Color Management with ICC Profiles: Can’t Live without It so Learn to Live with It

Paul D Fleming and Abhay Sharma

Department of Paper and Printing Science and Engineering

Western Michigan University
Kalamazoo, Michigan 49008
Abstract
Managing and controlling color, from scanner to monitor to output devices, are
serious issues for the printing and imaging industries. Accurate color control is
vital for predictable quality of final product, whether printed or displayed. Costs of
controlling color are significant. Printing has become much more science than art
and any company, especially those involved in printing and other media, must
have a presence on the World Wide Web. Thus, controlling color between input,
display and print is more important than ever. As a result, spectrophotometers
and colorimeters are now necessary tools throughout the industry. Color
management systems have been developed to utilize these measurement
devices to characterize the color of input, display and output devices. These
hardware and software tools have not yet solved all problems of color
reproduction, but have made possible quantification, which is the essential
ingredient for controlling quality. Here we review color management, ICC profiling
and the important players in this technology.

1
 Introduction
The issue of managing and controlling color, from scanners to monitors to digital
printers/proofers and finally to printing presses, is a serious one for the printing
and imaging industries. Accurate color control is vital in order to have a
predictable quality of final product, whether it be printed on a substrate or
displayed in an image editing program or a Web browser, and ultimately to
satisfy customer expectations. The costs involved in controlling color for the
designer/publisher/printer are both soft (time spent by employees during pre
publication, hard or soft proofing, film or plate output for printing presses, and on
press), and hard (wasted ink and substrate used in makeready, as well as time
wasted in jobs that do not meet final customer specifications).
Great strides have been made in the printing industry in recent years toward
applying scientific methods in the pressroom, as well as in the prepress area.
Printing has become much more science than art and every company, especially
printing and media companies, must have a presence on the World Wide Web.
This makes controlling agreement between displayed and printed color more
important than ever.
The advent of digital video and computer-generated animation (Figure 1) has
further compounded the problem. Accurately matching color, between computer
CRT and LCD displays, Analog monitors, Cinematic projection, companion books
and wearing apparel, present an apparently insurmountable challenge. To meet
this challenge, standards have been established and Color Management
methods have been developed. Essential to utilizing these methods are
spectrophotometers and colorimeters, which were virtually unheard of by printers
twenty years ago. These have now become necessary tools in all media
industries (1-4).

Figure 1. A snapshot from Toy Story II, a computer generated animation from Disney and Pixar.
To address the issues of accurate color control throughout the production
process, color management (3,4) systems have been developed. The International

2
Color Consortium (ICC) was formed in 1993 by Adobe, Agfa, Apple, Kodak,
Microsoft, Sun Microsystems and Silicon Graphics (3-6) to define the standards for
color device characterization. This device characterization is presented in terms
of specially formatted files, which have come to be called profiles. The latest
version (7) of the ICC specification has been reviewed recently in this magazine
by McDowell (8).
Unfortunately, the use of color management systems has not yet solved all of the
problems of color reproduction. However, it has made possible the quantification
of problems. As always in quality control, with quantification comes the ability to
control and, with control, quality management becomes possible.
Some Important Issues

