screens are capable of representing
for virtual reality and augmented reality experiences photographic content accurately and completely. It’s
will have a major role in the photographic medium. The only in the last few years that mobile operating systems
display hardware is considerable in its boundary-pushing have incorporated mature color management software to
refresh rates, resolutions and user-input tracking. The ensure quality color reproduction. Much of this color man-
modern consumer HMD for virtual reality experiences agement and interpretation of any available, embedded
came to life thanks to driven-down manufacturing costs profiles is inaccessible to us as smartphone users. Still, we
of small display panels for smartphones, meaning that can get a general sense of their display capabilities with a
continued investment and adoption of such systems may simple set of measurements to find their color gamut.
lead to other unforeseen benefits to visual media. Display First, we measure the tristimulus values for the additive
pixel densities are still below the threshold for appearing primary colors: red, green and blue. We prepared a series of
seamless at the very close viewing distances required image files made in Adobe Photoshop where each consists

320
 c h a p t e r 15 : D i s p l ay T e c h n o lo gy

Figure 15-9 Measuring a smartphone display using prepared color Figure 15-10 A chromaticity diagram showing the gamuts of our
patch images and a photometer. measured displays: a high-end desktop monitor, the average gamut
for phones in 2012 and again in 2018.

of a single color. These colors are the purest red, green and LCD). We think it’s interesting to have a point of reference,
blue (respective color channel value: 255). We loaded these even if we don’t expect our phones to keep up with the
image files onto our devices and, one by one, measured capabilities of professional, dedicated display hardware.
them displayed full screen at 100% brightness with a pho- Figure 15-10 shows aggregate results from running this
tometer as shown in Figure 15-9. The photometer outputs exercise over a number of years. We took student-collected
tristimulus values for each patch. With XYZ values in-hand, data—hundreds of measured phone screens across a
we calculate chromaticity coordinates (x and y) and plotted variety of models and operating systems—and plotted the
them on a chromaticity diagram. average tristimulus values for the 2012 and 2018 school
Table 15-2 lists XYZ values averaged from the ones years. While this does not represent every available device
measured by our students in 2012 and 2018. It also shows from those years, the data is notably varied in models and
measured data from a professional desktop display (an screen types.

Table 15-2 Measured tristimulus values and calculated chromaticity values for the average mobile phone in 2018 and a professional desktop LCD.

Color Patch X Y Z x y

Mobile Phone (2018) Red 41.2 21.3 1.99 0.64 0.33

Green 33.3 60.1 9.85 0.32 0.58

Blue 15.4 7.38 77.5 0.15 0.07

24” Professional Grade Red 56.39 25.64 0.93 0.68 0.31

LCD Monitor
Green 21.03 68.08 10.32 0.21 0.68

Blue 17.36 5.69 92.79 0.15 0.05

321
 s ecti o n 3: O utp ut

This idea can be further extended by measuring a series 6 Chinnock, Chris. “Quantum dots will power display products
of grayscale patches ramping from black to white to gather to the next level.” Insight Media Releases Technical White
data on contrast, maximum luminance and white points. Paper on Quantum Dot Technology , Nov. 2016, www.
We additionally measure a pure white patch to investigate insightmedia.info/insight-media-releases-technical-white-
color casts and the behaviors of “night modes” or “true paper-on-quantum-dot-technology/., p. 18
7 Baker, Simon. “Response time measurement.” TFT Central
color” color temperature adjustment software frequently
Monitor Panel Part Database, Jan. 2013, www.tftcentral.co.uk/
found on mobile operating systems today.
articles/response_time.htm.
Questions to consider from this exercise and our col-
8 Nilsson, Mike. “Ultra high definition video formats
lected data:
and standardization ” Slidelegend.com, BT Media and
Broadcast Research Paper, Apr. 2010, slidelegend.com/
• What do you think is the primary contributing factor ultra-high-definition-video-formats-and-standardisation-bt-
to the increase in color gamut capabilities on mobile media-_59bf6bef1723dd5242bdfbc0.html.
screens? 9 Brennesholtz, Matthew. “HDR10 vs Dolby Vision.”
• Have the increasing capabilities of small displays led to DisplayDaily, Aug. 10, 2016, displaydaily.com/article/
their increased use for visual media viewing? display-daily/hdr10-vs-dolby-vision.
• Do you think that the average consumer notices the 10 “Members.” UHD Alliance, alliance.experienceuhd.com/
difference or improvements over the last few years? members.
• Do you believe that portable device displays are headed 11 BT.709: Parameter Values for the HDTV Standards for
toward professional-grade color reproduction and other Production and International Programme Exchange.
higher-end display characteristics? Are there character- International Telecommunications Union, June 2006,
www.itu.int/rec/R-REC-BT.709/en.
istics beyond color gamut described in this chapter that
12 The Society of Motion Picture and Television Engineers
are important for these displays?
(2010), SMPTE-EG-0432-1:2010 Digital Source Processing –
Color Processing for D-Cinema.
13 BT.2020: Parameter Values for Ultra-High Definition Television
Notes Systems for Production and International Programme
1 Allen, Elizabeth and Sophie Triantaphillidou. The Manual of Exchange. International Telecommunications Union, Oct. 14,
Photography. Elsevier/Focal Press, 2017, p. 292. 2015, www.itu.int/rec/R-REC-BT.2020-2-201510-I/en.
2 Young, Ross, et al. “White paper release: QLEDs – Quantum 14 Jackson, Harry. “Viewing angle defined and explained.”
dot technology and the future of TVs.” Display Supply Chain Monitors for Photo Editing, Helps You Choose the Best
Consultants, Nov. 15, 2017, www.displaysupplychain.com/ Monitor for Photo Editing and Photography, Feb.
white-paper-release.html, p. 4. 11, 2018, www.monitorsforphotoediting.com/
3 Young, Ross, et al. “White paper release: QLEDs – Quantum monitor-viewing-angle-defined-explained/.
dot technology and the future of TVs.” Display Supply Chain 15 “LED displays.” Christie – Digital Projection and Integrated
Consultants, Nov. 15, 2017, www.displaysupplychain.com/ Visual Display Solutions, www.christiedigital.com/en-us/
white-paper-release.html, p. 6. digital-signage/visual-display-technology/resources/
4 Chinnock, Chris. “Insight media releases technical white LED-displays.
paper on quantum dot technology.” Insight Media Releases 16 Ashe, Tom. Color Management & Quality Output: Mastering
Technical White Paper on Quantum Dot Technology, Nov. Color from Camera to Display to Print. Focal Press, 2014.
14, 2016, www.insightmedia.info/insight-media-releases- 17 Soneira, Raymond M. “Watching displays at night.”
technical-white-paper-on-quantum-dot-technology/. DisplayMate LCD Advantages and Disadvantages, www.
5 Young, Ross, et al. “White paper release: QLEDs – Quantum displaymate.com/Displays_At_Night_1.htm.
dot technology and the future of TVs.” Display Supply Chain
Consultants, Nov. 15, 2017, www.displaysupplychain.com/
white-paper-release.html. p. 7.

322
16 Printing and
Physical Media

Photograph by Eric Kunsman

323
 s ecti o n 3: O utp ut

This chapter was written in collaboration with Eric Kunsman, the number of megapixels (a measure of pixels dimen-
Lecturer at Rochester Institute of Technology and owner of sions, width multiplied by height). From pixel dimensions
Booksmart Studio, Inc. in Rochester, New York. we determine print size when choosing how many pixels
___ per inch (ppi) to send to the printer for a given print size.
There is great debate between photographers as to the
Printing technology is expanding rapidly to meet the proper pixels per inch needed to print. Today, Epson
demands of the average consumer, the professional printers render upwards of 360 ppi and Canon 300 ppi
photographer and everyone in between. Photographic of spatial resolution on a high-end glossy substrate. Print
prints, once created solely with light-sensitive materials, resolution is a logical place to start an overview of printing
now employ various forms of ink or toner and sophisti- fundamentals.
cated printer hardware. High-quality printing is now easily
achieved in a home studio or office. The print media indus-
try is constantly improving printing technologies, whether Pixels Per Inch and Dots Per Inch
they are printer manufacturers, ink and toner developers, Pixels per inch is a measure of the resolution of a display,
paper companies or print finishing companies. The push scanner or image sensor; it’s a count of how many pixels
for “larger is better” has never been as true in the output exist in a square inch of space. This is different from pixel
field as it’s been since the beginning of the twenty-first density which is a function of the number of pixels in height
century. and width that are then rendered using dots of inks when
This chapter aims to inform photographers about availa- printing. We must decide on the printed spatial resolution
ble printing media and technologies and to provide insight to use with a 3,600x4,800-pixel image file, for example. If
into their best strengths and applications. We review we send the file to the printer at 300 ppi, we divide the
the fundamentals of what it means to output images to pixel resolution by 300 to determine the print dimensions
print media, the printer hardware, the papers and other in inches as calculated in Equation 16.1.
substrates used and a variety of other factors to consider.
Printing to physical media represents a key balance of 3600 pixels
by
3800 pixels
= 12 x 18 (Eq. 16.1)
300 pixels per inch 300 pixels per inch
vision, craft and technical precision; it’s how a great deal of inch print
photography is discovered and experienced even in our
increasingly digital world. This calculation indicates the print dimensions possible
with the native pixel resolution, though we always have the
option of interpolating and upscaling the image data. Print
The Fundamentals of Printing size relative to viewing distance is discussed later in this
chapter.
Printing for photographic output is a subtractive color The number of ink dots that physically fit in an inch on
mixing process that lays down color material onto a receiv- the substrate is measured as dots per inch (dpi). This is a
ing substrate. A substrate is any surface onto which an ink logical way to talk about print resolution in terms of the
or dye can be printed. The fidelity of the printed reproduc- physical printing mechanics. As a printer does not print
tion of an image file is dictated by its detail, color, tonality pixels, dots per inch is an appropriate term to describe
and sharpness, among other criteria. printing behavior. The greater the dots per inch metric, the
A camera captures an image with a set resolution more dots of ink a printer lays down in a square inch area
defined by the size and density of the photosites across of the receiving substrate. Dots per inch does not tell us
the area of a digital sensor. The native sensor resolution anything about pixel resolution.
dictates the maximum available pixel dimensions of our
captured image file. Often, photographers only care about

324
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Continuous Tone and Halftone Printing

Printing dots is accomplished in one of two ways: con-
tinuous tone or halftone. Continuous tone digital printers
produce output similar to the analog color photographic
process by writing individual image dots directly to paper
or film. The dots vary in density and the dyes used are
transparent. This allows for placing one color directly on
top of the previous color. A printer using the three sub-
tractive primary colors of cyan, magenta and yellow is
theoretically capable of reproducing more than 16.7 million
colors. The color gamut of the dyes used determines the
actual number of reproducible colors. Photosensitive light
written and thermal transfer printing technologies are Figure 16-1 An example orientation for halftone screen ink dot
patterns.1
continuous tone options available today outside of the
traditional darkroom print. The use of a light source writing
the image to a photosensitive material aids in the creation
of continuous tone. created for each color used. These screens break the image
Halftone printing uses opaque dots that do not vary up into a series of dots of varying frequencies and sizes
in density. Halftone dots can’t be placed on top of one corresponding to the amount of that particular color in
another without obscuring the previous dots. These dots the image. The resulting films are used to expose printing
are printed in patterns that blend to the eye and are plates. Each color uses its own angle to prevent the dots
perceived as various colors when viewed from a distance. from perfectly overlapping. Figure 16-1 illustrates one
The process of printing these small dot patterns is called possible combination of halftone ink dot orientations. The
halftone screening. Halftone screening does not allow for angles are adjustable depending on the lines per inch to
continuous tone density because the ink dots are over- avoid introducing printing artifacts like moiré patterns.
lapping, sized differently or spaced differently to create The AM screening method is used today for traditional
the illusion of continuous tone. The finer the dot (or dot offset lithography and by the HP Indigo printing press.
size) and the higher the frequency (the dots per inch), the Offset lithography does not print directly onto the sub-
more the image appears as continuous tone to the viewer. strate or paper. Instead, a printing plate is created for each
The space between the dots of ink reveals the underlying color which initially receives the ink. The ink is then trans-
substrate is a part of the process whereby our visual system ferred onto the printing blanket, typically made of rubber,
creates the illusion of continuous tone. which then comes in contact with the substrate to transfer
Halftone printing uses amplitude modulation (AM) or the ink. The contact between the printing blanket and the
frequency modulation (FM) screening methods. A loupe or substrate is facilitated by an impression cylinder that creates
magnifying glass is necessary to distinguish between the contact pressure between them. The printer must trans-
two. The AM screening method was first used for early fer all of the toner from the blanket to the substrate and
offset lithography; the image was exposed through screens therefore employs the AM halftone method. The HP Indigo
onto film with a grid-like pattern and a uniform spacing supports on-the-fly changes to screening resolution and
between differently sized dots. This grid-like pattern is the angles of each color during a print job.
called lines per inch (lpi) as this trait determines the space FM screening is also known as stochastic screening. The
between dots and their size. The higher the lines per inch, method came about when computers were first used
the finer the dot size. The physical dot size is determined to create the films that then made the printing plates
by the exposure of that dot. A film, also called a screen, is for offset lithography. FM screening uses a random dot

325
 s ecti o n 3: O utp ut

Figure 16-2 Amplitude modulation screening patterns compared to frequency modulation screening patterns. Illustration courtesy of Tom Ashe

pattern to create a frequency modulation that provides an magenta and light cyan to improve color gamut and
improved, continuous tone appearance when compared tonality. This practice became less common over the last
to AM screening patterns. Inkjet printers and dry toner elec- decade, however, as the quality and gamut capabilities of
trophotographic printers (described later in this chapter) the traditional subtractive primary colorants improved.
use a newer method of FM screening that leverages dither-
ing to further improve the continuous tone appearance.
Dithering is a technique of printing a series of tiny dots Printer Technologies
to create the illusion of a color that can’t be directly pro-
duced by the printer. Altering the dither pattern changes There are numerous printing technologies used in the
the color appearance thanks to partitive mixing. Dithering, diverse world of printing applications. This chapter
combined with stochastic screening, creates a random dot concentrates on inkjet printing used for professional
pattern and spacing with variable dot sizes. Figure 16-2 photographic output but we’ll also describe light-sensitive
illustrates the traditional halftone AM screening, traditional printing, electrophotographic and thermal transfer printing
FM stochastic screening and the newer dithering and technologies.
stochastic method.
Manufacturers are constantly working on dot size and
halftone methods to create the most continuous tone Inkjet Printing
image possible with each new generation of inkjet printers. Inkjet printing uses a printhead that deposits small amounts
This requires developing ink formulations that allow for of colorant onto a receiving substrate. The mechanics of
smaller printer nozzles and incorporating the capabilities inkjet printing allow for a wide range of print sizes: since
into their screening and printing software solutions. the printhead is mounted and travels along a linear rail,
For many years, almost all technologies using ink the maximum print width is extended by making the rail
or toner introduced additional colors like orange, light longer. Conversely, inkjet printers can be sized to sit on a

