Scanning Guide
Scanning Guide
Scanning Guide
Introduction
RGB scanning offers many advantages over scanning direct to CMYK, provided
you have high quality custom ICC profiles for your scanner, monitor, and output
device. Note that today's scanners usually offer "professional" and "novice"
modes, and the chosen mode may affect how the profile should be created.
Creating a professional scanner profile begins by scanning an IT8 or HCT target to
a ‘raw’ RGB file, with no color adjustments or processing applied at the scanner. A
good raw RGB scan should maintain all the tonal values and color variations in the
original, but should look a little dull or de-saturated when first opened in an
application like Adobe Photoshop™. When the scanner profile is assigned to the
raw image, a very close match to the original should be seen, provided the monitor
profile and viewing conditions are good.
A novice-style scanner profile requires that any automated exposure or contrast
functions are ON when the target is scanned, but that all color management
functions are temporarily OFF, to produce an "automated-raw" image
The problem with some scanners is that their default RGB mode does not give raw
RGB, but rather ‘color-managed' RGB, that is bright and saturated but may ‘clip’
or ‘plug’ some colors. This paper offers some general advice for getting truly raw
RGB from virtually any scanner and gives detailed instructions for a few specific
scanners. Some hints are given for optimizing the quality of an ICC scanner
profile, and an optimized RGB workflow is outlined.
CONTENTS
Introduction 1
RGB scanner setup 3
Basic goals 3
White and black density settings 3
Measuring scanner tone response curve 3
Ideal tone curve 4
8 bits vs 16 bits per channel (bit depth) 7
Benefits of high-bit scanning 7
Disadvantages of high-bit scanning 8
How many bits are enough? 9
Optical flare 20
Problems caused by optical flare 20
Minimizing flare BEFORE scanning 21
Minimizing flare AFTER scanning 21
Basic goals
The basic goals of RGB scanner setup are to;
• Set white and black sensitivities to the widest density range possible for the material being scanned.
• Create an ideal tone response curve for optimum distribution of digital values.
• Eliminate any built-in color correction or color management (at least while profiling).
• Eliminate any automated functions, such as finding white and black, cast removal, etc. (except if creating a
profile for a "novice" workflow).
1
For most accuracy, measure your own scale on an accurately calibrated densitometer to get the density values,
otherwise divide the step numbers of the ST-34 by ten to get the approximate density of each step.
Blue (upper) line: Optimum tone curve for CMS scanning Red (lower) line: Typical CMYK drum-scanner curve
IT8 target
HCT target
2
To read the L* values, open the reference file in Microsoft Excel or a text editing program like BBEdit.
Is it really 16?
In most cases the useful bit depth will be somewhat less than 16, typically 10 to 14 bits per channel,
depending on the scanner’s optics, signal processing, and ADC (Analog to Digital Converter). To clarify this,
the late Bruce Fraser (co-author of Real World Photoshop and Real World Color Management) coined the
phrase “High-bit” to refer to images with anything greater than 8 bits-per color per pixel.
Because computer hardware is based on multiples of 8 bits, the next available file size after 8 bits is 16 bits
per channel, so this is what is delivered by the software in High-bit mode. If the number of “real” bits is less
than 16, the least significant bit registers are “padded” with zeros, meaning you will get a 16 bit per channel
file, but depending on your scanner it may contain less than 16 bits of useful data per channel.
Post-processing benefits
High bit scanning typically delivers many more levels than most output processes can resolve, hence by some
arguments the extra levels are wasted. However these extra levels can dramatically improve output quality
when significant tone or color edits are necessary. For example, imagine an extremely dark image scanned in
8 bits, with a highlight value of only 63, vs the optimum value of 255. After it is lightened in Photoshop, the
final image will only have 64 levels in its entire tone scale, instead of the normal 256. To make matters
worse, such an image will usually need extreme curve lightening, for example a middle gray (about 127 in a
normal image) may only be 8 in an extremely dark image. Moving 8 to 127 via Curves or Gamma correction
may restore the image to normal lightness, but there will only be 8 levels from mid-tone to black, instead of
the normal 127, hence the image will appear “posterized” in dark shadow areas (see illustration below).
By contrast, if the same original was scanned with 14 true bits per channel, the same highlight area would
start with about 4,048 levels on a 16,383 scale, and the middle-tone area would start with about 514 levels, so
even after extreme curve correction, a full range of smooth tones will be visible without posterization.
with no visible advantage. The following examples proves there really can be a major visible difference.
Quality difference between 16 and 8 bit images after extreme shadow tone expansion
Shadow areas show unwanted “posterization” in the 8 bit image (right) compared to the smooth high-bit image (left)
Here the raw 16 bit scan of an under-exposed transparency (left) was reproduced twice. The middle image
was edited in Photoshop CS in 16 bits. For the right image, the raw 16 bit scan was first converted to 8 bits,
then the same edits were applied. Lighter tone quality is about the same in both, but after correction in
Photoshop, deep shadow areas and dark greens are much smoother in the high-bit (16) version.
(typically 2GB or more) and keep as much scratch disk space available as possible.
If these issues are a problem, do your first round of edits in high-bit RGB, then convert the image to 8 bits for
storage and subsequent edits.
Dragging the black channel to the trash can in the Channels window
• Select Image ... Mode ... RGB Color. (This simply renames the CMY channels as RGB.)