A large part of the problem of handling color lies in the inherent differences
between the mechanisms by which different input, display and output devices
perceive color (9-13). Computer displays, scanners and digital cameras are
generally based on the Additive Color Theory (14-16) and are represented in terms
of differing amounts of Red, Green and Blue (RGB). On the other hand, printing
ink on substrate is based on the Subtractive Color Theory (14,15,17), which
generally employs differing levels of Cyan, Magenta, Yellow and Black (CMYK).
The problem is further complicated by the fact that the same RGB image looks
different on different monitors, even ones that are nominally identical. Growing
acceptance of Liquid Crystal Display (18) (LCD) monitors, in addition to the
standard Cathode Ray Tube (CRT) monitors, further complicates the situation.
Furthermore, the two color models span different portions, or gamuts, of the
visible color space (14-17). Some colors representable in the RGB space cannot be
printed in CMYK space, and vice versa. In particular, highly saturated primary
colors are readily displayed, but cause serious problems in printing.
Ambiguities in handling color start first when electronically inputting color into the
workflow. Whether using scanners or digital cameras, each device sees color
differently. Even nominally identical devices show slight, but measurable
differences in captured color. Imagine scanning a saturated solid red patch, such
as a corporate color. Scanner A may report (Figure 2) the color in terms of RGB
(Red, Green and Blue) as (250, 0, 3), while Scanner B reports it as (240, 5, 3). A
digital camera may indicate the patch should be (235, 10, 7). Because the CCDs
(Charge Coupled Devices) used in scanners and digital cameras are
fundamentally different than the phosphors used on CRT (Cathode Ray Tube)
displays and filters used for LCD displays, none of the reported colors will be
correct on any monitor.
Likewise, different digital printers, proofers and printing will produce different
looking results when printing the same CMYK values (Figure 3). This results
because these devices employ different printing processes, different inks and
print on different paper.

3
Figure 2. Different scanners report different RGB values for the same scanned original.

Figure 3. The same CMYK value printed on different printers will look different.
In the early days of Color Electronic Prepress systems (CEPS), high-end drum
scanners were used with a single printing press in a closed-loop (3) system
(Figure 4). Highly trained skilled specialists adjusted scanner characteristics for
color separations and halftones based on press characteristics. Using visual
observations and some measurement instruments on carefully chosen test
targets, they were able to achieve good color “matching”. When the company
traded in the old printing press on a new one, the scanner operator had to start
all over again. Color separations intended for an offset press were not correct for
a gravure press and vice versa.
Today everybody has a scanner, computer and a color printer. Electronic images
come from different places including the Internet, digital cameras, computer
generated art and different scanners. They are displayed on different CRT and
LCD monitors on Macintosh and PC computers. They are intended for multiple
purposes; printed on different printers, proofers and presses and to be read on
screen with different applications from CD-ROM, DVD disks and Web pages.
This intertwined network of connections is illustrated in Figure 5. This situation

4
can only be handled by an open color controlled system. The necessity of
accurately handling these different color devices has led to the development of
color management systems(3).

Figure 4. Illustration of a closed-loop color management system.

Figure 5. Illustration of an open Color Management System.

The color managed system is achieved by make use of a universal device
independent color Profile Connection Space (PCS). A special type of computer
file, called a Profile, characterizes the behavior of the device in terms of the PCS
and specifies how to pass into and out of the PCS (Figure 6). The PCS can be

5
thought of as analogous to the Hub city airport used by airlines to route
passengers to and from different destinations. Just as a new city can be added to
the airline’s service area, a new device can be added to the mix by merely
specifying a profile that gets color information into and out of the PCS.

Figure 6. Open Color Management System with central Profile Connection Space (PCS).

Color Spaces and Tolerancing

As stated above, Scanners, Digital Cameras and Monitors use the additive color
theory and the RGB color space. Digital Printers, Proofers and Printing Presses
use the subtractive color theory and the CMYK color space. These are all
examples of device dependent color spaces. As such, they cannot be used as
the PCS. The first device independent color model was developed by the
Commission Internationale de l’Eclairage (CIE) in 1931 (19). This space, called the
tristimulus space is defined by coordinates X, Y and Z. X is a redness coordinate,
Y is a greenness coordinate and Z is a blueness coordinate.
The tristimulus values are very useful, being very easy to directly measure with
colorimeters and spectrophotometers. However, unit variations in tristimulus
values do not scale uniformly with human visual sensitivity. This deficiency led
the CIE to define a new color space, the L*, a*, b* color space in 1976 (20), which
has been called a uniform color space. This space, often called CIELAB, is
equivalent to the tristimulus space. For a given illuminant, the L*, a* and b*
values are given by

