Real World

Color Management,
Second Edition

This page lefi blank. ..

I s Color?
Reflections on Life

"It's not easy being green," sang the velvet voice of Kermit the Frog, perhaps giving us some indication of how the frog felt to be a felt frog. While
none of us may ever know the experience of "being green," it's worth
reflecting (as we are all reflective objects) on the experience of "seeing
green."
You don't have to be a color expert to use color management. But if
you're totally unfamiiiar with the concepts behind the technology, color
management may seem l i e magic. We don't expect you to become as
obsessed with color as we are-indeed, if you want any hope of leading a
normal lie, we advise againstit-but we do recommend that you familiarize yourselfwith the fundamentals we lay out in this chapter.
b

They'll help you understand the problem that color management addresses. The whole business of printing or displaying images that look
like realistic depictions of the things they portray depends on exploiting specific properties of the way humans see color. Color management is just an extension of that effort.

They'll explain some of the terminology you'll encounter when using

color management software--terms like colorimetric, perceptual, and
saturation, for example. A little color theory helps explain these terms
and why we need them.

Real World Color Management, Scrond Edition

b

The strengths and weaknesses of color management are rooted in our

ability (or inability) to quantify human color Vision accurately. If you
understand this, your expectations will be more realistic.

Color theory explains why a colorviewed ina complex scene such asa
photograph looks "different" from the same color viewed in isolation.
Understanding this helps you evaluate your results.

You need to understand the insmmentsyou may use with color management. This chapter explains just what they measure.

Butwe have to 'Fess up to anotherreason forwritingthis chapter: color

is just really darned interesting. While this chapter sets the stage and lays
the foundation for other chapters in this book, we hope it will also spark
your curiosity about something you probably take for granted-your
ability to see colors. If you're intimidated by scientific concepts, don't
worry-we won't bombard you with obscure equations, or insist that you
pass a graduate course in rocket science. It's not absolutely necessary to
understand all of the issues wecoverin this chapterto use color management. But a passing familiarity with these concepts and terms can often
come in handy And you may well come to realize that, although you've
probably done it all your life, in reality"it5 not easy seeing green."

Where Is Color?
If you want to manage color, it helps to first understand just what it is. so
let's start by examining your current definition of color. Depending on
how much you've thought about it-if you're reading this book. you've
probably done so more than most-you may have gone through several
definitions at various times in your life, but they've probably resembled
one of the following statements:

Color is a property of obieds. This is the first and most persistent view
of color. No matter how much we mav have philosophizd about color, we
all still speak of "green apples," "red lights," and "blue suede shoes."

Chapter B What Is Color?

Color is a ptop.r(y dlight.This is the textbookcounterclaim tothe view

of color as a property of objects. Authors of color books and papers love
to stress that "light is color" or "no light, no color."

Color hapin the ob-er.

This concept captures our imagination
when we encounter optical illusions such as afterimages, successive contrast, and others that don't seem to originate in the objects we see. Color
is something that originates in the eye or the brain of the observer.
The correct answer, of course, is a blend of all three. All are partially
true, but you don't have to look far to find examples that show that none
of the three statements, by itself, is a complete description of the experience we call color.
Color is an event that occurs among three participants: a light source,
an object, and an observer. The color event is a sensation evoked in the
observer by the wavelengths of light produced by the light source and
modified by the object. If any of these three things changes, the color
event is different (see Figure I-1)-in plain English, we see a different
color.
We find it interesting that the three ingredients of the color event represent three of the hard sciences: physics, chemistry, and biology. Understanding how light affects color takes us into the physics of color;
ngure1.1

The color event

A eolor event always has three parricipants.

Real World Color Management, Sewid Edition

understanding how objects change light involves the chemistry of surfaces and how their molecules and atoms absorb light energy; and understanding the n a m e of the observer takes us into biology, including
the neurophysiologyof the eye and brain. and the threshold of the nether
regions of psychology. In short, color is a complex phenomenon.
The next sections explore this simple model of the color event in more
detail. We begin with light sources, then move on to objects, and then
spend a bit more time with the subject most dear to you and us, namely
you and us (the obsewersl.

Light and the Color Event

The first participant in the color event is light. The party jusl doesn't get
started until this guest arrives. But all light isn't created equal: the characteristics of the light have a profound effect on our experience of color.
So let's look at the nature of light in more detail.

Photons and Waves

Many a poorphysics smdenthasrelivedthedilemrnafacedby eighteenthcentury scientists as to whether Lightis best modeled as aparticle (theview
held by Sir Isaac Newton) oras a wave (as argued by Christian Hupgens).
Depending on the experiment you do, light behaves sometimes like a particle, sometimes like a wave. The two competing views were eventually
reconciled by the quanhlm theorists like Max Planck and Albert Einstein
into the "wavicle"concept called a photon,
You can imagine a photon as a pulsating packet of energy traveling
through space. Each photon is born and dies with a specific energylevel.
The photon's energy level does not change the speed at which the photon r r a z ~ l c t h r o u g hany given medium, the speed of light is constant
for all photons, regardless of energy level. Instead, the energy level of the
photon determines how fast it pulsates Higher-energy photons pulsate
at higher frequencies. So as these photons all travel together at the same
speed, the photons with higher energy travel shorter distances between
pulses. In other words, they have shorter wazwlengths. Another way to
put it is that every photon has a specific energy lwel, and thus a specific
wavelength-the higher the energy level, the shorter the wavelength (see
Figure 1-21,

The wavelengths of light are at the order of magnitude of nanometers,

or billionths of a meter (abbreviated nm).