326
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

desktop and output smaller prints. The former approach operating challenge with inkjet printers. The printhead
is considered wide format inkjet, the latter, desktop inkjet. switches to one of its many extra nozzles if a clogged
Wide format inkjet printers are categorized by their ability nozzle is detected. Despite these extra nozzles, thermal
to output print sizes 24 inches wide or larger. Inkjet printing printheads eventually run out of unclogged ones and must
is a rapidly evolving technology because of the enormous be replaced by the user; these printheads are a consum-
amount of development dollars spent on it. It’s also one of able printer component. Canon and HP home desktop
the few affordable technologies for the home user. printer printheads are built into the ink cartridges which
There are two inkjet printhead technologies to con- drives the cost of replacement cartridges higher. The
sider and both use drop-on-demand (DOD) mechanisms printhead hardware is replaced when the ink cartridges
to deposit ink in a controlled manner. The first printhead are replaced even though it may have useful operating life
technology is the thermal or bubble jet system found left. Professional-grade printers employ different printhead
in printers made by companies including Canon and design approaches. Canon uses either a single or double
Hewlett-Packard (HP). Each ink cartridge in a desktop printhead for up to 12 colors; HP models use one printhead
inkjet printer contains a series of tiny nozzles and an for every two colors.
electrical connection to the printer. The ink is routed into Instead of using a heating element to force ink out of
individual chambers, each with a tiny resistive heating the nozzle, piezoelectric inkjet printers use a small piezo-
element. Applying a pulsed current to the heating ele- electric crystal that expands and changes shape when
ment heats the ink until it vaporizes, causing a bubble to a current is applied. This change forces ink out of the
form. The expanding bubble forces the ink through the nozzle chamber while also refilling the chamber for the
small opening in the chamber and sprays it toward the next firing of the nozzle. The piezoelectric jet is like a small
paper as illustrated in Figure 16-3. The vacuum created in electromechanical pump that precisely delivers a small
the chamber draws more ink into the nozzle to repeat the amount of ink. Since the shape of the piezoelectric element
process. This process allows for depositing variable droplet changes based on the amount of charge applied, the size
sizes onto the substrate. of the ink dot is variable. This technology is used by many
Thermal printheads use self-detection mechanisms to printer manufacturers in their desktop, wide format and
monitor for clogged nozzles, a commonly encountered high-speed inkjet printers. The difference between man-
ufacturers’ offerings is the number of nozzles per inch for
various ink types and viscosities. Epson printers traditionally
use 360 nozzles per inch that force ink droplets out of the
chamber at a size of 3.5 picoliters (the physical size of a
picoliter is approximately 20 microns).
Current piezoelectric technology is capable of reso-
lutions up to 2880 dpi. The printer spatial resolution can
produce upwards of 360 ppi accurately rendered with dots
of ink as compared to the 300 ppi possible using thermal
printhead technologies. The print quality between piezo-
electric and thermal printheads is indistinguishable to the
human eye without the use of a magnifying loupe.
Continuous-flow inkjet printers pump a continuous
stream of electrically charged ink droplets (64,000–165,000
Figure 16-3 Drop-on-demand inkjet printheads use a heating element per second) at the paper as illustrated in Figure 16-4 (over-
to create a bubble of ink or a piezo element pushes a drop of ink onto
the paper when requested by the image data stream. Not illustrated is leaf). A magnetic field controls ink placement on the page;
the printhead’s ability to vary ink droplet size. unused ink is directed to a recovery reservoir.

327
 s ecti o n 3: O utp ut

The Durst Lambda printer is one popular laser imaging

photosensitive digital printer on the market. Using a series
of laser diodes to write an image onto photographic paper,
the Lambda printer handles rolls of paper up to 50 inches
wide, exposing the paper as it moves past the lasers. The
paper materials are sensitized depending on the light
source used. The resulting chromogenic print, otherwise
known as a digital c-print, uses color coupler dyes (cyan,
magenta and yellow) in an emulsion layer with light-sen-
sitive silver halide crystals. Digital chromogenic prints
are made on traditional paper substrates or translucent
materials.
Light-sensitive silver halide crystals are suspended in a
Figure 16-4 A continuous-flow inkjet printhead pumps a stream of ink gelatin emulsion which is then coated onto a carrier substrate
droplets that get redirected by a magnetic field to get deposited on
(the supporting material). A chemical layer of barium sulfate
the substrate or recycled.
also known as Baryta is applied to the carrier substrate to
create a surface on which the light-sensitive layer sits. This is
Continuous-flow inkjet printers are available for several different from inkjet papers that use a microporous coating
distinct applications. The high-speed, low-resolution variety atop the Baryta layer. Once the image is exposed, the paper
is used to print product identification (labels), personalized is developed with a chemical process that transforms the
direct mail and magazines. High-resolution, low-speed exposed areas into silver metal. The unexposed silver halide
printers like the Iris series by Scitex were used for proofing is removed. Fiber-based prints are considered much more
and fine art printing for many years. Iris printers produced archival than their resin-coated counterparts but they require
multiple spots per dot and achieved effective resolutions an extensive washing stage to remove residual chemistry.
of up to 1,800 dpi. In part because of the great number These prints are as archival as traditional darkroom prints,
of paper surfaces on which it printed (especially heavier around 25 years, and offer better scratch resistance than
weight, fine art watercolor papers), the Iris is the print- inkjet prints thanks to the emulsion layer.
ing technology that facilitated the acceptance of fine art Printer hardware like Kodak’s Pegasus LED Printer print
reproduction by museums and collectors. Continuous-flow directly onto special digital photographic paper using a
technology is also employed by some high-speed inkjet small array of LEDs. The paper is fed into a drum where a
printers to help deliver faster output speeds. vacuum pump pulls the paper tight against the inside of
the drum. The Pegasus handles roll sizes up to 20 inches
wide and the drum accommodates a 33-inch long piece of
Photosensitive Light Writing Printing paper, allowing for a maximum 33x20 inch print size. LVT is
One of the more cost-effective continuous tone printing another company manufacturing printers that use LEDs to
methods available today is photosensitive light writing expose photographic film and paper. Instead of the paper
printing. An RGB laser or set of light-emitting diodes being drawn by a vacuum to the inside of a drum, LVT’s
(LEDs) exposes an image onto light-sensitive material. printers wrap the paper around the outside of a drum.
The light-sensitive material is then processed using RA-4 LED printers expose the image onto the paper using
development chemistry just like a traditional darkroom a small array of four red, four green and four blue LEDs
print. This process produces prints on chromogenic mounted at the end of a counterbalanced arm. The output
papers as well as silver gelatin papers for black and white of the red, green and blue diodes is combined with an
reproduction. optical lens assembly to create a single, focused spot. The

328
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

arm spins on a spiral track at approximately 1,350 revolu- Electrophotographic Printing with Liquid
tions per minute (RPM) and the LED assembly is positioned Toners
just above the paper’s surface. The image is written as the The only liquid toner printing press available at the time of
arm travels down the length of the drum on the spiral writing is the HP Indigo digital press. HP Indigos are used in
track. Once the image is written and the paper exposed, it’s commercial production for print applications such as mar-
advanced into an attached processor for development. It keting collateral (brochures, business cards, posters), direct
exits from the machine fully dried in about 6 minutes. mail, labels, folding cartons, flexible packaging, books,
manuals and specialty applications. The Indigo’s ability to
print without films and printing plates enables personal-
Electrophotographic Printing with Dry Toners ized short runs with variable data by changing text, images
There are two main types of electrophotographic printing and jobs without having to stop the press. Photographers
technologies today: those that use dry toner and those that enjoy the HP Indigo press when printing photography
use liquid toner. Dry toner technology is used by a variety of books because of their ability to mimic a traditional offset
manufacturers with each having slight differences in their lithography press and to leave the substrate characteristics
mechanics. With either toner type, electrophotographic intact.
technologies allow for print on demand (POD) workflows Each press consists of up to seven color stations which
because each time an impression or image is written with a use cyan, magenta, yellow, black and a variety of spot color
new charge to the imaging drum. inks. Spot colors are either premixed inks or specific pure
Dry toner technology was initially developed for pho- colors that may be difficult to create by mixing the avail-
tocopiers but found its way into desktop printers in 1985 able CMYK inks. Presses may also use special inks such as
with the introduction of the laser printer. A laser, directed white, UV red, security colors and transparent (for a spot
via a spinning mirror, alters the magnetic charge on a varnish). The ability to use spot colors is common with
rotary drum. Toner particles are attracted to these areas non-digital offset lithography presses and is one of the
of the drum. Toner particles are transferred to the paper features that distinguishes the HP Indigo digital press.
as the drum rotates into contact with it. The paper is then The Indigo’s electrophotographic liquid toner tech-
transported through a pair of heated rollers which fuse the nology is based on HP’s ElectroInk which uses charged
toner particles to the paper. The color process version adds pigment particles suspended in imaging oil that’s attracted
cyan, magenta and yellow toners and requires exposing or repelled by means of a voltage differential. The liquid
the electrostatic drum once for each color. toner solution starts out as a mixture of 80% imaging oil
Dry toner machines use FM screening just like inkjet and 20% toner along with the proper amount of imaging
printers. Initial dry toner technology was limited to CMYK agent to make sure that its conductivity is at the proper
toners; some printers now offer silver, gold, green, orange level. The ink particles are small to ensure that the printed
and white toners that expand color gamuts. Manufacturers image does not mask the underlying surface texture of the
are also working on toner development to increase the paper (as is possible with some toner-based processes).
color gamut of dry toner printers. Additionally, the press creates a thin, smooth plastic layer
Prints made with fused toner are relatively inexpensive on the substrate surface to accept the ink. The physical
per sheet compared to other technologies discussed in ink particles make up a halftone pattern and larger dots
this chapter. Dry toner electrophotographic machines are that mimic conventional offset lithography. The physical
migrating from older laser and mirror technologies to LEDs ink particles are smaller than dry toner because they are
to write to the imaging drum. Newer printers also use belts suspended in imaging oil and cannot become airborne.
instead of rotary drums to pass the substrate between the The process begins with electrostatic charging of the
imaging roller and a charge corotron (a component that electrophotographic photo imaging plate (PIP) mounted
regulates voltage). on the imaging cylinder. The initial charge is completed by

329
 s ecti o n 3: O utp ut

Scorotrons at 5,000–5,800 volts. The initial charge is nega- Thermal Dye-Diffusion Printing
tive and floats on top of the PIP and a thin layer of imaging Sony and Kodak both introduced printers in the mid-1980s
oil. The PIP is light-sensitive and the charge is controlled by that produced a continuous tone image from a video
an auto-bias. The PIP is exposed by a scanned array of lasers source using a thermal dye-diffusion process (also called
shooting through the negative charge. These lasers are thermal transfer). The thermal dye-diffusion process uses a
controlled by the raster image processor which converts two-part medium. The first part is a clear mylar film called
instructions from a digital file into on/off instructions for a donor ribbon onto which consecutive transparent cyan,
the lasers (described in more detail later in this chapter). magenta, yellow and optimizer dye patches are coated. The
The positively charged, exposed image area attracts the ink. second part is a paper-based material called the receiver.
Inking of the PIP occurs via the highly charged binary ink The donor material resides on a motor-driven spool. A
developer (BID) roller. thermal printhead containing thousands of individually
Printing begins when the mechanical roller pushes ink controllable resistive elements heats the dyes on the donor
through the BID while applying a negative charge to the material which causes the dye to transfer from the donor
ink. A squeegee removes the imaging oil and applies a to the receiver (see Figure 16-5). The amount of transferred
negative charge to it, forcing the imaging oil away and the dye is controlled by varying the amount of current (and
toner particle toward the developer roller. The developer thus heat) sent to each of the resistive elements on the
roller creates the positive charge for the toner particles as printhead. The process prints one color at a time and the
the imaging oil is removed. At this point, the ink solution paper is rewound back to precisely the same starting point
percentages are reversed to 80% toner and 20% imaging so that the next color is placed in registration.
oil. The BID moves closer so that the toner can jump to the The dyes in a thermal dye-diffusion printing process
PIP’s image areas. are transparent and their printed densities are variable by
Next, the PIP transfers the inked image to the blanket up to 255 different levels for each color. The color dots are
on the blanket cylinder which is heated and helps to placed on top of one another, combining to create any
create a thin layer of plastic. The heated inked image is of the 16.7 million theoretically possible colors in a 24-bit
transferred to the substrate held by an impression cylin- image. Figure 16-6 shows the resulting image viewed
der. The impression cylinder must have the right amount under magnification. The image is somewhat blurred
of contact to properly transfer the image. Most HP Indigo due to the nature of the solid-to-gas-to-solid events of
presses transfer each color one at a time using this process. dye-diffusion.
However, roll-to-roll and other presses allow for printing
using a process called one-shot. One-shot collects all of the
colors onto the blanket and only requires one round of
contact with the substrate. This allows for perfect registra-
tion and the ability to work with substrate rolls. If duplexing
is required, the substrate waits in the perfecting unit while a
second sheet is started. Once the second sheet’s first side is
printed, it releases the first sheet to print the second side.
The technical process described earlier gets the toner
onto the substrate. There are many consumables on these
presses and many controls not discussed here that ulti-
Figure 16-5 The donor ribbon transfers dye to the receiver paper
mately determine print quality. The electrophotographic when heated. Illustration courtesy of the Image Permanence Institute
liquid toner press requires a trained technician to maintain
it—unlike most dry toner machines.