6
 L* = 116f(Y/YI) – 16 (1a)
a* = 500[f(X/XI) - f(Y/YI)] (1b)
b* = 200[f(Y/YI) – f(Z/ZI)], (1c)
where XI, YI and ZI are the tristimulus vales of the illuminant and
f(x) = x(1/3), x > (6/29)3 = .008856…
= (29/6)2x/3 + 16/116, x < (6/29)3. (1d)
Being device independent, CIEXYZ or CIELAB are candidates for the PCS. In
fact, profiles generally use either CIEXYZ or CIELAB as the PCS.
As stated above, the CIELAB space more uniformly represents the human
perception. Thus, the difference between two colors is usually reported as the
Cartesian distance in L*a*b* space. This distance is generally denoted by DE,
which is defined by
DE = ÷(L1*-L2*)2+(a1*-a2*)2+(b1*-b2*)2), (2)
for the distance between colors L1*a1*b1* and L2*a2*b2*. Generally, a DE value of
around unity or less means that the two colors are indistinguishable to essentially
all human observers.
Instruments
Measurement devices used to characterize color are densitometers, colorimeters
and spectrophotometers.
Reflection densitometers measure the light reflected from the printed sample.
The reflection density is defined as
D = -log10(R/I), (3)
where R is the intensity of light reflected and I is the intensity of incident light.
Color information can be obtained by determining the density of light passed
through red, green and blue filters.
Colorimeters and spectrophotometers can actually measure CIE coordinates.
Generally, they measure X, Y, Z and calculate the L*, a*, b* values with built in
electronics. In addition, spectrophotometers actually measure the spectrum of
light reflected from or emitted from the sample.
ICC Profiles
In this section, we illustrate how ICC profiles function. First, we consider input
profiles. Suppose that a digital camera, a drum scanner and a professional
flatbed scanner each capture the same red patch discussed earlier. Each device
sees the color differently within its color space as illustrated in Figure 7. The
effect of the profile is to map all three to the same color in the PCS.
To make an input profile a standard target such as the standard IT8.7/1 (for color
transparency scanning) or IT8.7/2 (21) (for color reflective scanning) illustrated in
Figure 8. The different patches in this target have known colorimetric values,
either provided by the manufacturer or custom measured by the user. The profile
making software creates a mapping between device RGB values and L*a*b*
values, using the known values for the patches. Usually, this mapping involves a

7
parameterized fitting function. This function is then used to create a lookup table
that is stored in the profile. The input profile created by this process is then used
to interpret the color of all originals scanned or photographed with the device.

Figure 7. The effect of an input profile is to transform the color in the input device’s color
space to the correct color in the PCS.

Figure 8. Making an input profile using a standard target and using it to scan an actual
photograph.

8
Monitor profiles and output profiles are constructed by measuring colors
displayed or printed. For monitor profiles, a colorimeter, such as the X-Rite
Monitor Optimizer, or a spectrophotometer, such as a GretagMacbeth
Spectrolino, is attached to the screen by suction cups (Figure 9). A set of 40-50
RGB values is displayed and the colorimetric values are recorded. These are
used to construct the monitor profile. Monitor profiles are generally characterized
by white point color temperature, gamma (g) value and tristimulus values for the
Red, Green and Blue primary colors. The white point is interpreted as the
temperature of a perfect Planckian Black body radiator (22) that produces the
same relative tristimulus values (the Y value is usually normalized to 100).
Generally color temperatures of 5000 K (D50) and 6500 K (D65) are used to
simulate “natural” daylight under different assumed conditions. The g value is an
exponent that is used to characterize the nonlinear response of phosphor
brightness to electron energy (voltage) in a CRT monitor.

Figure 9. X-Rite DTP92 Monitor Optimizer.