The Spectrum
The spectrum refers to the full range of energy levels (wavelengths) that
photons have as they travel through space and time. The part of this spectrum that tickles our eye is a small sliver from about 380 nm to about 700
nm that we call the visible spectrum, or simply, light (see Figure 1-3).

mure 1-P

@f-V-deR

RUIR -a-cLh (dRbce@f-V-deR

IIWWWV\I1/WWWWV\I

Wavelengths

Figure 1-3

The spectrum
long wauelengths flow energy)

short wavelengths (high energy)

x-rays

radar
microwaves

gamma rays

longwave
radio
radio

visible l i g h y

infrared fIR)

Real World Cdor Management, Second Edition

Our eyes respond only to this tiny sliver of the full electromagnetic
spectrum, and they have varying responses to different parts of this sliver-the different wavelengths evoke different sensations of color. So we've
come to associate the different wavelengfhs with dle colors they evoke,
from the reds at the low-energy end (longer wavelengths at about 700
nm) thmugh the oranges, yellows, and greens to the blues and violets at
the high-energy end (shorter wavelengths at about 380 nm). Of course,
there's nothing in the electromagnetic spectrum itself that prevents us
bom namingmore or fewer than sixbands. Newton, for example, labeled
a seventh band, indigo, between the blues and violets. (Many historians
believe that Newton was looking For symmetry with the seven notes of
the musical octave.)
But no matter how many bands you label in the spectrum, the orderreds, oranges, yellows, greens, blues. and violets--1s always the same.
(Fred and Bruce spent early years in a British school system, and were
taught the mnemonic "Richard ofYork Gained Battles in Vain," while in
the US., Chris was introduced to the shange personage of Mr. "ROY G.
BiV") We could reverse the order, and list them from shortest to longest
wavelength (and hence from highest to lowest energy and frequenq-the
lowerthe energy thelowerthe frequency, and the longer the wavelength),
but green would always lie between blue and yellow, and orange would
always lie behveen yellow and red.
In the graphic arts, we're mainly concerned with visible light, but we
sometimes have to pay attention to those parts of the spectrum that lie
just outside the visible range. The wavelengths that are slightly longer
than red light occupy the inpared m)region (which means, literally,
"below red"). IR often creates problems for digital cameras, because the
CCD (charge-coupled-device) arrays used in digital cameras to detect
light are also highly sensitive to infrared, somost digital cameras include
an IR filter either on the chip or on the lens.
At the other end, just abwe the last visible violets, the range of highenergy (short-wavelength) photons known as the ttltrauioler (W)
region (literally,"beyond violet"] also raises some concerns. For example,
paper and ink manufacturers (like laundry detergent manufacturers)
often add W brighteners to make an extra-white paper or exha-bright
ink. The brighteners absorb non-visible photons with W wavelengths.
and re-emit photons in the visible spectrum-+ phenomenon known as

OHpkr 1: what b CdorI

fluorescence.This practice creates problems for some measuring instruments, because they see the paper or inkdifferentlyfrom the way our eyes
do. We address these issues in Chapters 5 and 8.

Spectral Curves
Other than the incredibly saturated greens and reds emitted by lasers,
you'll rarely see light composed of photons of all the same wavelength
(whatthe scientists call monochromaticlight).Instead, almost all the light
you see consists of a blend of photons of many wavelengths. The actual
color you see is determined by the specific blend of wavelengths-the
spectral energy-that reaches your eye.
Pure white light contains equal amounts of photons at all the visible
wavelengths. Light from a green object contains few short-wavelength
(high-energy) photons, and fewlong-wavelength (lowenergy1p h o t o n s
but is comprised mostly of medium-wavelength photons. Light coming
from apatch of magenta inkcontains photons in the short and long wavelengths, but few in the middle of the visible spectrum.
AU of thesespectral energies can be represented by a diagram called the
spectral curueof thelight reflected by the object (see Figure 1-41.

Spectral curves

white object'

406 m

sad nm

600 nn

wavelength
Spectral curves of three objecu

Light Sources
A lightsource is just something that emits large quantities of photons in
the visible spectrum.Just as with objects, we can draw the spectral curve
of the light energy emitted by the light source at each wavelength (see
Figure 1-5).

Light sources

400 nm

500 nm

60;

nm

700

wavelength

We care about several main kinds of light sources:

Blackbody radiators are light sources whose photons are purely the
result of thermal energy given off by atoms. Lightbulbsand stars such
as our sun are examples of near-perfect blackbodies.The wavelength
composition (i.e., "color") of radiation emitted by a blackbody radiator
depends only on its temperature, and not what it's made of. So we use
color temperatureas a way of describing the overall "color" of a light
source. (See the sidebar 'The Color ofWhite" later in this chapter.)

Dayiight is the result of our most familiar blackbody radiator, the

sun, and an enormous filter we call the atmosphere. It's probablythe
most important of all light sources since it's the one under whichour

Chapter 1: Whal Ir Color?

11

visual system evolved. The exact wavelength composition of daylight

depends on the timeofday, the weather,and thelatitude (seethe black
curve in Figure 1-51.
r Electric dischargelamps consist of an enclosed tube containinga gas

(such as mercury vapor or 'enon) that's excited by an electric charge.

The charge raises the energy level of the gas atoms, which then re-emit
the energy as photons at specific wavelengths, producing a "spikey"
spectral curve. Manufacturers use various techniques, such as pressurizing the gas or coating the inside of the tube with phosphors, to
add other wavelengths to the emitted light. Flirorescerlt Inrrrps are the
most common form of these 1amps.The phosphors coating the inside
of the tube absorb photons emitted by the gas and re-emit them at
other wavelengths.
r Computer monitors are also light sources-they emit photons. CRT
(cathode-ray tube) monitors use phosphors on the inside of the front
glas to absorb electrons and emit photonsat specific wavelengths (either
red, green, or blue). The red phosphor in particular is charncreristically
spikey (see the red curve in Figure 1-5l.We'll describe monitors in more
detail, includmgothertypes of monitors such as LCDs, in Chapter 6.