330
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

pigments are all types of colorants. Photographic printing

technology takes advantage of this variety of colorants
depending on the application and intended appearance.
When discussing photographic printers, a dye-based
ink is a water-based or aqueous solution into which a dye,
the colorant, is dissolved. The dye attaches to the water
carrier and stays suspended in the solution. This is similar
to adding sugar into a cup of coffee: it cannot be removed
or separated once dissolved. Different dye colorants create
a variety of ink colors. Dye-based inks are absorbed into a
paper substrate while the water evaporates, leaving the
Figure 16-6 A magnified look at a thermal dye-diffusion printed colorant behind. Dye-based inks offer great vibrancy in
image. Photograph courtesy of the Image Permanence Institute color but are prone to damage from exposure to ultraviolet
(UV) radiation.
A pigment is a compound of many molecules clumped
together and forms a finely ground powder colorant.
Dye Sublimation Printing Pigment particles are often many times larger than dye
Dye sublimation printing, also called dye diffusion, is a pro- particles. A pigment-based ink is comprised of a pigment
cess that first prints onto transfer paper with unique dye suspended in water in the context of photographic print-
sublimation inks. Using a heat press, typically around 400ºF, ing. The pigment settles out from the water given enough
the transfer paper is placed on top of the receiver substrate. time. When printed, pigment often sits on top of the paper
Heat and applied pressure cause the dyes to sublimate into and bonds to the surface in a thin film. Pigments are not as
the substrate fibers. This yields an image with vibrant color severely affected by UV exposure relative to dye and do not
and good durability. The heated printing process turns easily smear. They tend to be more expensive than dye-
solid ink into a gas, avoiding a liquid stage. based inks.
In traditional dye sublimation printing, the fabrics
needed to be white or light in color and made of 100%
polyester or a 50% polyester, 50% cotton mix. Today, Aqueous Inkjet Inks
microfiber and nylon are potential options as well. The pro- Aqueous inkjet printers are the most widely used inkjet
cess can also be used on ceramic, wood, glass and metal technology and are commonly found in homes and pho-
as long as they have a special polymer coating to accept tography studios. They are affordable entry-level options
the ink in its gas stage. Transfer papers and ink formulations but the substrates and inks used are the most expensive
now exist for popular elastic fabric materials and for all of all inkjet printing technologies. Aqueous inkjet inks
kinds of fabrics including natural fiber fabric like cotton, silk, use a water-based carrier for the colorant. Special coat-
linen and wool. ings on the paper are used to accept the colorant while
allowing the water to evaporate. These coatings drive up
the cost. Both dyes and pigments are used as colorants
Ink Technology in aqueous inkjet inks with most printer manufacturers
moving to pigment-based inks. Canon, Epson and HP are
Colorants, inks and pigments are three commonly used constantly introducing new inkjet printers to the aqueous
terms when explaining ink technologies and materials. inkjet market. Often times, new versions or generations
It’s important to clarify the distinctions between them. A see upgrades in the ink technology used in an effort to
colorant is any substance used to add color. Inks, dye and improve printer color gamut capabilities. Inkjet inks are

331
 s ecti o n 3: O utp ut

not 100% opaque, allowing for the blending of colors and low-solvent or light-solvent inks. Outgassing continues for
densities by overlaying inks. up to 24 hours on both solvent and eco-solvent inks, so
Dye-based inks offer a larger color gamut than there is a waiting period before beginning any finishing
pigment-based inks but tend to fade at a faster rate. processes, particularly lamination.
Pigment-based inks are more stable as they have larger The popularity of eco-solvent inks grew in the last
insoluble organic particles, however, this makes them more decade as their color quality and durability increased while
prone to scratches and scuffs. They also require a micro- their cost concurrently decreased. Eco-solvent inks contain
porous coating rather than a swellable polymer. Therefore, solvents and binders made from renewable crops like corn.
pigment-based inks are never truly encapsulated within However, the liquid solution in eco-solvent ink comes from
the inkjet receptive coating. Both dye-based and pig- ether extracts taken from refined mineral oil, so it’s inac-
ment-based aqueous inks offer a wider color gamut when curate to think that this is an entirely ecologically friendly
compared to other available inks thanks to the introduction alternative. Printers using eco-solvent inks use heat to
of colorants such as green, orange, red or blue into the evaporate the solution. Eco-solvent ink is usable on many
printers. uncoated substrates but takes longer to dry than solvent
ink. Prints made with eco-solvent ink last for approximately
two to three years before beginning to fade.
Solvent Inks
Solvent inks are the backbone of the digital inkjet print-
ing industry for outdoor signage applications. Solvent Ultraviolet Curable Inks
inks use an oil-based solution to hold the pigment and Ultraviolet curing is a photochemical process that uses a
resin. Solvent inks are used to print on different uncoated high-intensity ultraviolet energy source to instantly cure
banners, papers, flexible vinyl substrates, vehicle wraps or dry inks. Liquid monomers and oligomers are mixed
and adhesive decals. This technology is fade-resistant, with a small percentage of photoinitiators and exposed
waterproof and abrasion-resistant: all important quali- to UV energy. The ink instantly hardens to become solid
ties for outdoor usage. Unlike many aqueous inks, prints polymers in just a few seconds of exposure. This process
made using solvent-based inks are generally waterproof is widely used in the silkscreen industry. In fact, UV cure
and UV-resistant without special over-coatings. Solvent inkjet printing is now the fastest growth area in both
inks emit volatile organic compounds and ventilation is usage and technological advancements since 2015.2
required when working with them. The substrate used The inks and light sources are constantly improving in
when printing with solvent inks is often heated immedi- performance.
ately before and after the printheads apply ink; the high UV curable inks do not evaporate but rather set as a
print speed of many solvent printers demands special result of the chemical reaction. No material is evaporated
drying equipment. or removed, therefore 100% of the delivered ink volume is
Solvent inks are offered in the standard cyan, magenta, used to produce the printed dot. The curing reaction hap-
yellow and black. Additional colors such as orange, green, pens very quickly, leading to instant drying that results in a
violet, light cyan, light magenta, light black, white and even completely cured print in seconds. This allows for imme-
silver or metallic are also available options. diate print quality inspection. The fast chemical reaction
Solvent inks are subdivided into two categories: hard prevents ink from penetrating the substrate, allowing for
solvent ink offers the greatest durability without special- highly controlled dot placement.
ized over-coatings but requires special ventilation of the UV curable inks were created as an alternative to
printing area to avoid exposure to hazardous fumes. Mild solvent-based products for the signage industry. They
or eco-solvent inks are intended for use in enclosed spaces contain no solvent or water, meaning that there is no loss
without special ventilation. They are sometimes called of ink thickness or substrate shrinkage and substrate edges

332
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

do not curl. UV curable ink manufacturers are constantly This is good for items used indoors such as printed
working to improve on flexibility, durability, color gamut, wallpaper or signage and items are durable for indoor or
reducing surface texture and printability. outdoor usage.
UV cure printers are expensive compared to aqueous Latex inks require heaters in the printers as the substrate
and solvent printers and the cured ink’s thicker viscosity becomes receptive to the ink through a pre-heating step. A
leaves a slight relief on the print surface. The viscous inks post-heating step allows the ink to properly dry. A substrate
are more susceptible to cracking if applied to a flexible sub- that exhibits rapid expansion or contraction due to heat
strate relative to solvent inks, though the new formulations is problematic when using latex inks. At the same time,
to address this issue are being actively developed. There these machines have higher electricity demands compared
are two main classifications of inks used in this process: free to other inkjet technologies because of the need for the
radical and cationic; there are likely to be more alternatives heating elements.
in the coming years. HP was the first printer company to introduce latex
The free radical UV curable ink is sensitive to oxygen, so printing technology and only within the past few years
the curing process must be quick. This often means using have competitors started to offer their own versions. Ricoh
a greater intensity UV source and more photoinitiators in and Mimaki introduced a machine that requires a lower
the ink that allow for oxygen quenching. The increased amount of heat; Mimaki’s latest machine states that it only
intensity causes more heat on the substrate and some needs 60°C while the HP printers need around 120°C.
substrates are not built to accept that heat. Adding more
photoinitiators to the ink adds to the cost and produces
more of an odor. Dye Sublimation Inks
Cationic inks are not sensitive to oxygen, require less There are various types of ink chemistry used for the dye
UV energy to initiate the curing process and are fully sublimation process when printing on textiles. Using the
cured after exposure. These inks are also known to provide proper combination of inks with the transferred fabric type
improved flexibility, adhesion, hardness, chemical resist- is important. Cotton, rayon, linen and silk must undergo
ance and opacity with a higher gloss and less of an odor a pretreatment prior to printing and often use reactive
than free radical inks. However, cationic inks are known to dyes in their manufacturing. Acid and reactive dyes are
cure slower than free radical inks and may require pre-heat- essentially steamed to fix the dyes to the fabric and must
ing of the substrate to accept the ink. The ink formulations be washed afterwards to remove any extra or non-fixed ink
are often more expensive and are sensitive to moisture remaining on the fabric. Dye sublimation manufacturers
along with acids and bases. offer a variety of fluorescent colors to help with matching
of a specific brand color, as well as expanding the overall
color gamut of their machines.
Latex Inks
One of the latest developments in inkjet technology are
aqueous latex inks that use latex as the pigment car- Printing Substrates
rier solution. These inks offer an alternative to solvents
and UV curable inks in the signage market. The goal is Each printing technology presents unique requirements for
to replace solvents and reactive ingredients with less preparing the substrate surface to accepting the colorant.
hazardous materials and water. This technology offers Inkjet substrate coatings, for example, help the dye or
comparable durability to solvent inks on numerous sub- pigment colorant to penetrate the surface without pen-
strates, including uncoated vinyl. The use of water-based etrating too far thus preventing proper drying. Choosing
formulations in latex inks allows for a more environmen- the proper paper for your work is overwhelming as there
tally friendly finished print that requires no ventilation. are many choices available for photographic output. Here,

333
 s ecti o n 3: O utp ut

we explore the primary physical media substrate types to

consider before diving into specific characteristics in the
next section.
Note that the term photo paper previously referred
exclusively to light-sensitive substrates used to print in
the darkroom. Today, the term includes these options but
also extends to papers used with ink-based digital printing
processes. Fine Art photo papers used in inkjet printing are
specially coated to receive dye-based or pigment-based
inks and can be made of cotton, alpha cellulose, bamboo,
sugar cane, silk, rice and many other materials.
Figure 16-7 A 200x magnified cross section of RC swellable polymer
coated paper. Photograph courtesy of www.graphicsatlas.org
Aqueous Inkjet Substrates
There are two main coating types for aqueous inkjet print-
ing: swellable polymer used for dye-based inks and porous
inkjet used for dye or pigment-based inks. The subcatego-
ries of porous substrates compatible with pigment-based
inks are microporous and nanoporous. Macroporous is
compatible with dye-based inks.
A swellable inkjet coated paper absorbs dye-based
ink as it is sprayed onto the paper. This results in the
paper swelling around the ink. Swellable paper requires
time to dry and smears if handled while wet (though
some instant-drying papers exist). Swellable papers offer
high gloss surfaces and are resistant to scratching and
fading once dry. Swellable papers cannot be used with Figure 16-8 A 200x magnified cross section of uncoated paper
printed with dye ink. Photograph courtesy of www.graphicsatlas.org
pigment-based inks as the pigment is not absorbed into
the paper and can rub off; the process of swelling and
encapsulating works because of how thin dye colorant is
compared to pigment.
Both swellable and porous paper types may embed the substrate without completely absorbing into the paper
resin-coated (RC) layers. The top RC layer sits directly under support.
the swellable or porous coating to prevent moisture from Microporous coatings are commonly used for pig-
absorbing into the paper base (moisture causes the paper ment-based inks. The ink is absorbed into the coating layer
to curl). The second RC layer is applied to the back of the and remains there once dry. Figure 16-9 shows pigment ink
paper for the same reason. This layer is clear. printed onto the coating; it does not absorb past it to the
Figure 16-7 shows how the printed dye sits in a swella- polyethylene or paper support base layers below. Micro-
ble polymer coating and does not absorb lower into the porous coatings are often employed with RC glossy and
paper support. Attempting to print with pigment ink luster papers. You can often identify them by the backing
results in the colorant sitting on top of the coating that polyethylene coating because it feels like plastic.
will not dry. Figure 16-8 shows the result of printing onto Figure 16-10 shows a matte surface paper where the
an uncoated paper—the dye ink sits on the top surface on microporous coating sits directly on top of the paper

334
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Figure 16-9 A 200x magnified cross section of RC microporous Figure 16-11 A 200x magnified cross section of a Baryta microporous
coated paper. Photograph courtesy of www.graphicsatlas.org coated paper. Photograph courtesy of www.graphicsatlas.org

b*

Matte Paper
Luster Paper

a*
Figure 16-10 A 200x magnified cross section of matte microporous Figure 16-12 A comparison of measured color gamuts between luster
coated paper. Photograph courtesy of www.graphicsatlas.org and matte photo papers using an Epson 9900.

support. The pigment ink absorbs into the microporous the noticeable difference in color gamut when ink is
coating and possibly some of the paper support depend- absorbed into the coating compared to when it sits in the
ing on the coating quality. Higher-end papers allow for the microporous coating.
pigment ink to absorb to the bottom of the microporous
coating and for the top to harden allowing for a more
scuff-resistant print. Chromogenic Substrates
Figure 16-11 shows a substrate with a special chemi- Chromogenic substrates are light-sensitive papers that
cal coating typically comprised of barium sulfate (Baryta) require a photochemical development process to create a
under the porous ink-receiving layer. Such substrates are final image. Chromogenic paper consists of a gelatin emul-
commonly used by photographers because they yield an sion layer containing color dyes coated onto a cellulose
image that renders similarly to a photographic process acetate support seen in Figure 16-13 (overleaf). Exposing
without using darkroom chemistry. Figure 16-12 highlights the emulsion to light using lasers or LEDs activates the dye

335
 s ecti o n 3: O utp ut

Figure 16-13 A 200x magnified cross section of chromogenic paper. Figure 16-14 A 200x magnified cross section of a matte paper treated
Photograph courtesy of www.graphicsatlas.org for liquid toner. Photograph by www.graphicsatlas.org

layer. The paper is then developed using the RA-4 process Dry Toner Substrates
to create the print. Chromogenic papers used for digital Dry toner substrates are the easiest to find and can be pur-
image printing have a different chemical makeup com- chased in any office supply store by the ream for just a few
pared to those used in a traditional darkroom context. dollars. The paper does not require any special coatings or
treatment as the dry toner printing process fuses the toner
to the paper as seen in Figure 16-15. This fusing makes the
Liquid Toner Substrates final print more scratch-resistant compared to liquid toner
A special treatment is required for substrates used with prints. The substrates used are often standard color copier
liquid toner unless super calendering (rolling and compress- paper but higher-end options exist; pearlized paper, Kraft
ing the surface) of the substrate allows for the toner/plastic paper, transparency, vellums and Fine Art papers are availa-
layer to transfer from the blanket to the substrate. The ble for dry toner printing.
treatment also allows the toner to adhere to the substrate
to create a scratch resistance surface.
Substrates for the HP Indigo press, which use liquid
toners, are subject to a certification process specified by
HP. The Printing Applications Lab at Rochester Institute of
Technology is responsible for this certification process.3
Certified substrates are treated to allow the thin plastic film
to adhere to it. Therefore, the qualifications for the certifica-
tion are: runablity (does it feed through the press), blanket
wear (does the substrate destroy the blanket), adhesion Figure 16-15 A 200x magnified cross section of a matte paper with a
(does the toner adhere to the substrate and how well at fused image from dry toner. Photograph by www.graphicsatlas.org

different time intervals). Figure 16-14 shows that both sides

of the substrate are treated. The surface treatment that
accepts liquid toner is not the same as surface coatings
needed for aqueous inkjets.