• Set Black density to between 400 (4.00) and 450 (4.50) for 98%
• OPTIONAL: On the Separation Gradation curve, change the output value for density 1.00 from 47% to
55% for a darker basic scan, or make any other desired curve change
• In the USM dialog, switch the first two values (white line and black line) to zero
• TRANSMISSION: Perform a transmission ‘Basic alignment’ on the clear drum, or through the oil and
overlay film, (if oil-mounting)
• REFLECTION: Perform a reflection ‘Basic alignment’ on two layers of super-smooth, high-brightness (97
or higher) Laser-printer paper, or equivalent
• Scan the target and save it as a TIFF CMYK file
• Open the CMYK file in Photoshop and select Window ... Palettes ... Show Channels.
• Drag the black channel from the list into the trash can at the bottom of the palette
• Select File ... Mode ... RGB. (This simply renames the CMY channels as RGB)
• Save the new RGB image under the same name
• CHECK WHITE DENSITY: If R, G or B reads 253 or higher in the white patch of the target, reduce the
White density by 2 (e.g. from 10 to 8) and repeat the scan
• CHECK BLACK DENSITY: If R, G or B reads 5 or lower in the infinite density patch, increase the Black
density value by 10 (e.g. from 400 to 410) or reduce the Black % value (e.g. from 98% to 96%) and repeat
the scan
• SAVE THE FINAL SETUP for future use
IMPORTANT: When using the ‘Strip-K’ setup you must save the scan as TIFF CMYK, not TIFF RGB. If
you save an RGB file direct from the scanner (as implied by the manufacturer’s default RGB setup), the
results will be over-saturated and UNACCEPTABLE for good ICC profiliing.
Normal
Production scanning
• Scan with these settings but always manually check the White and Black density values before scanning,
as most versions of Celsis software have a nasty habit of ‘forgetting’ these vital values.
• Perform an ‘Autobalance’ on the clear drum (or through the oil and overlay film, if oil-mounting.)
In recent Celsis software it is safe to export the file as RGB, however if this is an older Crosfield scanner, or
if there is any doubt, export as CMYK and strip the black channel as described elsewhere.
iQscan Setup dialog (left) and typical endpoints for transparency (middle) and reflection (right)
NOTE: The White, Black and Exposure controls are inter-active and may require several iterations
• Save the End Points with a name like ‘TranEndpts_(date)’ or ‘RefEndPts_(date)’.
• In the Setup Dialog check the End Points and Sharpness settings have the names you just saved.
• Make a high-resolution scan of the HCT (or IT8) scanner target and create an ICC profile.
• Repeat this process for both transparency and reflective targets and save the resulting ICC profiles with
names like ‘iQTran_(date)’ and ‘iQRef_(date)’.
Optical flare
Optical flare is an undesirable by-product of all optical systems including cameras, scanners and even the
human eye. Flare is caused when light scattered by microscopic imperfections on mirrors, lenses and filters in
the optical path, strays unwanted into darker image areas.
Flare in a typical 35mm scanner, with normal gamma (left) and gamma 3.0 (right)
Flare in a drum scanner, with normal gamma (left) and gamma 3.0 (right)
layered PSD file (in case you have to try different settings later) then flatten it and save a copy as a TIFF
file with the letters 'RF' in the name (for 'Reduced-Flare').
• Make a profile from the 'Reduced-Flare' file and compare it to a normal profile. You should see an
improvement in dark detail - more apparent when lightening a dark image in Photoshop.
• If the results are not to your liking, repeat the experiment with different Amount and Radius values.
3
Calibration is variously known as ‘Autobalance’, ‘Zero’, ‘Reset’, ‘Basic Alignment’, etc, on different scanners. On
some scanners this is an automatic function with no user control.
Comparing the ‘infinite density’ and the darkest gray scale RGB values
• Subtract the infinite black RGB values from the RGB values measured in the blackest step of the gray scale
to get three numbers indicating the R, G and B differences.
4
This trick is not necessary with reflective targets.
• In Levels, enter the smallest or average of the R, G and B differences in the top left box and click OK.
Basic rules
The scanner must be set up exactly the same way in production as when the profile target was scanned. Drum
scanners should always be calibrated on the clear drum before each scanning session.
Remember that for the profile to work accurately, the scanner's controls must NOT be adjusted to suit the
output device. Likewise, for maximum scanning efficiency and to eliminate any need for color skills at scan
time, no adjustments should be made to correct for over- or under-exposure or color casts5.
Rendering intent
When converting from RGB to CMYK, a good rule of thumb is to use Perceptual for small-format (35mm
and 6x7) transparencies and Relative with Black Point Compensation ON for larger transparencies.
Reflective prints may be reproduced with Relative with Black Point Compensation ON or OFF, depending
on subject contrast. Note that the two intents may produce different density and contrast curves, depending on
the profiling software used. Experiment to find the settings that work best for you.
5
These edits will be made later in an ICC-aware editing application like Photoshop.
Sharpening in Photoshop
Some traditional drum scanner operators will swear their hardware-based USM is better than Photoshop’s
USM, but in the end, the differences are small and the advantages of previewing USM on a screen display
and being able to change sharpness after scanning outweigh any advantages of on-scanner sharpening.
A strong argument against scanner sharpening is that it reduces the archival value of the raw RGB file. For
example, a halo width that is acceptable at 200% will be unacceptable at 400%. USM applied in the original
scan can lead to a re-scan if the client requests a lower USM amount.