Illurninants
The word ill~rrnirznntrefers to a light source that has been measured or
specified formally in terms of spectral energy The CIE(CommissionIntenmtionnle de I'Eclairuge, or the International Commission on Illumination)--a
body of color scientists and technologistsfrom around the world that has
accumulated a huge amount of knowledge about color since the 1 9 2 0 s
has specified a number of CTE Standard Illuminants.
b

NuminantA represents the typical spectral curve of a tungsten lamp

(a standard Lightbulb).This is the green curve in Figure 1-5.

Illuminant B represents sunlight at a correlated color temperature of

4874 K. This is seldom used, if ever.
Illuminant C is an early daylight simulator (correlatedcolor temperature 6774 K). It has been replaced by the D illuminants, although you
occasionally still find it.

19

Real World Cdor Management, kcond Wtion

b u n 1-6
Color temperature

'9000 K

WOK.

400 n m

500 n m

600 m

70

wavelength

IlluminantsD is a series of illuminants that Gpkseiii- various modes

of daylight The most commonly used D illuminants are D50 and D65
with correlated color temperatures of 5000 K and 6504 K, respectively.
The D65 spectral curve is the black curve in Figure 1-5.

b IUuminant E is a

theoretical "equal energy" illurninant that doesn't

represent a real light source, and is mostly used for calculations.

IlluminantsF is a series of "fluorescent" illuminants that represent

the wavelength characteristics of various popular fluorescent lamps.
These are named F2,F3, and so on, up to F12.

The Object
- and the Color Event

The second participant in the color event is the object. The way an object
interacts with light plays a large role in determining the nature of the
color event, so in this section we examine the various ways that objects
interact with light, and the ways that this interaction affects our experience of color.

Chapter b What Is Color?

13

The Color of White

Many nf the light sources
we use--suchaslightbulbsorsunlight-produce light in a cbaracteristic way that dves us n handy
terminology to describe the color
of light: color tenrpernn~re.Every
dense object radiateswhat5 called
fher~nnlenergy. Atoms re-emit
energy that they'veabsorbedfrom
some process such as combustion
(burning of fuel) or metabolism
(burning off those fries you had
for lunch). A1 low temperatures.
this radiation is in the infrared
region invisible to humans, and
we call ir hear. But at higher temperatures the radiation is visible
and we call it light.
To study this phenomenon,
physicists imagine objects where
they have eliminated all other
sources of light and areonly looking at the radiation from thermal
energy. They call these objects
blarkbody radiators. If you're
standing in a pitch-black room,
you are a blackbody radiator
enlining energy that only an infrared detector,or an owl, cansee.
Stars (suchas our sun) are almost
perfect blackbodies as they aren't
illuminated by any other light

source. and the light the!, emit

is almost entirely from the heat
the~ratoms have absorbed from
the furnaces in the stars' cores.
A lightbulb in a dark room is an
almost-perfect blackbody radiator-all the light is from a heated
filament of tungsten. A candle is
mostly a blackbody (although if
you look closely, you can see a
small region of blue tight that's
due to direct energy released by
the chemical reaction of burning
wax rather than absorption and
re-emission of energy). To see a
blackbody in acrion, turn on your
toaster in a darkened kitchen.
Figure 1-6 shows the spectral
curves of a blackbody at various
temperatures. (Temperatures
are in kelvins fk7, where a kelvin is a degree in the physicist's
temperature scale from absolute
zero.) At lower temperatures, the
blackbody gives off heat in the
low-energyllong-wavelength
part of the visible spectrum, and
so Is dominated by red and yellow wavelengths. At 2000 K we
see the dull red we commonly
call "red hot." As the temperantre
gets higher, the curve shifts grad-

ually to the higher-energylshorter

wavelengths. At 3000 to 4000 K,
the light changes color from dull
red to orange to yeUow.The tungsten filament of an incandescent
lightbulb operates at about 2850
to 3100 K,giving itscharacteristic
yellowish light. At 5000 to 7000 K.
the blackbody's emitted light Is
relatively flat in the visible spectrum, producing a more neutral
white. At higher temperatures
of 9000 K or above, short wavelengths predominate, producing
a bluer light.
This is the system weuse to describe colors of "white light." We
refer to their "color temppratuce"
to describe whether the light is
orange, yellowish, neutral. or
bluish. Purists will remind you
!hat the correct term is actually
correlnted color ce~nperanrreas
most emissive light sourcesincluding daylight (which is
filtered bythe earth'satmosphere),
fluorescent lamps. and coniputer
monitors--aren't true blackbody
radiators, and so we're picking
the closest blackbody temperaturr to the apparent color of the
light source.

Reflection and Transmission

An object's surface must interact with lightto affect the lighr'scolor. Light
strikes the object, travels some way into the atoms at the surface, then
re-emerges. During the light's interaction with these surface atoms the
object absorbs some wavelengths and reflects others (see Figure 1-7),so
the spectral makeup of the rellected light isn't the s a m e as that of the

14

Real World Color Hanagcmcnt, Second Edition

Flgure 1-7

Reflection

Incoming white light

containsall wavelengths.

of speculnr reflecrion
may be unaffectedby surface.

Scattered reflection

surfaceabsorb long
and short wavelengths.

Surfaceof a reflectiveobject
incoming light. The degree to which an object reflects some wavelengths
and absorbs othersis called its spectral reflectance. Note that if you change
the light source, the reflectance of the object doesn't change, even though
the spectral energy that emerges is different. Reflectance, then, is an
invariant property of the object.
A transmissive object affects wavelengths in the same way as the
reflective object just described, except that the transmissive object must
be at least partially translucent so that the light can pass all the way
through it. However, it too alters the wavelength makeup of the light by
absorbing some wavelengths and allowing others to pass through.
The surface of a reflective object or the substance in a transmissive
object can affect the wavelengths that strike it in many specific ways.
But it's worth pausing to examine one phenomenon in particular that
sometimes bedevils color management-the phenomenon known as
jluorescence.

Chapter I:What Is Cdor?

