Advanced Computer Graphics

Light and Color 1

Matthias Teschner
Outline
 Context
 Light
 Color

University of Freiburg – Computer Science Department – 2


Light

 
specular diffuse

Light travels Light is Incoming light /

along rays Cameras
emitted at is absorbed and capture light
light sources scattered at surfaces

 How to quantify light transported along a ray?  Radiance

 How to quantify surface illumination?  Irradiance
 How to quantify light at a pixel?  Radiance
University of Freiburg – Computer Science Department – 3

Surface Reflection Properties
 How much incident light from a particular direction
is reflected into a particular direction?
 Bidirectional Reflectance Distribution Function BRDF

Incoming light
from direction Outgoing light into direction

Position

University of Freiburg – Computer Science Department – 4


Rendering Equation
 How to compute reflected light into a particular
direction given incident light from all possible
directions?  Rendering equation

Rendering equation:
Outgoing light into direction is a sum of
incident light from all directions weighted
with material properties
Incoming light
from direction
Position

University of Freiburg – Computer Science Department – 5


The Importance of Light Modeling
Light

Color

Material

Rendering equation

Solving the rendering equation

University of Freiburg – Computer Science Department – 6
Outline
 Context
 Light
 Color

University of Freiburg – Computer Science Department – 7


Light
 Radiation modeled with photons
 Light particles
 Energy parcels
 Travel along a straight line at the speed of light
 Characterized by a wavelength
(perceived as color in the visible spectrum)
Photons travel
along rays

University of Freiburg – Computer Science Department – 8

Quantifying Light
 Radiometric quantities characterize the
propagation of electromagnetic radiation
 Flux, irradiance, radiance
 Radiation with wavelengths between 390 nm
and 750 nm is visible to humans
(blue light  green light  red light)
 Radiometric quantities are represented by a spectrum
 A distribution function of wavelengths
 Amount of light at each wavelength interval
University of Freiburg – Computer Science Department – 9
Flux
 Radiant flux 
 Power
 Radiant energy, i.e. number of photons, per time

Flux is actually radiant energy per time.

As photons carry varying energy depending

on their wavelength, number of photons
per time is an approximation that improves
the intuition behind flux.

University of Freiburg – Computer Science Department – 10


Flux Density
 Rate at which flux enters, leaves or passes an area

 Describes strength of light with respect to a

surface area (existing or virtual surface)
 No directional information

University of Freiburg – Computer Science Department – 11

Flux Density - Variants
 Irradiance - incident / incoming flux per surface
 Radiosity - outgoing flux (reflected plus emitted)
per surface
it is important in diffuse reflectors

Irradiance – Incident flux per area Radiosity – Outgoing flux per area

University of Freiburg – Computer Science Department – 12

Spatially Varying Flux Density
 Irradiance at a position ?
 Issues: position with zero area, no flux per position
 Solution: infinitesimals, differentials, small quantities
 Consider a small amount of flux incident
to a small area around position
 For , we have ,
and the ratio converges to the
irradiance at :
Irradiance
at a position
University of Freiburg – Computer Science Department – 13

Overall Flux Incident to a Surface
 Infinitesimally small amount of flux at a position
Conceptually, dA converges to zero, but
we can still think of a small surface patch.

 Flux over an area

Area is discretized
into surface patches
with constant
irradiance per patch.
University of Freiburg – Computer Science Department – 14
 Towards Directional Quantities watch video again for this slide

 How to quantify light from / into a direction?

 E.g., light towards viewer or towards surfaces
 Issue: Flux from / into a particular direction is zero
 Analogous to flux per position
 Solution: Flux from / into a range of directions
 Represented by angles in 2D
 Represented by solid angles in 3D

irradiance or radiosity does not encode any information aboubt the direction of flux or photon.

University of Freiburg – Computer Science Department – 15

Solid Angle
 Area of a sphere surface divided
by the squared sphere radius

 E.g., solid angle of the entire

sphere surface
 Independent from the radius
 E.g., solid angle of a hemisphere
Wikipedia: Raumwinkel

University of Freiburg – Computer Science Department – 16

Solid Angle and Surface Area
 E.g., from which directions does a point receive light
from an area light source?
 Solid angle of an arbitrary surface

University of Freiburg – Computer Science Department – 17


Infinitesimal Solid Angle and Surface Area

 is an approximation
 If an infinitesimally small area at position
converges to zero, then the solid angle also
converges to zero and the relation
is correct in the limit

University of Freiburg – Computer Science Department – 18

Solid Angle Subtended by a Surface
 How big does an object appear in an image? From
which solid angle does a point receive light from a
light source?