336
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

100

90
Semi-Gloss
80

L * Printer driver media type

HW Semi-Gloss
70 Poster Semi-Gloss
60

50

40

30

20

10

0
0 20 40 60 80 100
*
L Target
Figure 16-16 A 200x magnified cross section of dye sublimation Figure 16-17 The measured and plotted data from a tone
paper. Photograph by www.graphicsatlas.org reproduction scale printed with different printer driver setting media
types.

Dye Sublimation Substrates tweaks to the available ink controls may be necessary if
Dye sublimation substrates are coated to allow for the over- or under-inking is observed.
encapsulation of the dye when it’s in its sublimation state. Other key substrate characteristics include paper
Figure 16-16 shows the dye image sitting on top of the weight, thickness, surface sheen, texture and base tint or
substrate surface—dye doesn’t get absorbed into the white point. Manufacturers often offer samples to exam-
paper support. Both sides of the paper have a polyethyl- ine and assist in selecting a photographic paper and we
ene layer. The back of the paper features a back coating recommend seeking out these paper sample books when
while the front features an insulating layer below the top planning photographic output.
coatings. The insulating layer is critical during the printing
process when heat is applied for each dot of transferred
dye. The amount of heat determines how much dye is Paper Weight and Thickness
transferred. Paper stock is characterized by the base substrate thickness
in accordance with ISO 18055-1:2004 Photography and
Imaging – Inkjet Media: Classification, Nomenclature and
Photo Paper Characteristics Dimensions – Part 1: Photo-Grade Media (Paper and Film).
One way to measure thickness is using the mil, a unit of
We consider a variety of paper characteristics beyond the length where 1 mil is equal to one-thousandth of an inch
basic swellable and porous categories. Importantly, paper (0.001 inch). A photo paper may be approximately 10 mil
types dictate how much ink is appropriate to lay down thick, for example, while a U.S. dollar bill is 4.3 mil.4 Paper
(controlled by printer ink settings). In Figure 16-17, the weights are described by basis weight in grams per square
same tonal reproduction scale is printed with different meter (g/m2). Paper used for documents and text is typically
media types chosen from the printer driver software and around 80 g/m2; photo papers range in weight from 190 to
the resulting L* values are measured. The difference in 270 g/m2.
media type additionally impacts color gamut. Some media
type settings offer subtle differences and making small

337
 s ecti o n 3: O utp ut

Table 16-1 Surface sheen categories and their appearance. Surface Sheen
Category Appearance
A substrate’s ability to reflect light defines its surface sheen.
The categories of sheen are specific to traditional printing
Coated Papers Gloss High sheen surface
processes for coated or uncoated printing papers. Coated
Satin Less shiny than gloss
papers are categorized as gloss, satin or matte. Uncoated
Matte Flat, little sheen papers exhibit surface sheen categories of smooth or
Uncoated Papers Smooth/Woven Very smooth surface woven, laid and linen. Table 16-1 identifies the main
Laid Textured lines categories of surface sheens and their appearances. Inkjet
Linen Fine lines substrates borrow the gloss, satin and matte descriptors
despite differences in material composition. Figure 16-18
shows two photo papers photographed using a high
magnification scanning electron microscope to reveal their
topographic differences.

Surface Texture
Texture is the feel of the paper surface. Some photo papers
have the look and feel of traditional light-sensitive photo
paper where others offer an appearance of a paint canvas
or watercolor paper. While many photographic prints are
not intended to be handled by viewers, texture is often
visually inferred based on the play of light and shadows
across the paper surface. On the other hand, texture and
paper weight are key elements of the photo book experi-
ence as the reader holds and turns the pages.

Brightness and Optical Brightening Agents

Paper is not inherently white and its color varies depending
on the region of the world from which its raw materials are
purchased. To produce a paper with a pure white, papers
are often tinted with chemicals like titanium dioxide.
Papers may also be dyed to create a warmer paper base
like an off-white or cream color.
It’s common to incorporate some amount of optical
brightening agents (OBAs) in paper surfaces with the goal
Figure 16-18 Highly magnified scanning electron microscope images of intensifying the paper’s apparent brightness. These
of photo paper substrates: a semi-gloss (top) and a matte (bottom).
Photographs by Rochester Institute of Technology photography
colorless chemical compounds are similar to laundry
alumnae Lia DiRico detergent brighteners that promise to keep white fab-
rics looking new. The effect is produced by using a dye
or chemical substance that fluoresces upon exposure to
ultraviolet energy (present to some extent in most light
sources). That is, the chemical component absorbs incident

338
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

110
is cut that may cause production issues in large-scale
100 printing operations and their presence dulls paper cutting
blades faster.
90
Spectral Reflectance

80

70 Paper White Point

The color of a paper’s base tint is called the paper white or
60 Matte Paper with OBA paper white point. The paper white point is identified in a
Gloss Paper with OBA
50 Matte Paper without OBA paper’s ICC profile. ICC profiles and the color management
module software use its color characteristics to correctly
40
render the white point in an image and to adjust all colors
30 accordingly for the most accurate reproduction when soft
350 400 450 500 550 600 650 700 750
Wavelength (nm) proofing and printing. The paper white point is typically
Figure 16-19 Measuring the spectral reflectance of different photo provided using CIELAB values. Newspaper is a notable
papers reveals the impact of OBAs. A 100% reflectance line is
example of a paper substrate with a paper white that is far
included for reference.
from bright or pure white.

UV energy and re-emits it as visible light energy, primarily Understanding Printer Settings
in the blue range. This causes the spectral reflectance of
papers with OBAs to exceed 100% as measured in Figure Navigating software options presented to the user
16-19. The presence of brightening agents results in paper moments before sending a file to print is daunting even
looking brighter than materials nearby in its environment. with a solid understanding of the materials and technolo-
OBAs are not archival and fluoresce less over time due to gies covered thus far. Here, we describe some of the major
airborne pollutants or simply exposure to oxygen or light. options available when configuring the software to output
The lessening chemical reaction reveals the natural, likely to your paper of choice. These include selecting between
yellowish substrate base color.5 On the other hand, framing photo or matte black inks, printhead height, bit depth,
a print behind archival, UV-blocking glass prevents the manufacturer-specific options like Super MicroWeave,
fluorescence reaction altogether and negates the utility of high-speed printing, output resolution and proprietary
any present OBAs. printer drivers.

Mineral Whiteners Photo Black and Matte Black Inks

Paper companies alternatively or additionally use The popular photo printer manufacturers, Epson, HP and
chemical additives such as calcium carbonate to main- Canon, use two distinct blacks in their printers: Photo Black
tain consistency in substrate whiteness. This degree of (PK) and Matte Black (MK). The difference lies in the type of
substrate color control prevents the need to create a new paper each of the inks is formulated to work with. Photo
ICC profile for each batch of paper. These mineral whiten- Black (or non-absorbent black) is meant for papers where
ers are more stable than the optical brightening agents the ink sits on top of the microporous coating and doesn’t
described earlier and do not necessarily modify the spec- penetrate into the paper fiber. Resin-coated and Bary-
tral reflectance of the substrate. The Canson® Infinity line ta-type papers should receive Photo Black ink. Matte Black
of archival photo papers uses mineral whiteners and is (or absorbent black) is formulated for papers in which the
OBA-free. Mineral whiteners create a dust when the paper ink sits within the microporous coating. It’s also intended to

339
 s ecti o n 3: O utp ut

100
importance of controlling the temperature and humidity
90
Gloss Paper with Photo Black in a printing environment. Head strikes indicate a need to
Gloss Paper with Matte Black
80 adjust the printhead height setting to prevent damage to
70
the printheads and the substrate. If the printhead it too far
away, on the other hand, the inkjet spray exhibits over-mist-
60
L Sample

ing which results in a much softer printed edge.

50
*

40

30 Printing Bit Depth

20
Bit depth settings control how much tonal data is sent to
the printer. The printer uses the image data to create the
10
maximum tonal range for the selected substrate. Current
0
0 10 20 30 40 50 60 70 80 90 100 printers are really only capable of printing upwards of
L* Target 12-bits worth of image data when working on a luster or
Figure 16-20 A comparison of tone reproduction when printing on glossy photo paper.
Hahnemühle Photo Rag Baryta using MK and PK inks.

Super MicroWeave and Finest Detail

get somewhat absorbed into the paper fibers. Matte Black The Super MicroWeave and Finest Detail settings offered by
should be used with cotton and matte papers. Epson printers dictate how the dots of inks are laid down
Each of the ink formulations provides the maximum on the substrate using screening and variable dot size
black when used on the correct paper and helps to avoid controls. The effect is more easily observed on a luster or
printing defects like bronzing. Accidentally printing with glossy paper surface than on a matte surface. Figure 16-21
Matte Black on a paper that is non-absorbent means that shows the Finest Detail setting turned on and off when
the black ink rubs off with the slide of a finger across its printing with Super MicroWeave enabled, highlighting the
surface. Figure 16-20 demonstrates how using PK ink on difference in how the ink dots are laid down.
a matte paper drastically changes the black density and
therefore the overall tonal reproduction. Many paper pro-
files pick the proper ink automatically to avoid accidental High-Speed Printing
use of the incorrect black. The high-speed printing option dictates whether the
printer uses unidirectional or bidirectional printing. Think
about eating corn on the cob: do you eat it in both
Printhead Height directions or in one direction and return back to the other
The printhead height setting controls the height of the side to start again like a typewriter? The two approaches
printhead in relation to the substrate. This can be important demonstrate unidirectional and bidirectional action. The
to adjust depending on the thickness of the paper and envi- latter method gets more accomplished with a similar range
ronmental conditions. The thicker the substrate, the higher of movement as the former. Bidirectional printing trans-
the printhead height must be set. Changing the printhead lates to high-speed printing because the printhead lays
height also helps prevent head strikes where the printhead down ink in both directions as it moves back and forth over
comes in contact with the substrate and creates ink buildup the paper. Some photographers swear that bidirectional
on the edges of the paper or scratched lines in the print printing creates image softness if the head is not laying the
area. Head strikes are also introduced by high humidity in ink down in the proper spot on the return passage of the
which the edges of the substrate curl, emphasizing the printhead. That is a possibility, however, a software utility

340
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Figure 16-22 High magnification views of prints made at 1440 dpi

(top) and at 2880 dpi (bottom) on Epson Premium Luster photo paper.
Photographs by Eric Kunsman

Output Resolution
Figure 16-21 High magnification views of prints made with an The most common output resolutions for photographic
Epson 4880 at 1440 dpi and using Super MicroWeave on Ilford
purposes are 1440 or 2880 dpi. Printing at higher output
Gold Fiber Silk. The test target is printed fine detail turned off
(top) and with fine detail turned on (bottom). Photographs by Eric resolutions slows the print speed down. The print reso-
Kunsman lutions offered by an example Epson printer are 360, 720,
1440 and 2880 dpi. For viewing distances greater than
10–15 feet, print output resolutions as low as 360 or 720
exists to ensure proper alignment. Running a printhead dpi yield acceptable quality.
alignment when set up for bidirectional printing negates Some believe that photographers must use 2880 dpi
this potential print quality problem. High-speed printing for the finest prints. In the end, think about the substrate
can reduce total print time (which is considerable for large and how well it holds the dot definition. There is little to
print sizes). no difference in detail between 1440 and 2880 dpi when

341
 s ecti o n 3: O utp ut

working with matte fine art papers, for example. Working considered as a function of showing the image, not as an
with a luster RC paper, on the other hand, and you may exercise in print characteristics.
distinguish the difference. Figure 16-22 shows the differ­ The traditional calculation for viewing distance mul-
ence in print resolution settings on a Luster RC paper tiplies the diagonal of the print by 1.5. For example, a
under magnification. Most users only see a difference 17x22 inch print with a diagonal of 28 inches equates to
under a loupe or magnification and therefore use the a viewing distance of 42 inches. This is a general rule of
1440 dpi resolution setting most of the time. Photogra- thumb as many factors impact viewing distance including
phers should test the available settings using their paper type (texture and gloss level), print resolution and a
selected substrate and printer combinations to establish a photograph’s contrast or dynamic range. The higher the
preferred resolution. pixels per inch and dots per inch, the closer one can view
the image and discern detail.
In the early 2000s, Nikon introduced a point-and-shoot
Printer Drivers digital camera called the Nikon 950. They printed a pho-
Canon, Epson and HP provide printer drivers to best config- tograph taken with the camera onto a large billboard on
ure the printer hardware for optimal print media output. the side of a building several stories tall about two blocks
These are custom software commands that interface from the Photo Plus Expo in New York City. People walking
with the printer hardware. It’s worth the time to test and out of the trade show buzzed about buying that camera
compare the different controls offered by the printer because it created photographs capable of being printed
driver to understand how they alter the image quality on a so large. They failed to consider the substantial viewing
given substrate. For instance, choosing your substrate type distance as a primary contributor in making the image look
effectively sets an ink limitation and ink conversion table so good at that size.
specific to that paper control. Therefore, when building a We use this anecdote to consider another factor related
color managed workflow, make sure that you use the same to viewing distance: the pixels per inch required for the
paper controls that were used to build the ICC profile at viewing distance as a function of the human visual system’s
the start. If you use generic paper profiles downloaded resolving power. Equation 16.2 is used for a high contrast
from the paper manufacturer’s website, use the settings print with optimal lighting conditions.
they used to create that profile to maximize the profile’s
potential. Ultimately, the printer driver is a black box with Pixels Per Inch = 1 (Eq. 16.2)
(Distance x 0.00029)
predefined controls that we try to adjust for a specific sub- 2

strate. Once you determine the options that you feel create
the best image, stick with them. Alternatively, Equation 16.3 rewrites this equation to solve
for distance:

Print Size and Viewing Distance Distance = 2 (Eq. 16.3)

Pixels per Inch x 0.00029
Always consider viewing distance when talking about a
print’s pixels per inch and dots per inch characteristics. At These equations are based off of studies conducted by
what distance do you expect viewers to experience the Ed Grainger in the 1970s regarding the human bandpass
print? The answer varies wildly depending on the venue, characteristic whereby we see certain spatial frequencies
presentation and photographic subject matter. There’s better than others for subjective print sharpness. Table 16-2
always an individual at an exhibition that looks at a 24x36 provides guidelines for a range of viewing distances based
inch print from half an inch away, judging the character- on Equation 16.3.
istics of the print. Prints on display should be viewed and

342
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Table 16-2 Viewing distance guidelines based on pixel resolution. ink channel along with ink drop size, dot structure and
screening patterns. Higher tiers of RIP software typically
Viewing Distance (inches) Resolution (pixels per inch)
correspond with higher price tags.
6 1,145

12 573

18 382 Hard Proofing

36 191

48 143 We’ve talked a lot about creating output profiles and this
is the stage where they get to flex their color-managed
60 115
muscles. An image must be interpreted by the output
96 72
profile before sending the information to the printer. This
120 57
profile, remember, is a characterization of the specific
printer and paper combination you’re using to print the
image. Sometimes this step of converting the photo to the
output profile is not obvious: Adobe Photoshop waits until
Raster Image Processing you’re at the “Print” screen to ask about the output profile
to apply, for example.
Raster image processing transforms the image data into A hard proof, also called a proof print or match print, is a
a language that the printer speaks, compatible with its printed simulation of your final intended output in order to
method of printing (ink drops, laser, etc.). The raster image review image processing, print settings and any other rele-
processor (RIP), a software solution, takes the image file and vant details. A hard proof ensures that the final product is the
translates it to the various controls offered by the printing highest quality physical media output of your photographic
hardware: resolution, ink colors, screening pattern and color work. Soft proofing always faces an uphill battle by virtue
profiles are all factored into the preparation. RIP functionality of relying on additive color mixing (via display) when the
comes in the form of printer firmware or desktop software. final output uses subtractive color mixing (via colorants and
The act of “RIPing” means passing the image file through this substrates). Options to simulate paper color or paper texture
layer of interpretation as it heads to the printer hardware. are inherently less accurate than making actual print. Soft
A printer driver provides a starting point with controls proofing is a valuable precursor to making a test print (or
predetermined by the manufacturer. A RIP, depending two, or ten) while keeping an eye on the prize: optimal color
on the level of control desired, may offer finer tuning and tone reproduction of your images. As each photograph
and a greater level of low-level settings to maximize your is unique and circumstances regarding print size, file size and
substrate and printer combinations’ output quality. Pho- tone reproduction are variable, hard proofing is a critical part
tographers opt to work with a RIP when the manufacturer’s of the output process to get everything optimized.
printer driver doesn’t offer the control they need or when Hard proofs are different for commercial use and fine
working in a lab environment where queue management art use. A commercial hard proof is often made using
and auto-layout to maximize print area are used. inkjet printers with an aqueous inkjet paper that emulates
RIPs can be organized into three tiers with each level the intended output substrate for traditional printing
adding additional controls and capabilities. A basic RIP processes. A hard proof is not created on a printing press
provides minimal control of ink amounts and drop layout. due to the make-ready process of RIPing the job; creating
Mid-level RIPs provide custom ICC profile creation tools printing plates and mounting the plates on the press is
and the ability to linearize the printer. Mid-level RIPs pro- prohibitively expensive. If the final output is on a digital
vide the ability to set ink limitations, and create custom ICC press, hard proofs are often made using the same printer
profiles. High-end RIPs additionally offer control of each hardware and with the substrate used in production. These

343
 s ecti o n 3: O utp ut

are used as contract proofs which act as an agreement

between the printing house and the client regarding how
the production should ultimately look.
In a fine art print studio, the printer and substrate com-
bination are used to hard proof and all final adjustments
are judged on the exact substrate. This is critical for artists
and photographers that work with open editions because
the print studio makes copies as demand for the print
arises. The fine art hard proof is used to ensure that the
edition always matches from start to finish; printing tech-
nology may change if the edition is open for a long period
of time and this practice facilitates consistency. Figure 16-24 A visible gloss differential. Photograph courtesy of the
Image Permanence Institute

Printing Issues

There are three common issues that occur in physical

media output: abrasion, bronzing and gloss differential.
Let’s step through each one.
Bronzing describes when the ink appears bronze in color
when viewed at an angle. Figure 16-23 shows bronzing
due to printing with a matte black ink on a glossy photo
paper. Bronzing occurs when certain light wavelengths are
trapped under the ink that sits on the microporous coating
and reflect at a different angle. This creates more of a
metallic look relative to the ink sitting on the substrate sur-
face. Additionally, some printer companies are formulating

Figure 16-25 Minor abrasion on a print surface. Photograph courtesy

of the Image Permanence Institute

their new inks to sit on the paper surface which can also
inadvertently introduce the characteristic.
Gloss differential is an unwanted appearance that occurs
when the printed ink changes the gloss level of the paper
Figure 16-23 An example of bronzing introduced by printing with a
matte black ink on a glossy photo paper. Image courtesy of the Image
surface. It’s a consequence of the pigment’s gloss level
Permanence Institute differing from that of the substrate. If the pigment sits on

344
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Figure 16-26 An example test print created based on the recommendations found in
the textbook Color Management & Quality Output.6

top of the microporous coating, the different gloss levels Evaluating Paper and Print Quality
are visible as unequal reflectivity as seen in Figure 16-24.
Printer companies combat this problem by applying a clear One question that we frequently hear from photography
coat ink (called a gloss optimizer or chrome enhancer) after students is: what’s the best paper for printing? There is no
all of the color inks are applied. The gloss optimizer creates simple or easy answer as there are many factors to consider
a uniform surface off of which light reflects. including personal preference and artistic intent. A classic
Abrasion occurs when ink sits on top of the microporous paper study helps answer the question on a technical,
coating or even within the coating. The surface is damaged physical media level, which offers an insightful starting
by physical contact and the colorant is scratched or com- point. We first run a customized test print like the one
pletely removed. Figure 16-25 shows an example of minor shown in Figure 16-26 for every paper and printer combi-
abrasion to the surface of a printed image. Abrasion comes nation we use or intend to evaluate for use in our practice.
from head strikes or a lack of careful handling and storage The test print provides us with everything necessary
of the final print. to objectively and subjectively evaluate image quality as

345
 s ecti o n 3: O utp ut

Figure 16-27 The presence of OBAs is easily detected using a standard UV blacklight.
The photo paper on the left fluoresces while the one on the right does not.

a function of paper characteristics. It includes high key helpful to know if long-term durability and print quality are
and low key images, skin tones, black and white images, priorities and if the product description is unclear on their
challenging saturated colors and memory colors like blue use. Figure 16-27 shows our low-tech test for identifying
sky and green grass. The test print layout also includes photo papers with OBAs.
ColorChecker patches, grayscale and a full color gradients.
We suggest building a test print document in Adobe Pho-
toshop and replacing or revising elements as your printing Measuring Color Gamut and White Point
skills and needs evolve. Next, we measure CIELAB values for the red, green, blue,
cyan, magenta and yellow patches using a spectropho-
tometer to create a gamut plot. The measured CIELAB
Testing for OBAs values for a white patch indicate the white point or
We test for the presence of optical brightener agents by neutrality of the paper or it identifies a notable base tint.
placing the paper under a UV light source. A paper with A perfect white in CIELAB space is indicated by a* and
chemical brighteners fluoresces or glows brightly under a b* values of 0. Knowing the paper’s white point assists in
UV source such as a blacklight. We find that manufacturers’ determining if a print on the same photo paper will be
claims here are not always reliable. Printing with papers neutrally color balanced.
containing OBAs is not necessarily problematic but it is a

346
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Measuring Black Point Density and Gloss paper due to the first surface reflections of the substrate.
Following gamut plots and white point measurements, we Prints are typically viewed on the perpendicular with light-
use a densitometer and the grayscale patches on our test ing off-axis. Most of the incident light that strikes a high
print to determine the black point density and to evaluate gloss paper with a smooth top surface bounces off at the
the dynamic range of the paper and printer combination. same angle as it approaches with very little light reflecting
The black point density is largely driven by the paper to the viewer’s eyes (see Figure 16-28). This results in a
surface. A glossy paper yields a darker black than a matte deeper, darker black and a higher dynamic range because
areas of high ink or pigment density appear dark to the
viewer. A textured paper surface like the kind found on
a matte paper causes the incident light to reflect back
at many different angles. Some of that light reaches the
viewer’s eyes as flare and some of it scatters, resulting in a
perceived low black point.
Photo paper manufacturers typically label their papers
as gloss, luster (or satin) and matte. The definitions of
these categories are not dictated by the ISO and are at the
Figure 16-28 The interaction of incident light on paper surfaces
discretion of the manufacturer, so going purely off of mar-
changes the perceived density. keting terms could mean that one photographer’s satin is

Figure 16-29 A compilation of print quality measurements designed to reveal key characteristics of our substrates using the same
inkjet printer. Clockwise from top left is a tone reproduction plot, a gamut plot, a table listing Dmax, white point chroma and gloss
readings and a paper white point plot.

347
 s ecti o n 3: O utp ut

another’s glossy. In light of this, the Technical Association of professional workflow involving hard proofing means
Pulp and Paper Industry’s T-480 offers a published standard making derivative image files that tweak and revise edits
for measuring gloss for papers, both coated and uncoated.7 when the test prints are subpar. If you’ve followed the text
A gloss meter is used to classify a paper surface according thus far, you likely have an appreciation for holding onto
to its reflectance at specific angles. It measures the specular detail and image information from the beginning of the
reflections on a surface at a range of angles and reports its image-making process—don’t leave anything on the table
measurements in gloss units (GU). at this last output stage.
The gloss meter offers us the chance to create a Simultaneous contrast, defined in Chapter 13, is impor-
uniform gloss name categorization across our most-used tant to keep in mind when hard proofing and when
papers. It takes measurements at 20o, 60o and 85o and considering display environments for your work. The
returns gloss unit values ranging from 0–100. Semi-gloss background or surrounding area in front of which a print is
surfaces are measured at 60° and should fall between viewed changes the perception of that print. The phenom-
10–70 GU. Matte surfaces measuring below 10 GU at 60° enon can change your perception of both color (hue and
should be measured at 85°; highly reflective surfaces with saturation) and lightness/density.
measured values exceeding 70 GU at 60° should be meas- The best option when doing critical print evaluation is to
ured at 20°. use a designated light booth (sometimes called a viewing
Figure 16-29 shows the compiled results from conduct- station) with neutral surround and controlled illumination.
ing some of the tests and measurements described here These offer the best chance to evaluate the print without
on copier paper, a luster photo paper and a matte photo outside factors introducing bias or misleading representa-
paper. It borrows concepts and practices described in tions. A calibrated viewing station is shown in Figure 16-30.
Chapter 13 and Chapter 14, using CIELAB measurements The best practice is to evaluate prints under the illumi-
and chroma calculations. From this data we begin to draw nation type that will be used to illuminate them when
conclusions and form an understanding of the photo viewed by others. As a consequence, viewing stations are
papers we regularly work with: their potential tonal range, equipped with multiple standardized illuminants like D65.
white point neutrality, color gamut and gloss character- Newer models include UV sources for detecting the pres-
istics. While we wouldn’t recommend seeking to create ence and effect of OBAs.
high-quality photographic reproduction on cheap copier
paper, we included it in our investigation as a point of
reference—it emphasizes the power of the substrate in
addition to the capabilities of the printer.

Critically Evaluating Color and Contrast

Additional characteristics such as contrast, color casts and
accurate color reproduction are important to evaluate
with the test print. In an image with skin tones, getting the
proper color balance is key. Look for artifacts like poster-
ization or banding in smooth tones and gradual tone or
color transitions. Keep a critical eye on how dark and how
bright the high and low key details render; maintaining
image pixel information in these extremes is not equiv-
alent depending on the paper type, color management Figure 16-30 Evaluating a test print using a calibrated viewing station.
decisions or editing decisions made prior to printing. A Some viewing stations include multiple light sources including UV.

348
 c h a p t e r 16 : P r i n t i n g a n d P h y s i c a l M e d i a

Finally, there is more to print quality than just objective Notes

measurements. There are many subjective qualities that 1 “Real world print production.” Real World Print Production, by
make a great print. We have touched on many of them Claudia McCue, Peachpit Press, 2009, p. 31.
throughout this text. The dynamic range of a paper may 2 “Global UV cured printing inks market (2018–2023): The
be considerable, for example, yet a poor print or poorly market is expected to grow at a CAGR of over 10% –
ResearchAndMarkets.com.” Business Wire, Sept. 7, 2018,
edited file can render everything too dark or muddy and
www.businesswire.com/news/home/20180907005221/en/
low contrast. Texture is another area that must be seen and
Global-UV-Cured-Printing-Inks-Market-2018-2023.
experienced to appreciate if it enhances or distracts from
3 “Printing applications laboratory.”, www.rit.edu/gis/printlab/.
the printed content.
4 “United States one-dollar bill.” Wikipedia, Wikimedia
Our favorite resource that goes into more detail with Foundation, Feb. 13, 2019, en.wikipedia.org/wiki/
excellent exercises and examples is Color Management & United_States_one-dollar_bill.
Quality Output: Working with Color from Camera to Display to 5 Work, Roy. “To brighten or not to brighten.” LexJet Blog, July
Print by Tom Ashe, published by Focal Press in 2014. 2000, https://blog.lexjet.com/wp-content/uploads/2009/10/
sbjuly43-45.pdf.
6 Ashe, Tom. Color Management & Quality Output: Working
with Color from Camera to Display to Print. Focal Press, 2014,
p. 51.
7 “Specular gloss of paper and paperboard at 75 degrees,
test method TAPPI/ANSI T 480 Om-15.” Thickness (Caliper of
Paper, Paperboard, and Combined Board, Test Method TAPPI/
ANSI T 411 Om-15, imisrise.tappi.org/TAPPI/Products/01/
T/0104T480.aspx.