15

Fluorescence
Some atoms and molecules have the peculiar ability to absorb photons of
a certain energy(wavelength),andemit photonsof alower energy (longer
wavelength). Fluorescence,as this phenomenon is called, can sometimes
change one type of visible wavelength into another visible wavelength.
For example, the fluorescent coating inside a sodiumlamp absorbs some
of the yellow wavelengths emitted by the electrically excited sodium
vapor, and re-emits photons of other wavelengths in order to produce a
more spectrally balanced light.
But fluorescence is most noticeable when the incoming photons have
wavelengths in the non-visible ultraviolet range of the spectrum, and the
cmitted photons are in the visible range (usually in the violets of blues).
The result is an object that seems to emit more visible photons than it
receives From the light source-It appears "brighter than white."
Many fabric makers and paper manufacturers add fluorescent brighteners to whiten the slightly yellow color of most natural fibers. To compensate for the slow yellowing of fabrics, many laundry detergents and
bleaches have fluorescent brighteners, often called "bluing agents"
(because they convert non-visible UV light to visible blue). We all have
m
fond memories of groovy "black lights -lamps designed to give off light
energy in the violet and ultraviolet wavelengths-and their effects on
posters printed with fluorescent inks, on whiteT-shirts. and yes, even on
teeth. depending on the brighteners in the toothpaste we had used!
Fluorescence crops up in unexpected places in color managementwe'll alert you when they aresomething to lookoutfor. For now it'senough
to know that fluorescence can be an issue in three cases:
b

Whenever a measurement instrument (a spectrophotometer, colotimeter, scanner, digital camera, or film1 is more responsive to W light
than our eye is (which has no response at all).

Whenever artificial light sources (such as lamps, flashbulbs, or scanner lamps) emit more or (more likely) less LN than daylight, which
includes a sizeable amount of UV
Whenever a coloranr (ink, wax. toner. etc.) or paper used for printing has fluorescent properties that make it behave unpredictably
dependingon thelight source used to view it (as unpredictability is [he
nemesis of color management).

16

Real World Color Management, I-nd

Edltlon

The Observer and the Color Event

Of the three participants in our simple model of the color event, you, the
observer, are by far the most complex. Your visual system is way more
complexthana blackbodyradiator or a hemoglobin molecule, so much so
that we still have a great deal toleam about it. It starts with the structures
of the eye, but continues through the optic nerve and goes deep into the
brain. In this section. we lookat various models of human vision that form
the basis of color management.

Trichromacy: Red, Green, Blue

If there's one lesson you should take from this chapter, it's that the fundamental basis for nllcolor reproduction is the three-channel design of the
human retina. Other features of human vision-such as opponency color
constancy and nonlinearity, all of which we cover in this section-are
important, but the fact that the human eye has three types of color sensors (corresponding roughly to reds, greens, and blues) is what lets us
reproduce colors at all using just three pigments on paper, or just three
phosphors in a monitor.

The gr.Your eye is one of the most beautiful structures in nature. (We
hope )ou don't think we're being too forward.) Contraw to popular belief, the main task of focusing light into an image at the hack of the eye is
handled not by the lens, but by the cornea, the curved front layer of the
eye. The lens makes minor focus adjustments as the tiny muscles that
hold it in place adjust its shape, but it does two important thingsfor color
vision. First, the lens acts as a UV filter, protecting the retina from damaging high-energy ultraviolet light-so even if the retina could see into
the UVrange (and some experiments show that it can),the lens is partly
responsible for our inability to see UV light, unlike other visual systems.
such as honeybees, birds, scanners, and digitalcameras. Second, the lens
yellows as we age, reducing OUI ability to see subtle changes in blues and
greens, while our ability to see reds and magentas is hardly affected. Our
discrimination in the yellows is always fairly weak, regardless of age.

The retina: rods and cones. The retino is a complex layer of nerve cells
lining the back of your eye (see Figure 1-8). The nerve cells in the retina
that respond to Light are called photoreceptors, or often just receptors.

&urn 1-8

The human eye

cornea,
I
iris

lens

visual axis

Cmss section of the right eyefrom above. Most color vision happens in the fouea,
where the three hmes ofmnes faroutnumber the rods. Inset shows theshawe
, o fthe rod
and cone cells, u;hich & mhe; they get their names.

Receptors come in two types, called rodsand conesbecauseof theirshapes.

Rods provide vision in low-light conditions, such as night vision. They're
sensitive in low levels of light and are largely blinded by daylight conditions. Cones are a more recent development in the evolution of the mammal retina and function in bright light conditions. We have far more rods
thancones throughout most of the retina (about120 mlllion rods to about 6
million cones),except in alittle indentation in the very center of the retina,
called the fouea, where cones outnumber rods (about 150,000 cones, with
a small number of rods, falling off to a completely rod-free region in the
center of the fovea called thefouwla).This center region, where you have
the highest density of photoreceptors, also provides the best acuity (for
example, to read these letters, youtrefocusing the image of the letters on
your fovea). It's also where your primary color vision happens.

18

Real World Cdor Manag.menl, Second Edition

Three types of cones. While all the rods in your retina are essentially
the same, the cones fall into three types. One responds primarily to the
long wavelengths of light and has little response in the middle or short
wavelengths. One responds primarily to the middle wavelengths,and the
third responds to the short wavelengths (seeFigure 1-9).Many people call
these the red cones, green cones, and blue cones, respectively,because of
the colors we normallyassociate with these three regions of the spectrum,
but it's less misleading to refer to them as the long-, medium-, and shortwavelength cones (or L, M, and s),respectively.

Mchromacy and trlsllmulus. 'Iko related and often confused terms are
trichromacy and iristimulus. The term trichromacy (also known as the
three-component theory or the Young-Helmoltz theory of color vision)
refers to the theory, now well verified, that we have three receptors for
color [the three types of cones).
The term tristimulus refers to experiments and measurements of
human color vision involving three color stimuli, which the test subject
Figure 1-9
Peak sensitivities

wavelength
The peak wavelengths of the photoreceptor pigments. The three cone
photoreceptors have peak absorptions at 420 nm, 530 nm, and 565nm,
respectively.The rod photoreceptor (gray line) has a peakabsorption
at499 nm.