For output profiles, it is necessary to print a file of know CMYK values, such as
the IT8.7/3 (23) or the newer IT8.7/4 (24) (Figure 10). The printed patches are then
measured with a colorimeter or spectrophotometer, such as the GretagMacbeth
Spectroscan. Much like the case with input profiles, a mapping between CMYK
values and L*a*b* values is generated and used to populate lookup tables in the
profile. The mapping is more complicated in this case, because in order for it to
be one to one and invertible, some sort of GCR (Gray Component
Replacement)(25,26) or UCR (UnderColor Removal) (25,27) prescription must be
specified to make the effective CMYK space a function only of three variables.
Specifying the L*a*b* value for a given CMYK value is no problem, because that
is what is specifically measured. However, specifying the CMYK value for a given
L*a*b* requires a constraint on the CMYK values. It is important that the output
profile be reversible, so that transforming between CMYK and L*a*b* and then
back again to CMYK yields the same (or equivalent) color values. This feature is
crucial for any digital proofing, hard or soft (on monitor). Sometimes, an
annealing process must be applied to assure this reversibility (28).

9
Numerous utilities are available to view the contents of ICC profiles. ColorSync
Profile Inspector (written in 1994 by Steve Swen and Friends) for Classic
Macintosh Operating Systems, ProfileInspector for Mac OS X Operating Systems
and ICCinspect (from Iphoto), for Microsoft Windows Operating Systems, allow
users to view the contents of ICC profile tags. These are very useful for
interpreting the behavior of profiles. The ColorSync Utility, distributed with Mac
OS X, displays the tags present in a profile, but not the content of the tags. It
does, however, have some useful features including Profile First Aid and Device
association with profiles. Ben Griffin’s Profile Editor (written in 1995) actually
enables editing of some profile tags under Mac classic operating systems.

Figure 10. The new IT8.7/4 output target.

Gamut and Rendering Intent

When working in a color managed workflow with ICC profiles, it is important to
remember the “Three Cs” of Color Management (3). These are calibration,
characterization and conversion. Calibration is the process of ensuring that the
device in question conforms to a specified state or condition. The profiling
process characterizes the color behavior of the properly calibrated device.
Conversion is what is needed for applications to convert between different device
color spaces. This is required in order to display or print scanned images or to
preview (soft proof) what the printed image will look like on a calibrated and
characterized output device. Varying degrees of support for these conversions
are available in Photoshop, InDesign, QuarkXPress, Illustrator, FreeHand and
PageMaker. Of these, Photoshop 6 and 7 are the most advanced and the easiest
to manipulate and probe. In addition, there is some degree of support at the
operating system level from Mac OS 9.2.2 (or earlier), Mac OS X and Windows
98/Me/2000/Xp.

10
The conversion process also depends intimately on the Color Management
Module (CMM) (3) of the system. The CMM accepts color data and translates
them to another color space using the device profiles. It is like a language or
dialect translator that interprets the profile information. Different CMMs will
produce different results, even with the same profiles (29, 30). CMMs are available
from Apple, Adobe, Heidelberg, Kodak and X-Rite. Currently, only the Apple
CMM is available for Mac OS X.
In order for the CMM and applications to convert between device spaces, it is
necessary to recognize that different devices sample different volumes, or
gamuts (31), of the full human visible color space. Thus, it is necessary to
determine how colors that are outside the color gamut of one device are mapped
to the color space of that device. This process is generally called gamut mapping
and the strategies for gamut mapping are referred to as Rendering Intents. The
latest ICC specification (7), version 4.0.0, defines four rendering intents;
perceptual, media relative colorimetric, saturation and ICC absolute colorimetric.
These are illustrated in Figure 11.
Perceptual Rendering – According to the ICC, the exact gamut mapping for
perceptual intent is vendor specific. It generally involves compromises, such as
trading off preservation of contrast in order to preserve detail throughout the tonal
range. It is most useful for general reproduction of images, particularly pictorial or
photographic-type images.
Media Relative Colorimetric – According to the specification, this intent re-scales
the in-gamut, chromatically adapted tristimulus values such that the white point of
the actual medium is mapped to the PCS white point (for either input or output). It
is especially useful for colors that have already been mapped to a medium with a
smaller gamut than the reference medium (and therefore need no further
compression).
ICC Absolute Colorimetry - The ICC specifies that for this intent, the
chromatically adapted tristimulus values of the in-gamut colors are unchanged.
This is useful for spot colors and when simulating one medium on another
(proofing). Note that this definition of ICC-absolute colorimetry is actually called
“relative colorimetry” in CIE terminology, since the data have been normalized
relative to the perfect diffuser viewed under the same illumination source as the
sample. Profiles do not contain a separate transform for the ICC-absolute
colorimetric intent. When this intent is needed, it is generated using the media-
relative colorimetric intent and scaling the PCS values by the ratio of the
destination profile white point values to the source profile white point values.
Saturation Rendering – As noted by the ICC the exact gamut mapping of the
saturation intent is vendor specific and involves compromises such as trading off
preservation of hue in order to preserve the vividness of pure colors. It is useful
for images, which contain objects such as charts or diagrams.
Since the Perceptual and Saturation rendering intents are vendor specific, only
the colorimetric rendering intents for a given device can be meaningfully
compared between different profile making vendors. Thus, in the following
section, where we discuss profile quality, we employ absolute colorimetric