349
Index

3:2 Pulldown, 153 Amount, unsharp mask filter, 216–217 B

Amplitude Modulation (AM), B, see Byte
A 325–326 B-frame, see Bidirectional frame
A/D, see Analog to Digital Analog to Digital (A/D), 73, 75–76, 83, 96, Back nodal point, 37–44
AAC, see Advanced Audio Coding 114–115, 231–236 Back principal focal point, 38, 50
Abrasion, 332, 334–334 Anomalous trichromats, 271 Back-focusing, 89, 245
Absorbed (light), 7, 27–28, 127–129, 244, ANSI, see American National Standards Backside Illuminated (BSI), 78
266, 268, 271, 273 Institute Bandwidth (data), 149–151, 153, 169, 172,
Absorbed, Reflected, Transmitted, Anti-aliasing filter, 83, 163, 215 185–187, 263, 312–316, 320
Scattered (ARTS), 127 AOMedia Video 1 (AV1), 150 Bandwidth (light), 5
ACE, see Adobe Color Engine Aperture, 84–85, 102–105 Baryta, 328, 335, 340
Active autofocus, 89 Aperture diaphragm, 84–85 Bayer pattern, 74–76, 157
Active Pixel Sensor (APS), 76, 79 Aperture Priority, 117 BD, see Blu-ray Disc
Acutance, 214 APS, see Active Pixel Sensor Beam intensity, 10
Adaptive bitrate, 151 Aqueous inkjet printers, 331–332 Bicubic Interpolation, 165–166
Additive color mixing, 271–272, 288, 343 AR, see Augmented Reality BID, see Binary Ink Developer
Additive primary colors, 271, 320 Archiving, 158, 175, 185 Bidirectional frame, 147
Administrative metadata, 179 ARTS, see Absorbed, Reflected, Bidirectional Reflectance Distribution
Adobe Color Engine (ACE), 296 Transmitted, Scattered Functions (BRDF), 297
Adobe RGB, 315 AS-IPS, see Advanced Super In-Plane Bilinear interpolation, 165–166
Advanced Audio Coding (AAC), 150 Switching Binary Ink Developer (BID), 330
Advanced Photo System, 79 Aspect ratio (camera), 79–80, 95, Bioluminescence, 9
Advanced Super In-Plane Switching (AS- 205 Bit, 81–83, 107
IPS), 310 Aspect ratio (display), 312–313 Bit-Plane encoding, 253–255
Advanced Video Coding (AVC), 149–150, Aspect ratio (video), 143–145 Bitmap, 168
152 Augmented Reality (AR), 153, 320 Bitmap Image File (BMP), 169, 171–172
AE, see Automatic Exposure Autofocus points, 89 Bitrate, 150–152
AEB, see Automatic Exposure Bracketing Automatic Exposure (AE), 85, 88, Bits per channel, 83, 171–172, 252
Aerial haze, 46, 204–205, 282, 288 116–118, 230 Bits per pixel, 83
Aerial haze adjustment, see Dehaze Automatic Exposure Bracketing (AEB), Bitstream, 148–150
AF, see Automatic Focus 119 Black level subtraction, 232
Afocal attachments, 45 Automatic Focus (AF), 88, 230, 245 Blackbody, 7, 10, 16, 18, 20, 25, 280
Afterimage, 138 -139 Automatic White Balance (AWB), 117, Blackbody curve, 17–18
Aliasing, 83–84, 163, 215, 239 121–122, 230, 244 Blackbody radiation, 2, 6–7, 10
Alpha channel, 223–224 AV1, see AOMedia Video 1 Blend modes, 219–220
AM, see Amplitude Modulation AVC, see Advanced Video Coding Bloch’s law, 138
American National Standards Institute AWB, see Automatic White Balance Blocking artifacts, 151–152, 262
(ANSI), 28, 349 Blooming, 108–109

351
 i n d ex

Blu-ray Disc (BD), 182 CF, see Compact Flash CMYK, see Cyan, Magenta, Yellow, Key
Blur, 65–66, 83–84, 90–91, 103–105 CFA, see Color Filter Array CMYK Profiles, 297
Blur Filter, 218–219 CFL, see Compact Fluorescent Lamps Codec, 145–146, 148–150, 233
BMP, see Bitmap Image File Channels, 82, 171–173, 180–181, Coded picture frame type, 147
BRDF, see Bidirectional Reflectance 208–210, 235 Coding redundancy, 252, 255–257
Distribution Functions Characteristic curves, 195 Coding tree, 149
Brewster angle, 23–24 Characterization, 292, 295, 343 Cold body radiation, 9
Brightness (compression), 250–252 Charge-Coupled Device (CCD), 75–76, Cold Cathode Fluorescent Lamp (CCFL),
Brightness (defintion), 14 78, 196, 233 309
Brightness (digital), 73–76, 168, 173–174, Chemiluminescence, 9 Color balance, 18–20, 178, 196, 203,
221–224, 233 Chroma (C*), 277–281, 303 206–208
Brightness (display), 92, 288–290, Chroma channel, 251, 258, 260–261 Color bleeding, 151–152, 263
299–300, 308–310, 314–318 Chroma subsampling, 148, 152, 258–259, Color channel histogram, 109
Brightness (histogram), 109–114, 190, 263 Color constancy, 5, 8, 244
196 Chromatic adaptation, see Color Color contrast, 283–284
Brightness (image formation), 64, 104, Constancy Color correction, 119, 180, 205–206,
108, 118–119, 128, 160–161 Chromaticity coordinates, 280–281, 293, 303–304
Brightness (light source), 17, 19, 28, 30 321 Color Correction Matrix (CCM), 161
Brightness (scene), 102, 107–108 Chromaticity diagram, 161–162, Color Filter Array (CFA), 74–76, 159, 160,
Brightness (visual), 5, 82–83, 138, 282 280–281, 293–294, 316, 321 169
Bronzing, 340, 344 Chromogenic print, 328 Color filters, 127–130, 221
BSI, see Backside Illuminated Chromogenic substrates, 335–336 Color gamut, 162, 280–282, 292–295,
Bubble jet system, 327 CIE, see International Commission on 299–302
Byte (B), 180–181, 184–185, 254 Illumination Color interpolation, 160
CIE Illuminant A, 5, 17, 274–275 Color management, 291–296
C CIE Standard Illuminant Daylight 5500K Color Management Module (CMM), 296,
Calibration, 245–246, 292, 295, 298, 307, (D50), 6, 25 305
318–319 CIE Standard Illuminant Daylight 6500K Color Management System (CMS),
Camera exposure, 102–104, 115, 118, 128 (D65), 5–6, 17, 25, 274–275, 319, 348 295–296
Camera movement, 91, 140 CIE Standard Illuminant Fluorescent (F2), Color Matching Module (CMM), see Color
Camera-based image stabilization, 90–91 17 Management Module
Candela, 10–16, 19, 314 CIE System of Color Specification, 276, Color non-uniformity, 78, 240
Candlepower, 10–11 279–280 Color profile, 158, 292, 295, 319
Canned profiles, 297, 342 CIELAB, 22, 210, 233, 243–244, 281–282, Color Quality Scale (CQS), 21–22
Capture sharpening, 163, 215 293–296, 299, 339 Color Rendering Index (CRI), 20–22, 31,
Catadioptric lenses, 60 CIEXYZ tristimulus values, 280–283, 296, 32
Cathode Ray Tube (CRT), 308 299–300, 319–321 Color saturation, 276
CCD, see Charge-Coupled Device Clarity, 204 Color shading, 77, 240
CCFL, see Cold Cathode Fluorescent Clipping, 110, 116–119, 197–199, 204, Color shading, 77–78, 240–241
Lamp 300–301 Color space, 22, 161–162, 282, 293–296,
CCM, see Color Correction Matrix Cloud-based storage, 185 300
CCT, see Correlated Color Temperature CMM, see Color Management Module or Color space mapping, 161
CD, see Compact Disc Color Matching Module Color Temperature (CT), 2, 16–21, 25–32,
Center of curvature, 37 CMOS, see Complementary Metal Oxide 119, 121, 319, 322
Center-weighted metering, 116 Semiconductor Color vision deficiencies, 270
Central Processing Unit (CPU), 188–189 CMS, see Color Management System Color vision theories, 268

352
 i n d ex

Colorant, 331, 333–335 D55, see CIE Illuminant D55 Diffraction limit, 64
ColorChecker, 115, 205–206, 233, D65, see CIE Illuminant D65 Diffuse, 22–23, 28, 94, 218
294–295 D65, see CIE Standard Illuminant Daylight Digital Asset Management (DAM), 156,
ColorChecker Passport, 206 6500K 167–168, 173, 265
Combing, 198 DAM, see Digital Asset Management Digital count, 73, 110, 168, 196, 199, 233,
Commission Internationale de l’Eclairage, Dark pixel defects, 74 236, 251, 300, 319
see International Commission on Dark Signal Non-Uniformity (DSNU), Digital Light Processing (DLP), 272
Illumination 235–236 Digital Negative (DNG), 158, 169, 170,
Compact camera, 78, 81, 96, 98 Data compression, 141, 160, 168, 251 189–190, 253
Compact Disc (CD), 98, 182 Data interpolation, 158 Digital Single Lens Reflex (DSLR), 80, 94,
CompactFlash (CF), 98–100 Daylight, 5–6, 17–18, 20–21, 25–29, 96, 98, 205, 232, 247
Compact Fluorescent Lamp (CFL), 28 119–121, 129, 274, 280, 304 Digital Versatile Disc (DVD), 182
Complement mapping, 201 DCT, see Discrete Cosine Transform Digital video resolution, 143–144, 151
Complementary Metal Oxide Dead pixel, 232, 251 Digital Visual Interface (DVI), 186, 187,
Semiconductor (CMOS), 76–78, Decamired, 19 312
87–88, 100, 132–133, 233 Decibels (db), 237 Discontinuous, 8, 18, 24
Compound lenses, 41, 97 Decoder, 147, 148, 151–152 Discrete Cosine Transform (DCT), 251,
Compression artifacts, 148, 191, 261–262 Dehaze, 204–205 259–260
Computational photography, 97 Delta E 2000 (ΔE00), 244, 302 Discrete Wavelet Transform (DWT), 260
Concave lens, 37, 60 Delta Eab (ΔEab), 243, 302 Dispersion, 5–6, 36, 62
Continuous servo, 90 Delta frame, 147 Display device profiles, 299
Continuous source, 6 Demosaicing, 74–75, 83, 113, 160–165, Display dynamic range, 314–316,
Continuous tone printing, 328 215, 230–234 318–319
Continuous-flow inkjet, 327–328 Density, 126–127, 130, 238, 298, 347 DisplayPort, 186–187, 312, 314
Contrast adjustment, 198, 204 Depth of Field (DOF), 33, 46, 53–60, 105, Dithering, 326
Contrast detection autofocus, 89 114, 117–119, 288 DLP, see Digital Light Processing
Convex lens, 37 Depth of Field and F-Number, 56–57 DNG, see Digital Negative
Corpuscular theory, 2, 34 Depth of Field and Perspective, 46 DOD, see Drop-on-Demand
Correlated Color Temperature (CCT), 16, Depth of Field determination, 53–55 DOF, see Depth of Field
20–21, 27, 29 Depth of Field Preview, 85, 92 Donor ribbon, 330
Covering power, 44–45, 61 Depth of Focus, 57–58 Dots Per Inch (DPI), 324–328, 341–343
CPU, see Central Processing Unit Depth perception, 268 DPI, see Dots Per Inch
CQS, see Color Quality Scale Descriptive metadata, 179–180 Drop-on-Demand (DOD), 327
Creative sharpening, 215–217 Destination profile, 296, 305–306 Dry toner, 326, 329–330–336
CRI, see Color Rendering Index Device Independent Bitmap (DIB), 172 Dry Toner Substrates, 336
Crop factor, 79–80, 103 Device-dependent color space, 293 DSLR, see Digital Single Lens Reflex
CRT, see Cathode Ray Tube Device-independent color space, 293, DSNU, see Dark Signal Non-Uniformity
CT, see Color Temperature 296 Durst Lambda printer, 328
Curves tool, 199–203, 208 DIB, see Device Independent Bitmap DVD, see Digital Versatile Disc
Cyan, Magenta, Yellow, Key (CMYK), 171, Dichroic, 24 DVI, see Digital Visual Interface
177, 273, 293–300, 329 Dichromat, 270 DWT, see Discrete Wavelet Transform
Cycles per millimeter, 241 DICOM, see Digital Imaging and Dye diffusion, see Dye sublimation
Communications in Medicine printing
D Differential encoding, 251, 255 Dye sublimation inks, 331
D50, see CIE Standard Illuminant Daylight Diffraction, 35, 63–64 Dye sublimation printing, 331
5500K Diffraction grating, 1, 31, 32 Dye sublimation substrates, 337