Chapler 1: What h Cdor?

19

uses to match a target stimulus (see Figure 1-10). In other words, trichromacy refers to our three color receptors, and tristimulus refers to the
experimentsthatuse three stimulito verify andmeasure trichromacy. The
most comprehensive tristimulus model has been defined by the CIE and
forms the basis for color management.
The importance oftrichromacyfor the graphic arts is that wecansimulate almost any color by using just three well-chosen primary colors of
light. TWOcolors are not enough-no matter how carefully you choose
them, you cannot duplicate all colors with two primaries.And four colors
is unnecessary-any color you can produce with four colors of light you
can reproduce with just a well-chosen three.
Addltlvr pdmarla. It's the trichromatic stmcture of the human retina
that makes possible what we know as the additive primary colors (see
Figure 1-11). If you choose three light sources with overlapping spectra
that divide up the visible spectrum roughly into thirds, each one adds
nguw 1-10

Tristimdus experiment

The obserueradjusts the intensities of the red, green, and blue lamps until they
match the target stimulus on the split screen.

PO

R u l World Color Hanagnnml, kcond Wition

wavelengths that tickle one or more of your eye's three receptors. Divide
the spectrum r o u m into thirds and you get three Light s o m a that we
would call red, green, and blue. Starting from black (no wavelengths),
the three colors udd wavelengths-hence "additive colorn-until you get
white (all wavelengthsin even proportions).
SuMnc(ivr pfirmrkr.Trichrornacyisalso the source of our subtmctiveprimaries-cyan, magenta, and yellow (see Figure 1-11).Rather than adding
wavelengths to black, they act to subtract wavelengths from an otherwise
white source of light. In other words, the term ''cyanink" is just a name for
"long-wavelength-subtractor,"or simp1y"red-subtractor"-it subtractslong
(red) wavelengths ram white light (suchas that reflected from otherwise
blankpaped. S i a r l y , magentainkisaUmedium-wavelength-subtractor,"
or a "green-subtractor." And yellow is a "short-wavelength-subaactor,"or
a "blue-subtractor"
But the bottom line is that both additive and subtractive primaries
work by manipulating the wavelengths that enter our eyes and stimulate
our three cone receptors. This manipulation, when done cleverly, stimulates our three receptors in just the right proportions to make us feel l i e
we are receiving light of a certain color.
There are still afew more points to make about trichromacy.

Color spaces. The three primary colors not only allow us to define any
color in terms of the amount of each primary, they also allow us to plot
the relationships between colors by using the values of the three primaries
Figure 1-11

Primary colors

additiue primary colors

submtive primary colors

Chapln1: Whal b Color?

P1

as Cartesian coordinates in a three-dimensional space. where each primary forms one of the three axes.This notion of color spaces is one you'll
encounter again and again in your color management travails.

Relationshipwith the spectrum: why no1GRB or YMCK?Have youhave

ever wondered why we have the convention that RGB is always written in
that order (never GRB or ERG)?Similarly, CMYK is never written YMCR
(except to specify the order in which inks are laid down, or when adding
a twist to a certainVillagePeople song). Well. now you know why:
The colors of the spectrum are usually listed in order of increasing
frequency: red. orange, yellow, green, blue, and violet-ROYGB\!
r The additive primaries divide this spectrum roughly into thirds, corresponding to the reds, greens, and blues. So ROYGBVleads to RGB.
We write the subtractive primaries in the order that matches them to
their additive counterparts (their opposites).Thus RGB leads to CMY.
You may find this last convention handy for remembering complementary colors. For example, if you're working on a CMYK image with a
blue cast, and can't remember which channel to adjust, just write RGB.
and then under it C M Y The additive primaries match their subtractive
complements (Yis under B because yellow subtracts blue). So to reduce
a blue cast, you increaseY. or reduce both C and M.

Artltkial trichromats (scanners, cameras, etc.). We also use trichromacy

to make devices that simulate color vision. The most accurate of these
are calorimeters, which we'll describe later in this chapter. A colorimeter
tries to duplicate the exact tristimulus response of human vision. More
common examples of artificial rrichromats are cameras, which use film
sensitive to red, green, and blue regions of the spectrum, and scanners
and digital cameras, which use sensors (CCDs)with red, green. and blue
filters that break the incoming light into the three primary colors. When
the "receptors" in these devices differ horn our receptors, the situation
arises where they see matches that we don't and vice versa. We discuss
this phenomenon, called metamerism.,later in this chapter.

22

Rcal World Color Managcnrmt, Second Edition

Opponency: Red-Green, Blue-Yellow

Some fundamental features of human color vision not only are unexplained by trichromatic theory, but seem to contradict it.

The strangc case of yellow. Many of us grew up with the myth that the
primary colors are red, yellow, and blue--not red, green, and blue. Studies have shown that more cultures have a name for "yellow" than have a
name even for "blueu-and our own color naming reveals a sense that
"yellow" is in a different, more fundamental category than, say, "cyan"
and "magenta". Even though ThomasYoung first demonstrated that you
can make yellow light by combining red and green light, it seems counterintuitive. In fact, it's diicult to imagine any color that is both red and
green at the same time.