11
rendering.

Figure 11. Illustration of different rendering intents for output devices. The outline in the
interior represents the printing gamut of a SWOP CMYK device printed on the “standard”
coated paper.

Measures of Profile Quality

Recently (29,30), we introduced some measures for evaluating the quality of ICC
profiles. This work was undertaken because no agreed upon method of
determination of profile accuracy was available. The distributed document (30),
called the WMU Profiling Review, provides an independent, scientific
assessment of the accuracy of profiles. In the current version of the review,

12
results of 8 packages used to make input, printer and monitor profiles are
compared. Updates to the WMU Profiling Review occur bi-annually on January 1
and July 1. Any vendor that would like to be included in future revisions is
welcome to contact us (30).
It is important to have a quality measure for ICC profiles, because this indicates
how well a device has been characterized and, therefore, how accurate its color
is likely to be in a color managed workflow. It is important for software vendors to
have a merit figure available and for the industry to agree on how the figure is
calculated. Some vendors quote a DE merit figure, and often programs will write
out a file with statistics following profile generation. However, there is no
indication to tell us how these figures were calculated and whether everybody is
measuring the same thing in the same way.
The aim of the WMU Profiling Review is to establish some baseline assessment
for ICC profiles and thus assist user choice, raise the standard for profiling
software and promote the wider acceptance of ICC color management.
It is inevitable in a survey of this type that some vendors appeared better than
others did. However, this should not be taken as an endorsement of any
particular product or manufacturer. In addition, whilst accuracy is important other
features such as visualization tools and profile editing should be considered
when making a purchasing decision.
The assessment of ICC profiles and color reproduction is a complex issue
involving everything from color science, psychophysics and image analysis to
‘preferred’ reproduction styles. The approach adopted in our work is to evaluate
the accuracy of profiles using the absolute colorimetric intent. This does not
provide an all-encompassing result but does provide an indicative set of metric
figures that can be used to make valid cross-vendor comparisons.
Photoshop 6.0.1, Mac OS 9.2.2 and ColorSync 3.0.4 were used throughout this
testing.
Scanner Profiles
Agfa, FujiFilm and Kodak IT8.7/2 (23) reflection test targets were scanned on a
Umax Astra 4000u scanner and profiles were made using normal procedures in
the different profiling packages. The methods are detailed in references 29 and
30.
The accuracy obtained for each vendor’s program is shown in the table.
Manufacturers are ranked in order. Thus, ScanOpen provided the best overall
result, whilst the generic profile was worst. How do we interpret these results? A
lower ∆E number is preferable. Profiles with a ∆E less than 2.0 are very accurate
input profiles. The error in these profiles is so small that the differences are
probably not even noticeable. For this scanner, the generic profile with a DE of
nearly 30 was very poor. Note that just because the generic scanner profile is
poor, this does not mean that the Umax scanner is poor. In fact, the scanner is
remarkably good for the price.