353
 i n d ex

Dye-based ink, 331–332 F Frequency, 2–3, 14

Dynamic range, see Scene Dynamic f-number, 104–105, 119, also see Frequency histogram, 109–110
Range Aperture Frequency Modulation (FM), 325–326
f-stop, 84–85 Front principal focal point, 38, 50
E F2, see CIE Standard Illuminant Front-focusing, 245
Eco-solvent ink, 332 Fluorescent Full frame sensor, 41–44, 58, 79–81, 94,
Effective aperture, 58, 104 Farnsworth-Munsell 100 Hue Test (FM- 96
Effective f-number, 58, 61 100 Hue Test), 269
Einstein, Albert, 2 FAT, see File Allocation Table G
EIS, see Electronic Image Stabilization Field of view, 41–45, 68–69, 80, 90, 145, Gamma correction, 162, 230, 233
Electrical discharge, 7, 27 153, 233, 267, 283, 317 Gamma encoding, 162
Electromagnetic spectrum, 3–4, 132 File Allocation Table (FAT), 181 Gamma rays, 4
Electronic flash, 2, 17, 29–30, 86, 93–94 Filter factor, 128–129, 134 Gamut clipping, 300–301
Electronic Image Stabilization (EIS), 91, Filter masking, 223 Gamut compression, 301
97, 144 Fine Art photo papers, 334 Gamut Handling, 300
Electronic shutter, 87–88 Finest Detail, 340 Gaussian blur filter, 219
Electronic Viewfinder (EVF), 35, 92, 95, Fixed pattern noise, 235, 247 GB, see Gigabyte
308, 320 FLAC, see Free Lossless Audio Codec General Requirements for Applications
Electrophotographic printing, 329 Flash guide number, 30 in Commercial Offset Lithography
Encoding, 145–151, 251 Flash memory, 98, 183 (GRACoL), 297
Equivalent exposure, 113–114, 129 Flash meters, 113 Generation loss, 170, 263
eSATA, see External Serial Advanced Flash-synchronous, 85 Generic profiles, see Canned profiles
Technology Flicker fusion, 138 Geographic Information Systems (GIS),
Ethernet, 99, 150, 186–187 Fluorescence, 9, 339 171
ETTR, see Expose To The Right Fluorescent lamps, 8, 16, 20, 27–28 Geostationary Earth Orbit Tagged Image
EV, see Exposure Value Flux, 11–12, 19, 315 File Format (GeoTIFF), 171
Evaluative metering, see Matrix metering FM, see Frequency Modulation GeoTIFF, see Geostationary Earth Orbit
EVF, see Electronic Viewfinder Focal length, 37–44, 50–52 Tagged Image File Format
Exchangeable Image File Format (EXIF), Focal length and depth of field, 56–57 GHz, see Gigahertz
179 Focal length and f-number, 38 GIF, see Graphics Interchange Format
EXIF, see Exchangeable Image File Focal length and object distance, 48–49 Gigabyte (GB), 171–173, 180, 182–183,
Format Focal plane shutter, 86–87, 93,118, 143 186–189
Expose To The Right (ETTR), 118–119 Focal point, see Principle focal point Gigahertz (GHz), 188
Exposure compensation, 116–117 Focus non-uniformity, 240, 246 GIS, see Geographic Information Systems
Exposure latitude, 107 Footcandles, 12–13, 15–16 Global enhancement, 215–216, 223
Exposure meters, 115 Footlambert, 14–16 Global Positioning System (GPS), 94,
Exposure time, 102 Four-quadrant tone reproduction digital 179
Exposure Value (EV), 113–114, 118, system, 195 Global shutter, 87
123 Four-Thirds, 79 Gloss, 338–339, 344–345, 347–348
Extensible Metadata Platform (XMP), 175, Foveon Sensor, 78 Gloss differential, 344
178, 180 FPS, see Frames Per Second Gloss Unit (GU), 348
External Serial Advanced Technology Frame rate, 140–143, 148, 151–153, 172, Glossy surface, 24
(eSATA), 186–187 232, 313–316 Golden hour, 25
External Serial Advanced Technology Frames Per Second (FPS), 99, 138– GOP, see Group of Pictures
Attachment (eSATA), 186 143,151–153, 232 GPS, see Global Positioning System
Free Lossless Audio Codec (FLAC), 150 GPU, see Graphics Processing Unit

354
 i n d ex

GRACoL, see General Requirements for High-intensity discharge lamps, 28–29 In-Plane Switching (IPS), 310
Applications in Commercial Offset High-pressure discharge sources, 28 Incandescence, 2, 7, 9, 27, 29
Lithography High-speed printing, 339–341 Incandescent, 7, 16, 18, 20, 26, 121, 280,
Graduated neutral density filters, 131 Highlights, 22, 84, 107–108, 110–111, 282
Graininess, 237 116–119 Index of refraction, 36–37, 63
Graphical drawing (lens), 47, 49–51 HMD, see Head Mounted Display Information theory, 250–251
Graphics Interchange Format (GIF), 172, Hot pixel defects, 74 Infrared, 4–5, 7–9, 89
256 HTML, see Hypertext Markup Language Infrared filter, 78, 132–133
Graphics Processing Unit (GPU), 188–189, Hue, 277 Infrared radiation (IR), 4, 30, 74–75, 77
233, 308 Hue angle, 303 Inkjet Printing, 326–327, 331–334
Gray card, 17–18, 115, 117, 237 Hue Value/Chroma (H V/C), 279 Integration time, 104
Group of Pictures (GOP), 147–148, 152 Huffman coding, 251, 255–256 Intensity, 10–14, 19, 102, 233
GU, see Gloss Unit Human eye, 4–5, 14, 42, 53, 107, 266–268, Interchangeable lens, 41, 72, 88, 91,
Guide numbers, 118 294, 327 94–95, 98
Huygens, Christiaan, 2 Interframe compression, 147, 152, 172
H Hyperfocal distance, 54–55 Interframe Encoding, 146–149
Halftone printing, 325 Hypertext Markup Language (HTML), 172 Interlaced, 144
Halo, 214, 262 Hz, see Hertz International Color Consortium (ICC), 292
Handheld shutter speed rule of thumb, International Commission on
103 I Illumination (CIE), 4, 279
Hard Disk Drive (HDD), 182–183, 189 I-frame, see Intra-coded frame International Press Telecommunications
Hard proofing, 305, 343–344, 348 ICC, see International Color Consortium Council (IPTC), 179
HD, see High Definition ICC-absolute colorimetric, 301–302 International Standards Organization
HDD, see Hard Disk Drive iccMAX Profile, 297 (ISO), 105–106, 194
HDMI, see High-Definition Multimedia Illuminance, 12–19, 35, 62, 102, 130, 271, International Telecommunication Union
Interface 295 (ITU), 258, 316
HDR, see High Dynamic Range Illuminant, 6, 11, 122, 280, 297 Internet Service Provider (ISP), 185
Head Mounted Display (HMD), 312–313, Illuminant A, 5, 17, 25, 274–275, 280, 300 Interpixel redundancy, 251
320 Image bit depth, 81 Interpolation, 158–160
HEIF, see High Efficiency Image File Image catalog, 177–178 Intra-coded frame, 147–149, 152
Format Image distance, 35–38, 40, 44 Intraframe compression, 146, 253
Helmholtz, Herman von, 269 Image histogram, 109–113, 196, 199, 204, Intraframe encoding, 146, 149
Hering opponent color theory, 268 208 Inverse square law, 13, 93, 118
Hertz (Hz), 3, 313 Image ingestion, 175, 177, 215 IPS, see In-Plane Switching
HEVC, see High Efficiency Video Coding Image noise, 106, 216, 234–237 IPTC, see International Press
HFR, see High Frame Rate Image processing pipeline, 114–115,161, Telecommunications Council
High Definition (HD), 144–145, 314–316 231 IR, see Infrared
High Dynamic Range (HDR), 119–120, 314 Image sensor, 73–81, 83, 233–234 IR cut-off filter, 133
High Efficiency Image File Format (HEIF), Image sensor size, 41–43, 58,78–81, 94, IS, see Image Stabilization
169, 172–173 97, 105, 158, 169, 247 Ishihara Color Vision Test, 269
High Efficiency Video Coding (HEVC), Image Signal Processor (ISP), 230–233, ISO, see International Standards
149–150, 172 236, 240 Organization
High Frame Rate (HFR), 140–142 Image size, 35, 39, 45–52, 57 ISO Base Media File Format (ISOBMFF), 172
High pass filtering, 217 Image size (file), 164–165, 171, 173 ISOBMFF, see ISO Base Media File Format
High-Definition Multimedia Interface Image stabilization (IS), 90–91, 103 ISP, see Image Signal Processor or
(HDMI), 186–187, 312–314 Impression cylinder, 325, 330 Internet Service Provider

355
 i n d ex

ITU, see International Lens axis, 37–40, 44, 50–51, 242 M

Telecommunication Union Lens cast, see Color shading MacAdam ellipses, 281
Lens speed, 105 MacAdam, David, 281
J Letterboxing, 145 Macro lenses, 58
JBOD, see Just a Bunch of Disks Levels tool, 203–204 Mapping, 199
Jerkiness, 151, 153 Light, 1–7, 10–14 Material Connection Space (MCS), 297
JND, see Just Noticeable Difference Light-Emitting Diode (LED), 1, 17, 21, Matrix metering, 115–116
Joint Photographic Experts Group – 30–32, 97, 309–310, 316, 328 Matte Black (MK), 339–340, 344
Lossless Compression (JPEG-LS), 170 Linear interpolation, 163–164, 165 Maxwell triangle, 129–130, 203,208, 272
Joint Photographic Experts Group (JPEG), Linear perspective, 46–49, 282, 284, 287 Maxwell, James Clerk, 134, 272
65, 74, 110, 113, 114, 145, 147, 149, Lines Per Inch (LPI), 325 MB, see Megabyte
151, 162, 169–175, 178, 179, 180–181, Liquid Crystal Display (LCD), 289, MCS, see Material Connection Space
189–191, 215, 218, 221, 231, 232, 234, 308–316, 321 Mean horizontal intensity, see
243, 253, 258–259, 260–263 Liquid toner, 329–330, 336 Candlepower
JPEG, see Joint Photographic Experts Liquid toner substrates, 336 Mean spherical intensity, 11
Group Local Area Network (LAN), 184, 186 mechanical compensation, 45
JPEG-LS, see Joint Photographic Experts Local enhancement, 215, 223 Media-relative colorimetric, 301
Group – Lossless Compression Lookup Table (LUT), 196, 199, 319 Medium format camera, 75, 92, 94–95,
Judder, see Jerkiness Lossless compression, 168–173, 250–256, 98, 186
Just a Bunch of Disks (JBOD), 183–184 261 Medium format sensor, 41–42, 79–80, 97
Just Noticeable Difference (JND), 243, Lossy compression, 148, 150, 152, Megabyte (MB), 180–181, 183
281 168–171, 173, 181, 232, 250–251, 253, Megapixel, 79, 81, 95, 97, 143, 158, 324
260–263 Mesopic vision, 267, 282
K Low pass filter, 83–84, 162, 163, 218–219 Metadata, 65, 79, 92, 145, 158, 168,
K, see Kelvin LPI, see Lines Per Inch 169–170, 172, 175, 176, 177, 178,
KB, see Kilobyte Luma, 148, 257–259, 316 179–181, 206, 232, 298
Kelvin (K), 6, 7, 16, 19 Luma-chrominance color space, 257–260 Metamerism, 274–275, 292
Keywords, 177, 179–180 Lumen, 11–12, 16, 19, 28, 304 Metercandle, 13–14, 16, 19
Kilobyte (KB), 180–181 Lumen method, 12 Micro videos, 153
Luminance, 14–19, 107, 138–139, 148, Micro-reciprocal degrees, 18–19
L 152, 160–161, 194–196, 252 Microlenses, 65, 77, 78
Lambertian surface, 15 Luminance and Light Adaptation, Midtone sharpening, 217
LAN, see Local Area Network 282–283 Mie scattering, 25
Large format camera, 35, 79, 86, 95–96 Luminance mapping, 162 Mineral whiteners, 339
Laser, continuous tone printing, Luminance noise, 235–238 Mini-Serial Advanced Technology
328–330. 335 Luminance non-uniformity, 240 Attachment (mSATA), 183
Lateral adaptation, 283 Luminance, Chroma Blue, Chroma Red Mini-Serial Advanced Technology
Latex inks, 333 (YCbCr), 257–260 Attachment (mSATA), 183
Layer mask, 168, 173, 201, 223–224, Luminescence, 9 Mired scale, 18–20
225–227 Luminosity histogram, 109, 110, 196–197, Mired value, 17, 18–20
LCD, see Liquid Crystal Display 208 Mirrorless camera, 78–79, 89, 96, 97, 245,
Leaf node, 255–256 Luminous flux, 11–12, 19 247
Leaf shutter, 85–86, 95 LUT, see Lookup Table MJPEG, see Motion Joint Photographic
LED, see Light-Emitting Diode LZW, see Lempel–Ziv–Welch Experts Group
Lempel–Ziv–Welch (LZW) compression, compression MK, see Matte Black
171, 172, 190, 255, 256–257 Mobile phone camera, 96–97