Other effeds uncxplald by trlchromaq. The fact that we can't imagine a reddish green or a greenish red is evidence that something more is
going on than just three independent sensors (trichromacy).The same
holds for blue and yellow-we can't imagine a yellowish blue. The effects
of simultaneous contrast and afterimages shown in Figures 1-12 and
1-13 are other examplesthe absence of a color produces the perception
of its opposite. Finally, anomalous color vision (color blindness) usually
involves the loss of color differentiation in pairs: a person with anomalous
red response also loses discrimination in the greens, and a person who
has no blue response also has no yellow response.
Figure 1-19

Simultaneouscontrast

The two Xs look dramnticallvdiffmentin brinhmess

and color, but ns shown by where they touch,-they are
identicnl. W e d on a painting byJosefAlbers)

b u n1-13
Successive contrast

Instructions: stare at the white cross in the center of the upper image
for a full minute. Then quickly look down at the lower area, and
stareat the black cross.
After a few seconds and some blinking, the afterimage should appear:
Note that the colors are chosen to demonstrategreen-red.yellow-blue,
and black-white opponency.

Opponcncy. Experiments and observations by Ewald Hering in the late

1800s focused on these opponentpairs. Why is it that we can have reddishyellows (orange),blue-greens, and blue-reds (purples),but no reddishgreens andno yellowish-blues?Tosay that something is both red and green
at the same time is as counterintuitive as saying that something is both
light and dark at the same time. The opponent-color theory (also known as
the Hering theory) held that the retina has fundamental components that
generate opposite signals dependingon wavelength.The key point is that

the color components in the retina aren't independent receptorsthat have

no effect on their neighbors, but rather work as antagonistic or opponent
pairs. These opponentpainare light-dark, red-green, and yellow-blue.

Recontiling opponency and trkhromaq. While the advocates of the

opponency and trichromacy theories debated for many years which theory best described the fundamental nature of the retina, the two theories
were eventually reconciled into the zone theory of color. This holds that
one layer (or zone) of the retinacontainsthe three trichromaticcones, and
a second layer of the retina translated these cone signals into opponent
signals: light-dark, red-green, yellow-blue. The zone theory has held up
well as we continue to learn more about the layer structure of the retina,
and use simple neural-net models that explain how opponent signals can
emerge from additive signals (see Figure 1-14).

Wchromacy and
opponency in the retina

Firstzone (or stage):

layer of retina with
three independent
lypes of cones

Second zone (or stage):

signakfrorn cones
either excite or inhibit
second layer of
neurons, producing
opponent signals

1'lza

red orgreen h g h r to
dark

Chapter 1: What I s Cdorl

45

Opponency and trichromaq in the CIE models. If all this opponency

and trichromacy stuff is about to make your head explode, let us assure
you that it$ very relevant to color management. In Chapter 2 we'll look
at the CIE models of color vision, which are used as the basis for color
management computations, but the important point here is that many of
the CIE modek-such as C E LAB used in Photoshop and in most color
management systems (CMSsI-incorporate aspects of both trichromacy
and opponency. CIE LAE4 is based on the results of tristimulus experiments such as the one described in Figure 1-10, but it represents colon in
terms ofthree values that represent three opponent systems: L* (lightnessdarkness), a* (red-greenopposition), and b' (hlue-yellowopposition).
So using LAB, we can convert trichromatic measurements [measurements made with instruments that duplicate the responses of our cone
receptors) to values in an opponent-color system. Achieving this wasn't
simple, and it's far from perfect (as we will discuss in Chapter 2), but it's
very clever, and surprisinglysuccessful as a model, particularly when you
consider that it was designed simply as a test of the reigning theories of
color vision, not as aworkhorsefor computercalculations For the graphic
arts industry.

Metamerism
II you've encountered the term "metamerism" you've probably heard it
referred to as a problem or as an "error" of human vision. But-as programmerslove to sap-it's not abug, it's afeature. Not only is metamerism
inherent intrichromaticvision,it's the feature that makes color reproduction
possible.
In simple terms, metomerismis the phenomenon whereby two different
color samples produce the same color sensation. By "different color samples" we mean two objects that have different spectnl characteristics. So,
rememberingourlight-object-observerdefinition of color,if the objects are
different but theyproduce the same"color" (the samecolor sensation),this
match may be dependent on (1) the light illuminating both color samples,
or (2) the observerviewing the two color samples. Under different lighting,
or to a different observer, the two samples may not match.
Twospectrally different color samples that produce the same color sensations are called mernmers. Or we say that the two colors are metarneric
under certain lightingor to a certain type of observer.

26

Real World Color Management, kcond Edition

You may run into different, seemingly contradictory deAnitions of

metamerism. For example, many books give one of the following hvo
definitions:
t Metamerism is when twocolorsamples produce the samecolorsensa-

tion under certain lighting.

Metamerism iswhen two coiorsamplesproduce different colorsensations under certain lighting.
Actually, metamerism is when both of these events happen to the
same pair of color samples. The two color samples match under some.
but not all. lighting. The first statement focuses on the match that can be
made using radically different spectra. The second statement focuses on
the fact that the match is tenuous. The reason you need to understand
metamerism is that virtuallyall our color-matchingactivitiesrely on makingametamericmatch between two colors orsets ofcolors-comparinga
chrome on a light tablewith a scanned imageon a monitor, or comparing
a proof with a press sheet. It's highly unlikely that the two samples will
have identical spectral cuwes, but thanks to metamerism. we can make
lhem match-at least in some lightingconditions.

Relationship between two color samples. Metamerism is always a

relationship between nuo color samples-a single color sample carit be
metameric any more than it can be identical. You may hear some people
refer to "a metameric color," or talk about a printer having "metameric
inks," but we think that this usage is both confusing and wrong. A printer
with inks that were rruly metamets of each other would be pointless--the
inks would allappear to bethe same color under some lightingcondition.
What they realkmean is that the inks have spectral properties that make
their appearance change more radically under different lighting conditions than most other inks.
Why metamerism happens. Metamerism happens because the eye
divides all incoming spectra into the three cone responses. ltvo stimuli
may have radically different spectral energies, but if they both get divided
up between the three cone types. stimulating them in the same way, they
appear to be the same color.

Chaplu 1: What L Color?