13
 Agfa FujiFilm Kodak Final 1
Scanner profile quality Price
IT8.7/2 Chart IT8.7/2 Chart IT8.7/2 Chart result

Mean (Max) Mean (Max) Mean (Max) Average

∆E ∆E ∆E ∆E

ScanOpen 4.0.5 0.69 (16.08) 0.63 (2.51) 0.69 (7.53) 0.67 Bundled
2
ScanOpen 2.1.0 0.99 (15.56) 1.12 (3.28) 0.96 (4.99) 1.02

GretagMacbeth
0.85 (2.87) 0.99 (10.13) 1.23 (4.12) 1.02 $3,000
ProfileMaker 4.0

Monaco Profiler 4.0 1.19 (9.95) 0.92 (4.70) 1.19 (7.10) 1.10 $4,250
2
Monaco Profiler 3.2 4.39(15.00) 5.04 (8.25) 4.79 (11.35) 4.75

FujiFilm ColourKit 2.3 1.15 (3.72) 1.23 (4.53) 1.43 (3.53) 1.27 $3,600
2
FujiFilm ColourKit 2.2 1.17 (3.98) 1.25 (4.53) 1.42 (3.66) 1.28

ColorBlind 4.2 1.53 (7.13) 1.95 (8.95) 1.37 (6.92) 1.62 $4,545

ColorSynergy 4.5 2.74 (11.17) 3.05 (10.09) 2.79 (10.43) 2.86 $1,495

Kodak Colorflow 2.2.1 6.86 (36.33) 11.20 (59.27) 6.75 (34.40) 8.27 $2,450

Generic Umax
3 29.80 (44.55) 28.93 (42.03) 29.38 (46.67) 29.37 Free
scanner profile
1
The price quoted is for the full product that will make input, monitor and printer profiles.
2
The ranking of the profile and the price is based on the latest release of the software. Older
versions of the software are included for reference only.
3
The generic profile was provided as part of the Umax scanner driver, Umax VistaScan 3.5.4.
Table 1. Results for scanner profiles.
Display Profiles
Eight monitor profiles were tested to see if they were able to achieve a requested
gamma and a requested white point. A Mitsubishi Diamond Plus CRT was used
on a Power Mac G4. Monitor profiles were made using different measuring
instruments as shown in the table and if offered a choice, the user requested a g
of 1.8 and white point of D65. After each profile was made it was selected as the
system display profile. Using Photoshop, a series of patches were displayed on
the monitor so that the actual white point and actual g of the display could be
measured.
Macintosh monitor profiles are distinguished by the use of a ‘vcgt’ tag that is used
to store the correction that converts the factory setting into the user desired
settings. Vcgt stands for video card gamma tag and has been part of the MacOS
since ColorSync 2.5. In terms of the g value, the monitor profiling results fell in to

14
two camps. Profiles with a vcgt tag produced a g of 1.8 as requested by the user,
whilst profiles without a vcgt produced a g of around 3.0, which is the inherent
gamma of the display. The difference between the color of the profiled monitor
and the desired white point of D65 was calculated for each vendor and is shown
in the table. The ∆E on a monitor should not be considered as critical as that for
a scanner. A monitor profile can be colorimetrically accurate for any white point,
even if its white point isn’t exactly the same as the target whit point.
Achieved
gamma ∆E difference in
Measuring
Monitor profile quality white point from vcgt tag
instrument (Target was a target of D65
1.8)

GretagMacbeth GretagMacbeth
1.81 0.58 Yes
ProfileMaker 4.0 Spectrolino

Monaco Profiler 4.0 Spectrolino 1.85 0.68 Yes

FujiFilm ColourKit 2.2 X-Rite DTP92 1.77 3.35 Yes

Mitsubishi monitor
None 2.99 3.83 No
generic profile

Kodak Colorflow 2.2.1 None 2.98 3.97 No

ColorSynergy 4.5 Spectrolino 2.99 4.07 No

X-Rite Colorshop 2.6.2 DTP92 1.76 5.81 Yes

ColorSync Monitor
Visual 2.03 12.76 Yes
Calibrator 3.0.4

Table 2. Results for Display profiles.