University of Freiburg – Computer Science Department – 19


Visibility Function
 Position only contributes to ,
if it is visible from
 Therefore,

with , if is visible
from and , if
is not visible from

University of Freiburg – Computer Science Department – 20


Directional Flux per Area
 Flux per area per
orientation of the sensor plane does not matter , here we do not
compute the illumination strength of surface but only illumination in
the grey area.

solid angle
 Photons per time that
hit an area from directions
within a solid angle
 Flux per projected area
per solid angle
 How much flux travels
through the grey area
 Independent from sensor orientation
University of Freiburg – Computer Science Department – 21

Radiance
 If the area around a position converges to zero
and the solid angle around direction converges to
zero, then the flux that hits (passes, is reflected
from) from (into) solid angle converges to
zero and is the radiance at position
from (into) direction
L(x, ) characterizes the flux that travels
through position x in direction  .

The notation d2 indicates that two integrations

(over area and over solid angle) are required to
get a non-infinitesimal value .
University of Freiburg – Computer Science Department – 22

Radiance at a Position in a Direction
 Actual setting

 Flux that is transported

through an infinitesimally
small cone
 Simplified notion

 Radiance at position in direction

 Flux that is transported along a ray
University of Freiburg – Computer Science Department – 23
Flux Density and Radiance - Terms
 Flux per area flux density variants is irradiance and radiosity
flux per area is flux density

 Flux density
 Incident / incoming flux density: Irradiance
 Exitant / outgoing flux density: Radiosity
 Flux per projected area (orthogonal to flux direction)
per solid angle flux per projected area is radiance

 Radiance
 Incident, outgoing radiance: Radiance

University of Freiburg – Computer Science Department – 24

Radiance and Oriented Surfaces
 Two areas around positions with
 Angles between surface normal and flux direction :
this explains that radiance depends on the flux along ray and not depend on the sensor plane.

 Radiance at : cos theta 1 is 1 since theta is 0

 Radiance at : da1 is equal to da2

Radiance describes the flux

within the grey area independent
from the plane (sensor) orientation.

 University of Freiburg – Computer Science Department – 25

Irradiance and Oriented Surfaces
here the orientation matters

da1 and da2 are of different sizes and hence the irradiance would differ.

 Irradiance at :
 Irradiance at :

i denotes
an arbitrary
orientation. Irradiance describes the effect of
 Lambert’s Cosine Law the flux within the grey area onto
a surface. I.e., the orientation of
 Irradiance on a surface is proportional the surface with respect to the
flux direction matters.
to the cosine of the angle between
surface normal and flux direction
University of Freiburg – Computer Science Department – 26

Lambert's Cosine Law
 Angle between surface normal and light source
direction influences the surface brightness
The same light source illuminates
a surface at different angles.
The same flux and the same radiance
is transported along the rays.

in both the cases the radiance might be same but the irradiance is different.

Surface receives more flux Surface receives less flux

per area. Appears brighter. per area. Appears darker.

University of Freiburg – Computer Science Department – 27

Discussion
 Radiance characterizes the flux that is transported
between infinitesimally small surface areas (along rays)
 Irradiance characterizes the effect of this flux at these
surface areas

Flux d is reflected / emitted from dA1 and incident to dA2.

Distance from x1 to x2 is r.
Angles between flux direction and surface normal are 1 and 2.
Size of dA1 seen from x2 is the solid angle d2. d1 analogous.

University of Freiburg – Computer Science Department – 28

Discussion – Conservation of Radiance
if angle theta 1 and theta 2 increases then the flux density decreases

 Radiosity at :
 Irradiance at :
 Radiance at :

 Radiance at :

Conservation of radiance.
Radiance describes flux
transported along a ray.
 University of Freiburg – Computer Science Department – 29
Discussion – Inverse Square Law
 Irradiance at an illuminated surface decreases
quadratically with the distance from a light source
 Surfaces appear darker with growing distance from light sources
 Flux generated at A, arriving
at A1 and A2:
 Areas
radiance./projected area./solid angle

we can see from figure that the area increases quadratically from a to a1 and a2

 Irradiances All planes are orthogonal to .

Thus, cos  = 1 for all planes.

area increases quadratically , irradiance decreases quadratically with distance from light source.

University of Freiburg – Computer Science Department – 30

Sensor Model
 Pinhole camera model
 Sensor elements with a
small area receive flux
from a small solid angle,
i.e. radiance if the sensor plane is orthogonal to the flux direction

 Radiance
 Is measured by sensors
 Is computed in computer- Idealized graphics model
of an imaging sensor
generated images
[Akenine-Möller et al.]
University of Freiburg – Computer Science Department – 31
Discussion – Sensors Measure Radiance
object in a photo captured by a camera by observer. if the observer changes the location then the ,brightness remains same.