37

In the terms of our light-object-observer model (remember Figure 1-11,

the color "event" is a product of three things: the wavelengths present
in the light source; the wavelengths reflected from the object or surface;
and the way the wavelengths are divided among the three receptors (the
cones in the eye).
What matters isn't the individual components, but the product of the
three. If the light reaching your eye from object A and object B produces
the same cone response, then you get the same answer-the same color
sensation (see Figure 1-15).

ngum 1.15
Metamerism

lluo diferent
colorsamples...

... viewed under lightsourceA...

... result in thesame
color experience.

This is a metameric

match.

... to an observer with

normal color vision...

The same pair of ... viewed under lightsource B ...

color samples...

... results in two different

color expe

'

?s.

This is a metameric
mismatch.

...to sameobserver...
Objects can only reflect those u~uelengthsthat are present in the light source.
Source B containsfewer red wavelengths than source A, so thesamples that
appear to match undersourceA produce a mismatch undersource B.

48

Real World Color Management, Second Edition

Metamerlrm in everyday life. If you've ever bought nvo "matching"

items in a store--a tie and a handkerchief, or a handbag and s h o e s o n l y
to lind that they look different when vien-ed in daylight or home lighting,
you've experienced metamerism. Fred has pairs of whlre jogging socks
that match when he puts them on indoors, but when outdoors one looks
noticeablybluerthantheother, probably (he theorizes)because they were
washed separately using laundry detergents with different W brighteners. Catalog-makers cite as a common problem the fact that a clothing
item shown in a catalog doesn't match the color of the item received by
the customer-perhaps a result of a metameric match that was achieved
in the pressroom that failed in the customer's home.
These examples illustrate the fragile nature of all color-matching exercises-when wc match colors, we're almost always creating a metameric
match under a specific lighting condition. That's why we use standard
lighting conditions when we evaluate proof-to-press matches.
In theory, ifthe catalog publisher knew theconditions underwhich the
catalog would be viewed, she could tailor the color reproduction to that
environment, but in practice, it's just aboutimpossible to know whether
prospective customers will look at the catalog in daylight on their lunch
hour, under office fluorescent lighting,at homeunder incandescentlighting, or curled up by a cozy fire under candlelight.
Ultimately, metamerism problems are something we simply have to
accept as part of the territory, but ifyou're working in an extreme c a s e
like designing a menu for a fancy restaurant that will be viewed mostly
under candlelight-you may want to test important matches under a
variety of viewing conditions.

Metamerism k your friend. You need to make peace with metamerism.

Many people first encounter metamerism as a problem-a color match
achieved at great effort under certain lighting fails under different lighting because of odd properties of certain inks or paper. But it's not, as
some people describe, an error of our visual system. It's just an inevitable
side-product of the cleversolution evolution producedforderivingwavelength, and hence color information. using only three types of sensors.
More importantly for color management. metamerism is what makes
color reproduction possible a t all. Metamerism lets u s display yellows or
skin tones on a computer monitor without dedicated yellow or skin-color

Chapter 1: What k Color?

99

phosphors. Metamerism lets us reproduce the green characteristic of

chlorophyll (the pigment found in plants) without chlorophyll-colored
ink--or even an ink we would call green (see Figure 1-16)!

Thanks to metamerism, we can reproduce

the color of this leaf in print without usinga
single drop ofgreen ink.

Without metamerism, we'd have to reproduce colors by duplicating

the exact spectral content of the original color stimulus. (Incidentally,this
is what we have to do with sound reproduction-duplicate the stimulus
of the original sound wavelength by wavelength.) If you think your ink
costs are high today, imagine if you had to have thousands of colors of
ink instead ofjust four!

Camera and scanna n w b ~ i s mW

. e mentioned that a metameric relationship between two color samples is dependent not only on the light
source, but also on the specific observer viewing the two samples. A pair
of color samples can produce one color sensation to one observer, but a
different color sensation to a second observer. This "observer metamerism" is sometimes an issue with color management when the observer is
one of our artificial trichromats.
If a scanner'sred, green, and blue detectors respond differently than our
cone sensors, the scanner can see a metameric match where you and I
see separate colors, or conversely, a pair of samples may appear identical

30

Redl World (dor Hanagrmcnt, kcond Edltlon

to us but different to the scanner (see Figure 1-17). This is sometimes

called scanner metamerism, and we shall see in Chapter 8 that this is the
reason it's difficult to use a scanner as a measurement device for making
profiles. Similarly, if a film or digital camera has different sets of metameric matches than we do, we would call this camera metamerism, and
while there's little that color management can do about it, you need to be
aware of it as an issue.

... seen under lightsourceA...

Figure 1-17

How scanner
metamerismhappens

A pair ofspemlly
differentcolor
samples...

... results in the same

color e.~pn-n---

Wesee the same colol:

... by an obseruer wrrn

certain color vision ...

... underascanner lamp

the sameas 1ightsourceA...

... pmduces different

measurements...

same pair of
color samples...

%i

RGB =
148,I34,4O)

A.

... bemuse tl

... nWhas a
different color response

RGB =
(XLi,l36,119)

The scanner sees two

d i f f i n t colors.

31

Chapla 1: Whal k Cob?

Won-Linearity: Intensity and Brightness

Another important property of the human visual system is that it's nonlinear. All this means is that our eyes don't respond in a one-to-one fashion to the intensity [the number of photons reachmg your eye as you
might count them with a light meter) by reporting brightness (the sensation you feel) to the brain. When you double the intensity, you don't see
the light as twice as bright. If things were that simple. if you drew a graph
of intensity versus brightness, this would be a straight line as in Figure
1-18a, and we would call this linear response. Instead the relationship
between intensity and perceived brightnesslooks like the graph in Figure
1 -18b. Ib perceive something about twice as bright, we h a w to multiply
the intens~tyby about nine!