Printer Profiles
Printer profiles contain three all three rendering intents – perceptual, colorimetric
and saturation. Each intent has a forward (Profile Connection Space to Device)
and reverse (Device to Profile Connection Space) look-up table. In this test, the
forward and reverse parts of the absolute colorimetric intent were evaluated. The
methods for testing the forward and reverse modes are detailed in the reference.
Whilst photographic images are normally processed using the perceptual intent,
the colorimetric intent is used during the facsimile reproduction of images, during
soft proofing, when images are evaluated on a monitor and during proofing when
press images are ‘returned’ to the PCS and printed on a proofing device. The
colorimetric intent may also be used when legacy CMYK images are opened.
Therefore, although the colorimetric intent is not normally used to process

15
photographic images it is used in number of very significant ICC workflows and
as such is a good indicator to use for profile accuracy evaluation.

1 Epson Stylus Pro 3

Printer profile quality HP Designjet 20ps 2 Final
5000

Mean ∆E Range Mean ∆E Range Mean ∆E Range

GretagMacbeth
2-4 2-4 2-4
ProfileMaker 4.0

FujiFilm ColourKit 2.3 2-4 2-4 2-4

PrintOpen 4.0.5 2-4 2-4 2-4

Monaco Profiler 4.0 2-4 2-4 2-4

Kodak Colorflow 2.2 4-6 4-6 4-6

ColorSynergy 4.5 6-8 10-12 8-10

4
Generic profile 12-14 12-14 12-14

1
HP Designjet using the HP RIP v 1.1 SP2 with HP Premium Photo Paper Glossy, batch C6039A
2
Epson Stylus Pro5000 using Fiery RIP SPv1.3 in CMYK mode with Epson Photo Paper, batch
Y1JL0U744
3
The final result is an average of four mean ∆E results.
4
The HP generic profile was part of the HP driver and was called HPGb2_out. The Epson generic
profile was downloaded from [Link]/de/[Link].
Table 3. Results for Printer profiles.

Output profiles were generated and tested for HP Designjet 20ps and Epson
Stylus Pro 5000 inkjet printers. The results are summarized in Table 3. It is
reassuring to note that, in many cases, vendors achieve similar results across
both printers despite each device having a different RIP and print process. The
results produced by the first group of vendors, with an average ∆E of 3.0 – 5.0,
are very good. In general, the forward process shows greater error than the
reverse. It is possible to speculate that the forward test involves printing and
measuring and therefore is more subject to instrument variability. The number of
grid points used in the profile may also be an issue if, for example, the forward
look-up table has more nodes in the look-up table while the reverse table is more
sparsely populated.

16
 Summary
We have reviewed fundamentals of color management and ICC profiles. As we
have seen, such a system is needed for today’s production workflow with multi
source digital color content and cross-purposed output. It provides needed
flexibility and is not as dependent on the skills of the users as were previous
closed-loop systems. It seems complicated at times, but color management is the
only way to provide the controls needed to have critical colors viewed faithfully at
all stages and for all purposes.
Managing color is essential for all methods of visual communication, especially
color printing, gravure in particular. Obtaining an ICC output profile for the
Gravure Press is an essential component of fingerprinting a (gravure) press (31).
References
1. C. Bak, “Color: From Art To Science”, In-Plant Graphics, Oct. 1997, p33
2. H. Tolliver, “Quantifying Color”, American Inkmaker, Oct. 1997, p16
3. R. M. Adams and J. B. Weisberg, The GATF Practical Guide to Color
Management, GATFPress, Second Edition, 2000.
4. S. Brues, Postscriptum on Color Management, GretagMacbeth, Second
Edition, 2000.
5. P. Green, Understanding Digital Color, 2nd Edition, GATFPress, 1999,
Chapter 7.
6. ICC home page, [Link]
7. ICC, Specification ICC.1:2001-12, File Formats for Color Profiles (version
4.0.0), downloadable from [Link]
8. D. McDowell, “Color Management: What's New with the ICC?”, Gravure,
April 2002, 50-54.
9. K. C. Williamson, “A Soft Proofing Study”, GATFWorld, Nov./Dec. 1997, p23
10. M. Samworth, “Toward True Color Management”, Part 1-3, Flexo, Sept.,
Oct., Nov. 1997, p74, p64, p57
11. D. C. Rich, “Today’s Color Management Systems: The Application of
Advanced Technology To the Science of Color”, American Inkmaker, Oct.
1997, p33
12. J. Mertens, “Color Management Using CIELab”, Flexo, Oct. 1997, p60
13. X. Zhang and B. A. Wandell, “A spatial extension of CIELAB for digital color-
image reproduction”, JSID, 5/1, 61 (1997)
14. P. Green, Understanding Digital Color, 2nd Edition, GATFPress, 1999,
Chapter 1.