 Surface brightness is independent from the distance

between surface and viewer / camera / sensor
 Flux at A decreases quadratically
with distance r, if A1 moves, e.g.,
from distance r1 to r2
 The area visible at A grows
quadratically with distance r. The sensor at A distance does not matter in this case hence
the radiance and flux remains the same at

receives flux from a quadratically growing area

the sensor.
condition : entire sensor element must be
considered only then this applies.

 Both effects cancel, if the same radiance

over the areas A1 and A2 is reflected onto A
 University of Freiburg – Computer Science Department – 32
Discussion – Sensors Measure Radiance

 Surface brightness is independent from the distance

between surface and viewer / camera / sensor
 Not true, if a surface does not
cover a sensor element, e.g.,
object appears smaller than
a pixel
 Flux at sensor area A decreases
quadratically with distance r
 Radiance decreases quadratically with r

University of Freiburg – Computer Science Department – 33


Discussion – Irradiance and Radiance
 Illumination strength at a surface can be characterized
by irradiance (flux per area)
 Depends quadratically on the distance between surface
and light source
 Illumination strength at a sensor element can be cha-
racterized by radiance (flux per area per solid angle)
 Does not depend on the distance between surface and
sensor

University of Freiburg – Computer Science Department – 34

Discussion – Irradiance and Radiance
Light 1 Object surface receives less flux from light 2
than from light 1 (inverse square law). Both
Object lights contribute to the illumination of the
surface same surface element.

Brightness Camera captures the same radiance for

characterizes both lights. d1 and d2 are of the same
irradiance. size into different directions. Light 1 and 2
Camera contribute to different sensor elements.
(Pinhole
model)

Sensor Light 2
response Direction:
characterizes Solid angle:
radiance.

University of Freiburg – Computer Science Department – 35

Radiometric vs. Photometric Quantities

 Radiometric quantities describe all types of radiation

 Preferred in graphics research
 E.g., flux, irradiance, radiosity, radiance
 Photometric quantities describe visible radiation
weighted with the sensitivity of the human eye
 E.g., luminous flux [lumen], illuminance [lux], luminous
exitance [lux], luminance [candela / m2]

University of Freiburg – Computer Science Department – 36

Summary
 Flux describes the number of photons per time
 More precisely photon energy per time
 Irradiance and radiosity describe the flux into, through
or from a surface per area
 Irradiance describes the illumination of surfaces
 Radiance describes the flow at a direction into
or from a surface orthogonal to that direction
per area per solid angle
 Radiance is measured by sensors
University of Freiburg – Computer Science Department – 37

Advanced Computer Graphics
Light and Color 2
Matthias Teschner
Outline
 Context
 Light
 Color

University of Freiburg – Computer Science Department – 39


Introduction
 Light consists of a set of photons
 Photons are characterized by a
wavelength within the visible
spectrum from 390 nm to 750 nm
 The distribution of wavelengths
within this set is referred to as
spectral power distribution (spectrum)
 Spectra are perceived as colors

University of Freiburg – Computer Science Department – 40


Spectral Quantities
 Flux, flux density and radiance depend on wavelength

Photons with a wavelength

in a range i around  i.
University of Freiburg – Computer Science Department – 41

Visible Spectrum
 If the spectrum consists of a dominant wavelength,
humans perceive a "rainbow" color (monochromatic)
390 nm 750 nm [Wikipedia: Visible spectrum]

 If all wavelengths are equally distributed, humans

perceive gray, ranging from black to white (achromatic)
 Colors "mixed from rainbow
colors" are chromatic
This spectrum corresponds to a ripe brown
banana under white light (reflectance).
[Akenine-Möller et al.]
University of Freiburg – Computer Science Department – 42
Spectral Power Distribution / Reflectance

 A spectrum can describe, e.g.,

 The wavelength distribution within flux characterize the color of light radiation

 The reflectance or absorbance of flux at surfaces reflectance

[Pharr, Humphreys]

Spectral power distribution

Reflectance of lemon skin
of a light source
University of Freiburg – Computer Science Department – 43
Representing a Spectrum
the shape of this function represents some color

 Spectrum

for computation we

 Uniform samples, e.g.

discretize the spectrum

mostly uniform samples are considered

 Non-uniform samples, e.g.,

University of Freiburg – Computer Science Department – 44

Flux vs. Spectral Flux
 Color (spectrum) is typically represented with
(RGB values, spectral flux values)
 Raytracing concepts are described with (flux)
 Can be flux
 Can also be interpreted as a spectral flux vector
 E.g., typically refers to
scalar value
scalar values can also be given as vector values

vector value

University of Freiburg – Computer Science Department – 45

Color Perception
 Perceived color is the radiation spectrum weighted
with absorbance spectra (sensitivity) of the eye
Photopic vision during daylight Scotopic vision during night
cones - daylight
rods - night