356
 i n d ex

Modulation Transfer Function (MTF), Noise Reduction, 119, 122–124, 161, 162– Paper weight, 328, 337–338
241–242 163, 230–232, 237–238, 246–247 Paper white point, 339, 347
Moiré, 83–84, 163, 215, 325 Normal focal length, 42–43, 44, 48 Parametric Image Editing (PIE), 178–179,
Monitor density, 196 NTSC, see National Television Standards 206
Monochromat, 270 Committee Partitive mixing, 271, 275–276, 288–289,
Monoscopic audio, 150 Nyquist frequency, 84 308, 326
Mosquito noise, 151, 152 Nyquist limit, 163 PB, see Petabyte
Motion effect, 138 PCIe, see Peripheral Component
Motion estimation vector, 147–148 O Interconnect Express
Motion jitter, 140 OBA, see Optical Brightener Agent PCS, see Profile Connection Space
Motion Joint Photographic Experts Object distance, 38, 40, 44, 46–48 Perceptual rendering intent, 301
Group (MJPEG), 149 Objective tone reproduction, 194 Peripheral Component Interconnect
Motion video, 137–138, 140–142, Offset lithography, 325, 329 Express (PCIe), 186–187
145–146, 148, 149, 150, 151–153, 202, OIS, see Optical Image Stabilization Persistence of vision, 272
316 OLED, see Organic Light-Emitting Diode Persistent image, 138–139
Moving Picture Experts Group (MPEG), OLPF, see Optical Low-Pass Filter Perspective, 46
145, 149, 172 Opacity, 126–127, 130–131, 333 Petabyte (PB), 180
MPEG, see Moving Picture Experts Group Operating System (OS), 97, 150, 162, 172, Phase detection autofocus, 73, 88–89,
MPEG-4 Advanced Video Coding (AVC), 173, 175–176, 178, 183–184, 188, 189, 97, 245
149, 150, 152 190, 300, 312, 318, 319, 320–322 Phi phenomenon, 140
mSATA, see Mini-Serial Advanced Opponent colors theory, 268–270 Phosphorescence, 9, 308, 311
Technology Attachment Optical axis, 37, 59, 93 Photo Black (PK), 339–340
MTF, see Modulation Transfer Function Optical Brightener Agent (OBA), 338–339, Photo Imaging Plate (PIP), 329–330
Munsell (hue, value, chroma), 277–279 346 Photo paper, 24, 218, 305–306, 333–339,
Munsell Color System, 276–278 Optical Image Stabilization (OIS), 90, 97, 344–349
Munsell, Albert Henry, 277, 290 103 Photodiode, 17, 73, 78, 106
Optical Low-Pass Filter (OLPF), 83, 163 Photoelectric effect, 73
N Optical viewfinder, 34–35, 72–73, 85, 89, Photographic exposure, 102
Nanometer (nm), 3, 19, 280 91–92, 94–96 Photoluminescence, 9
NAS, see Network-Attached Storage Optimal pinhole size, 35–36, 64 Photon noise, 235, 236
National Institute of Standards and Organic Light-Emitting Diode (OLED), 9, Photons, 2, 9, 30, 73–74, 75, 77, 81, 83,
Technology (NIST), 21 187, 308–309, 311–312, 314, 316, 317 87–88, 106, 107–109, 118, 162, 233,
National Television Standards Committee OS, see Operating System 236, 238–239
(NTSC), 315–316 Output resolution, 339, 341–342 Photopic luminosity function of the
ND, see Neutral Density Output sharpening, 217–218 human eye (VM(λ)), 4–5
Nearest neighbor interpolation, 164–166 Overexposure, 102, 108, 110–111, Photopic vision, 267, 282
Negative lens, 38 114–115, 130 Photosensitive light writing printing,
Network-Attached Storage (NAS), 184, Overscan, 91, 144 328–329
186 Overshoot, 152, 214 Photoshop Big format (PSB), 173
Neutral Density (ND), 128–130, 139 Photoshop Document (PSD), 173, 185, 224
Newton, Isaac, 2, 268 P Photosite, 64–65, 73–81, 108, 160–161,
NIST, see National Institute of Standards P-frame, see Predicted frame 232–233, 239
and Technology Packet loss, 152 PIE, see Parametric Image Editing
Nodal point, 37–41, 44, 51 Pantone System, 277 Piezoelectric inkjet, 327–328
Nodal slide, 39, 44 Paper dynamic range, 342, 347 Pigment, 139, 270, 272, 275–276, 329,
Noise, 73, 76, 106, 118, 234, 237 Paper thickness, 227 331–332, 333–335, 344, 347

357
 i n d ex

Pigment-based ink, 331–332, 334 PSD, see Photoshop Document Resolution (display), 312–313
Pillarboxing, 145 Psychological primaries, 268 Resolution (image), 64, 79, 81, 89, 91, 95,
Pinhole camera, 34–36, 63–64 Psychovisual redundancy, 251, 252–253, 97, 140, 158, 160, 163–166, 168, 169,
PIP, see Photo Imaging Plate 258, 260 170, 172, 178, 179, 232, 234, 241–242,
Pixel addressing, 224 258–259, 316, 324, 342–343
Pixel density, 312–313, 317, 324 Q Resolving power, 53, 64, 83, 241–242,
Pixel pitch, 81, 97, 317–318 QD, see Quantum Dot 246, 342
Pixel Response Non-Uniformity (PRNU), QE, see Quantum Efficiency Reversed telephoto wide-angle lens, 60
235–236, 239 Quantization error, 151, 152 Revolutions Per Minute (RPM), 182, 329
Pixel value, 73, 75, 110, 158–165, 198–203 Quantization of energy, 2 Rewritable (RW), 182
Pixels Per Inch (PPI), 312, 324, 342–343 Quantization stage, 251 RGB, see Red, Green, Blue
PK, see Photo Black Quantum Dot (QD), 308, 311–312 RGB histogram, 197
Planck, Max, 2 Quantum Efficiency (QE), 73, 75, 77, RGBAW, see Red, Green, Blue, Amber,
Planck’s equation, 6 311–312 White
PNG, see Portable Network Graphics Quantum electrodynamics, 2 Ringing artifacts, 151–152, 262
POD, see Print On Demand Quantum theory, 2 RIP, see Raster Image Processor
Point and shoot camera, see Compact RLE, see Run-Length Encoding
camera R Rolling shutter artifact, 87
Point source, 12, 13–14, 22–23, 104, 118 RAID, see Redundant Array of RPM, see Revolutions Per Minute
Pointer’s Gamut, 294–295, 316 Independent Disks Run-Length Encoding (RLE), 251,
Polarization, 23–24, 128, 131–133, RAM, see Random Access Memory 253–254
308–309, 311 Random Access Memory (RAM), 114, 188, RW, see Rewritable
Polarizing filter, 24, 128, 131–133, 308 308
Porous inkjet, 334–335 Rangefinder camera, 92, 95, 97–98 S
Portable Network Graphics (PNG), 172 Raster graphics, 168, 172 S-IPS, see Super In-Plane Switching
Positive lens, 37–39, 62, 63 Raster Image Processor (RIP), 330, 343 SaaS, see Software as a Service
Posterization, 198, 237–239, 262–263, Raw, 159–163, 168–181 Saccade, 140
315, 348 Raw file, see Raw Sampling rate, 140, 150, 163
PPI, see Pixels Per Inch Rayleigh scattering, 25, 133 SATA, see Serial Advanced Technology
Predicted frame, 147–148, 152 RC, see Resin-Coated Attachment
Prime lens, 45, 240 Read noise, 236 Saturation rendering intent, 301
Principle focal point, 38–41 Rear curtain synchronization, 118 Scale of reproduction, 39, 52–53, 56, 59,
Print On Demand (POD), 329 Red, Green, Blue, Amber, White (RGBAW), 288
Printer drivers, 339, 342 17 Scattered (light), 23, 25, 127, 132
Printer profiles, 296–297, 300 Redundant Array of Independent Disks Scene dynamic range, 107, 111, 118,
Printhead, 326–328, 330, 332, 339–341 (RAID), 183–184 198
Printhead height, 339–340 Reflected (light), 15, 24–25, 30, 127, 194, Scene luminance ratio, 107
Printing blanket, 325 235, 271–273 Scotopic vision, 93, 267, 282
Printing plate, 325, 329, 343 Reflection factor, 15 SD, see Standard Definition or Secure
PRNU, see Pixel Response Refraction, 36, 41, 62–63 Digital
Non-Uniformity Refresh rate, 92, 313–314, 320 SDHC, see Secure Digital High Capacity
Profile Connection Space (PCS), 296–297, Relative aperture, 45, 104 SDR, see Standard Dynamic Range
300 Rendering Intents, 292, 293, 300–302, SDUC, see Secure Digital Ultra Capacity
Progressive scan, 144 305–306 SDXC, see Secure Digital Extended
ProPhoto RGB, 161–162, 294 Resampling, 163–166 Capacity
PSB, see Photoshop Big format Resin-Coated (RC), 328, 334, 339 Secure Digital (SD), 98–99

358
 i n d ex

Secure Digital Extended Capacity (SDXC), Sparse color filter array, 76 Synthetic aperture, see Synthetic depth
98–99 Spatial audio, 150 of field
Secure Digital High Capacity (SDHC), Spatial color mixing, see Partitive mixing Synthetic depth of field, 105
98–99 Spatial frequency, 241 System on a Chip (SoC), 230
Secure Digital Ultra Capacity (SDUC), Spatial resolution, 81, 232, 312, 324, 327
98–99 Spatial sampling, 140 T
Sensor dynamic range, 107, 114, 233, SPD, see Spectral Power Distribution or Tagged Image File Format (TIFF), 171
238–239 Silicon Photodiode TB, see Terabyte
Sensor speed, 105 Specifications for Web Offset Telephoto focal length, 44, 95
Serial Advanced Technology Attachment Publications (SWOP), 297–298 Temporal redundancy, 146, 253
(SATA), 183, 186 Spectral colors, 5–6, 272 Temporal sampling, 140
Sharpening, 173, 287 Spectral Power Distribution (SPD), 6, 29, Terabyte (TB), 180–182
Sharpening filters, 214 73 Thermal dye-diffusion, 330–331
Shift register electronics, 75 Spectrophotometer, 133, 243, 274, 280, Through the Lens (TTL), 88, 115, 129
Shot noise, see Photon noise 298–299, 346 TIFF, see Tagged Image File Format
Shutter angle, 142–143 Specular, 22–23, 66, 84, 117, 166, 206, 348 TN, see Twisted Nematic
Shutter button, 73, 84–85, 90, 94, 103, Split toning, 208–209 Tonal range, 110–111, 119, 153, 198–199,
116–118 Spot color ink, 329 204–207, 238–240
Shutter Priority, 117 Spot metering, 15, 102, 115–117, 239 Tone curve, 199, 202–203, 208
Shutter release, see Shutter button sRGB, 161–162, 294, 296, 316 Tone mapping, 199, 208, 233
Shutter Speed (photo), 85–86, 90–96, SSD, see Solid State Drive Tone Response Curve (TRC), 319
102–105, 113–119 Standard Definition (SD), 144–145, 258, Tone Transfer Compensation (TTC), 199
Shutter Speed (video), 143 314, 316 Transcoding, 148
Shutters, 85–87 Standard Dynamic Range (SDR), 314–315 Transfer curve, 199
Sidecar file, 175, 178, 179, 180 Standard illuminants, 6,17 Transmission stops (t-stops), 85
Signal to Noise Ratio (SNR), 236–237, 247 Steradian, 12 Transmittance, 126–134
Silicon Photodiode (SPD), 17, 73 Stereoscopic audio, 150 Transmitted (data), 150
Simultaneous contrast, see Lateral Stereoscopically, 150 Transmitted (light), 2, 24, 126- 128, 271,
adaptation Stochastic screening, 325–326 298, 309
Single Lens Reflex (SLR), 72, 94 Stop, 102–103, 106, 113–114, 116, 129 TRC, see Tone Response Curve
Single servo focus, 90 Stretch mapping, 203, 210 Triboluminescence, 9
Skylight, 17, 23, 25 Stronger perspective, 48 Trichromat, 268, 270–271
Slide mapping, 203 Subject magnification, 41, 43–44 Tristimulus values, 280, 282, 293, 320–321
SLR, see Single Lens Reflex Subjective tone reproduction, 195 TrueSense, 76, 99
Small format camera, see Compact Camera Substrate, 305, 324–330, 332, 336–337 TTC, see Tone Transfer Compensation
Smart sharpening, 216–217 Subtractive color mixing, 271–272, 294, TTL, see Through the Lens
Snell’s Law of Refraction, 36 324, 343, 340–344 Tungsten, 5–7, 10, 13, 17, 20–21, 24,
SNR, see Signal to Noise Ratio Subtractive primary colors, 273, 325 26–31, 119–121, 275
SoC, see System on a Chip Sunlight, 6,12, 17–18, 23–26, 268, 282 Tungsten-Filament Lamps, 26, 31
Soft proofing, 305, 339, 343 Super In-Plane Switching (S-IPS), 310 Tungsten-halogen lamps, 27, 28, 31
Software as a service (SaaS), 178 Super MicroWeave, 339–341 Twisted Nematic (TN), 310, 317
Software filters, 92, 130, 213–218, 221, Surface sheen, 337–338
224–226 Surface texture, 329, 333, 338 U
Solid State Drive (SSD), 183, 189 Swellable polymer, 332–334 UDMA, see Ultra Direct Memory Access
Solvent inks, 332–333 SWOP, see Specifications for Web Offset UHD, see Ultra High Definition
Source profile, 296 Publications UHS, see Ultra High Speed

359
 i n d ex

UI, see User Interface VM(λ), see Photopic luminosity function

Ultra Direct Memory Access (UDMA), of the human eye
99 VP9, 149–150
Ultra High Definition (UHD), 144, VR, see Virtual Reality
315–316 VVC, see Versatile Video Coding
Ultra High Speed (UHS), 99–100
Ultraviolet (UV) radiation, 4–5, 7, 8–9, W
27–28, 75, 331–332, 338 Wave theory, 2,4
Ultraviolet filters, 132–133 Wavelength, 2–9, 17, 19, 22, 25, 27–30,
Underexposure, 102, 108, 110–111, 36, 280–281, 285
114–115, 128 Wavelength (filters), 127–133
Universal Serial Bus (USB), 186–189, 312 Wavelength (optics), 62, 64
Unpolarized light, 23 Wavelength (quantum efficiency), 73,
Unsharp masking, 216 77–78
USB, see Universal Serial Bus Wavelength (tristimulus values), 280–281
User Interface (UI), 305, 312 Wavelength (vision), 266–267, 272
UV, see Ultraviolet radiation WCAG, see Web Content Accessibility
UV curable inks, 332–333 Guidelines
WCG, see Wide Color Gamut
V Weaker perspective, 48
VA, see Vertical Alignment Web Content Accessibility Guidelines
Vapor lamp, 7–8 (WCAG), 283, 290
Vapor sources, 7–8 Weighted average filter, 218–219
Variable, 45 White Balance, 18, 20, 115–122, 205–206,
Variable focal length, 39, 41, 45, 105 230, 244–245
Vector, 147–148, 168, 173, 223 Whole stop, 102–104
Velocity, 2 Wide Color Gamut (WCG), 153, 202, 316
Versatile Video Coding (VVC), 149 Wide-angle lenses, 44–45, 58, 60
Vertical Alignment (VA), 310, 317 WiFi, 99, 186–187, 312
Video codec, see Codec WORM, see Write-Once-Read-Many
Video quality artifacts, 138, 150–151 Write-Once-Read-Many (WORM), 182
View camera, 35, 98
Viewing angle, 287, 308, 310, 316–317 X
Viewing distance, 53, 56, 58, 153, 214, X-ray, 4
218, 237 X-Trans color filter array, 74–75
Viewing distance (display), 275, 289, 312, XMP, see Extensible Metadata Platform
317, 320
Viewing distance (print), 324, 341–343 Y
Virtual image, 38 YCbCr, see Luminance, Chroma Blue,
Virtual Reality (VR), 153, 320 Chroma Red
Visible light, 4, 17, 25–31, 62, 132–133, YCbCr color space, 257–260
280, 339 Young-Helmholtz Theory, 268
Visible perception, 266–267,284, 286 Young, Thomas, 2, 268
Visible spectrum, 4–6, 77–78, 129,
132–133 Z
Visual memory, 138–140, 147 Zoom lenses, 41, 45, 105

360