17
15. P. Pfiffner and B. Fraser, How Desktop Publishing Works, Ziff-Davis Press,
1994, Chapter 11.
16. T. E. Schildgen, Pocket Guide to Color with digital applications, Delmar,
1998, chapter 3.
17. T. E. Schildgen, Pocket Guide to Color with digital applications, Delmar,
1998, chapter 4.
18. J. Widman, Ed., “Flat-Panel Power”, Publish Extra, Summer, 1998.
19. CIE Proceedings 1931, Cambridge University Press, Cambridge, 1932.
20. CIE, Recommendations on Uniform Color Spaces, Color-Difference
Equations, Psychometric Color Terms, Supplement No. 2 of CIE Publ. No. 15
(E-1.3.1) 1971, Bureau Ventral de la CIE, Paris 1978.
21. ANSI IT8.7/2-1993 (ISO 12641), Graphic Technology - Color Reflection
Target for Input Scanner Calibration (see
[Link] for ordering information), P. J.
Groff and Frank V. Kanonik, Test Images for Printing, GATFPress, 1990, p
13.
22. M. Planck, Verh. d. deutsch phys. Ges., 2, 202, 237, 1900. Ann. d. Physik,
(4), 4, 553 (1901).
23. ANSI IT8.7/3 Graphic technology – Input data for characterization of 4-color
process printing, also published as ISO 12642:1996 (see
[Link] for ordering information).
24. ANSI IT8.7/4-200X, Input data for characterization of 4-color process printing
of packaging materials, CGATS STF2 N 088. (Downloadable from
[Link]
25. J. A. C. Yule, “Theory of Subtractive Color Photography III; Four-Color
Process and the Black Printer”, J. Opt. Soc. Am., 30, 322-331 (1940).
26. G. G. Field, “Color Variability Associated with Printing GCR Color
Separations”, TAGA Proceedings, 145-157, (1986).
27. J. A. C. Yule and F. R. Clapper, “Undercolor Removal Requirements of the
Black Printer”, Graphic Arts Mon., 32, No. 5, 153-166, (1960), J. A. C. Yule,
“Special Methods of Undercolor Removal”, Graphic Arts Mon., 33, No. 12,
42-46 (1961).
28. A. Sharma, M.P. Gouch, and D.N. Rughani, “Generation of an ICC profile
from a proprietary style file”, J. Imag. Sci. Tech, 46, 26. 2002
29. A Sharma and P D Fleming, “Evaluating the Quality of Commercial ICC Color
Management Software”, Presented at TAGA Annual Technical Conference,

18
 North Carolina, April 11-14, 2002. To be included in the Proceedings of the
Conference.
30. A Sharma and P D Fleming, “Measuring the accuracy of ICC profiles and
color management software”, WMU Profiling Review 2.0, July 1, 2002. This
can be obtained by sending an e-mail to [Link]@[Link] with
“Subscribe WMU Profiling Review” in the subject line.
31. M. R. Pointer, “The Gamut of Real Surface Colors”, Color Research and
Application, 5, no. 3, 145-155 (1980).
32. R. Wiesmann, "Fingerprinting: The Rising Phoenix", Six part series, Gravure,
June 2002-April 2002. See also [Link]

19