Radiation spectrum and absorbance Radiation spectrum and absorbance

spectra of human cone cells (Zapfen). spectrum of human rod cells (Stäbchen).
Cone cells absorb (are sensitive to) Rod cells absorb a wider range of visible
blue, green, red radiation. radiation.
University of Freiburg – Computer Science Department – 46
Color Perception
 Basis functions map from infinite-dimensional space
to low-dimensional space (3D in daylight, 1D at night)
Photopic vision during daylight Scotopic vision during night

x(), y(), z() are the absorbance i() is the absorbance

spectra of human cone cells. spectrum of a rod cell.

 In daylight, three cone signals

are interpreted by the brain as color
University of Freiburg – Computer Science Department – 47
Color Perception
 Is a complex phenomenon

A and B are of the same Perception (partially) adapts

color / brightness. to changing illumination.
A and B are of
the same color.
Wikipedia: Color constancy
University of Freiburg – Computer Science Department – 48

CIE XYZ Color Space
 Proposed by the International Commission
on Illumination CIE in 1931
 Motivated by trichromacy model
 Three cone types  Three signals / numbers for a color
 Spectrum is converted to
with color-matching functions x(), y(), z().

 Color-matching functions x(), y(), z() have been

experimentally estimated to map all perceivable
colors to values in the range from 0 to 1
University of Freiburg – Computer Science Department – 49

CIE xy Chromaticity Diagram
 XYZ represents color and
brightness / luminance
 Two values are sufficient
to represent color

 Monochromatic colors are

on the boundary
 The center is achromatic
[Wikipedia: CIE 1931 color space]

University of Freiburg – Computer Science Department – 50


CIE RGB Color Space
 RGB color space

[Wikipedia:
CIE 1931
color space]

 Spectrum of is converted to
given the color-matching functions
r,g,b of lamda are primary colors and might vary based on the display.

 The color-matching functions consider the spectra

of real display primaries (e.g. LED, LCD, plasma cells)
University of Freiburg – Computer Science Department – 51
CIE RGB Color Space
 Different sets of primary colors result in different sets
of color-matching functions

[Pharr, Humphreys]

Spectra of red, green, blue Spectra of red, green, blue

for an LCD display for an LED display

University of Freiburg – Computer Science Department – 52

Conversion XYZ / RGB
 Depends on the particular set of spectra
of the primary display colors
 E.g., sRGB for HDTV
Negative matrix coefficients
indicate that XYZ values could
result in negative RGB values,
i.e. not all perceivable colors
can be represented / generated
with RGB.

University of Freiburg – Computer Science Department – 53


Display Devices colors change wrt the display devices

 xy chromaticity diagram
 Three display / primary colors
 Diagram indicates an example
 Can only reproduce colors
within the spanned triangle
(gamut)
 Colors outside the gamut [Akenine-Möller
et al.]

are not properly displayed

on the respective monitor
University of Freiburg – Computer Science Department – 54

RGB Color Space
 Three primaries: red, green, blue

Blue (0,0,1) Cyan (0,1,1)

Magenta (1,0,1)
White (1,1,1)

Black (0,0,0)
Green (0,1,0)

Red (1,0,0) Yellow (1,1,0)

University of Freiburg – Computer Science Department – 55

RGB Color Space - Flux
 Light source color
 E.g., yellow light (1, 1, 0)
 Emits a spectrum with maximum red and green components
 The spectrum does not contain any blue
 The RGB values describe the amount of the
respective color component in the emitted light

University of Freiburg – Computer Science Department – 56

RGB Color Space - Surfaces
 Surface color / reflectance
 E.g., yellow object (1, 1, 0)
 Perfectly reflects red and green components
of the incoming light
 Perfectly absorbs the blue component of the incoming light
 The RGB values describe how much of the respective
incoming color component is reflected ("one minus
value" describes how much is absorbed)

University of Freiburg – Computer Science Department – 57

Summary
 Distribution of wavelengths within the perceived radiance
is referred to as spectral power distribution or spectrum
 Spectra are weighted with absorption spectra
of the eye and perceived as colors
 Three cone types for daylight vision motivate XYZ space
 XYZ space can represent all perceivable colors
 RGB space represents displayable colors
 Colors of display devices are restricted to a gamut
that does not contain all perceivable colors
 Ray tracers can work with arbitrary representations
 Conversion to RGB for display purposes
University of Freiburg – Computer Science Department – 58