20

10

60

light intensity

00

20

40

60

80

100

light intensity

a. Linem msponse-light inlensin) b, Non-linear response--increased

translrrresdirectly to brig11tness.
light intensity prodrrces progressively
s~nallerincr~nsesin brig11 mess.

This non-linearity is common in human perception. If you double the

intensity of asound, you don't hear it asmice as loud. If you put two sugar
cubes in yourcoffee instead ofyour normal one, you don't taste it as twice
as sweet. The degree of non-linearity varies for different senses, but they
often shape as Figure 1-18b.
Our non-linear responses are what allow our sensory systems to function across a huge range of stimuli. The difference m intensity between
a piece of paper illuminated by daylight and the same piece of paper
illuminated bymoonlight is about 1.000.000:1. But nerve cellscan respond
(inthe number of nerve firings per second) only at a range of about 100:1,
so a huge number of inputs have to map to a small number of outputs.

39

Real World Color Management, Second Ediion

Non-linearitylets our sensory systems operate over a wide range of environments without getting overloaded.
The good news is that just because our response is non-linear doesn't
mean it's complicated. We could have evolved a totally wacky response
curve instead of the nice simple one in Figure 1-18b. Instead, this simple
curve resembles what mathematicians call n power nrrve. In fact, using
the trickof graphing the logarithms of intensity and brightness simplifies
the curve even more, turning it into a straight line that can be character[zed by its slope. (If you aren't familiar with logarithms, see the sidebar,
"What's a Logarithm, and Why Should I Care?")Many of the scales we use
to measure perception are logarithmic scales, including decibels. which
measure perceived loudness, and optical density (OD),which measures
how dark or light an object appears.
The non-linear nature of our response to light impinges on color
management in several ways, but the most important is that the various
devices we use to measure light have linear responses. To relate these
measurements to human perception, we have to translate them from the
linear realm oflight to thenon-linear realm of perception (see"We Cannot
Truly Measure Color, Only Light,"later in this chapter.
So next lime someone accuses you of being non-linear, you can respond, "yes I am, and glad to be that way!"

Achromatic Component: Brightness

The term Origlihessrelers to our perception of intensity (thenumber ofphotons reaching our eye). Of the three color atuibute+brighmess, hue, and
saturation-we tend to think of brightness as different from the other two.
in part because we can detect variations in brightness even when there isn't
enough light (orenough differentwavelengths in the light)to see color.
On a dark night, our vision is produced by our rods, which have no
color response, but we can still see differences in brightness, and if we
view objects under monochromatic light. everything takes on the color
of the light, but again we still see differences in brightness.
Brightness describes the qimntiry of light ("how much"). while hue
and saturation describe the qunlityoflighf ("what kind"). Detectingvariations in brightness is the fundamental task of vision itself, while color, as
established by hue and saturation, is just icing (albeit some tasty icing).
Vision is fundamentally about "countingphotons," whilecolor deals with
"categorizing" these photons into differing types.

Chapter 1: What Is Color?

33

What's a Logarithm,
and Why Should I Care?
A lojiarirhnl is a handy way to
express numbers that vary over
huge ranges. For example, if a
set of values ranges from 0 to
1,000,000, you'll probably want
much more precision when you
compare values between 1 and
10 than you will when comparing
values in the 100,000 to 1,000,000
range. For example, you might
notice when the price of a movie
ticket goes up from 57 to $8, but
when buying a house you probably don't care as much about the
difference between 5247,637 and
$247,638. The numbers at the
upper end are not only unwieldy.
they have an unnecessary amount
of precision.
Logarithms are a nice way to
deal with this problem, Instead of

imagining all the numbers from

0 to 1.000.000 l i e d up on a huge
~ l esor that the distance between
any two numbers is always the
same, a logarithmic scale compresses the distance between any
two numbers as the numbers get
higher. In this way, the distance
between 0 and 10 is the same as
the distance between 10 and 100.
So between 0 and 10 you have
Uck marks for 1, 2, 3, etc. hut
between 10 and 100 you have
tick marks for 10. 20.30, etc. You
end u p w t h a rulerthat looks like
Figure 1-19.
The nifty thing about such
a scale is that when you take a
non-linear response function like
that shown in Figure 1-lab, and
graphthelogarithmsofthevalues

(what'scalled a log-IoggraphLthe
result is a straight line again (see
Figure 1-20)!
This is why, when we measure some physical value with
the purpose or predicting how
our nervous system perceives
that stimulus. we use some sort
of logarithmic quantity. For
example, what we call dmsity is
a logarithmic function derived
from a measurement of light
intensity and gives us a measurement of perceived darkness.
Another example is the unit we
call a decibel, which is also a
logarithmic function derived
froni sound intensity, and which
gives us a measurement of
perceived loudness.

Alog-log graph
ofFigurc 1-10b

-.?
5

,
1.0

2.0

3.0

log of light intensity

Logarithms nmm the response curve
into n straigh~line.

34

Real World Color Mmagcment, Second Wlilon

So color scientists often speak of the achromatic and chrornaric attributes of a source of light or color.The acl~romaticattributeis brightness as
are
we perceive it independently from color, and the rl~ron~aticamibutes
those that we commonly associate with color, independent of brightness.
Most mathematical modelsofhuman colorvision, including the ones that
lie at the heart of color management, treat brightness separately from the
chromatic attributes. We discuss these models in Chapter 2, Compriters
a n d Color.

Brlghtneu US. lightness. In color science, we draw a distinction between

brightness and ligl~rrfess.For most purposes, the hvo words mean the
same thing-they both refer to the eye's (non-linear) perception of intensity. But the strict definition is that lighrness is relative brightness. In
other words, lightness is the brightness ofan object relative to an absolute
white reference. So lightness ranges from "dark" to "light" with specific
definitions of black and white as the limits, while brightness ranges from
"dim" to "bright" with no real limits. The distinction matters because we
can measure lighmess and assign specific numerical values to it, while
brightness is a subjective sensation in our heads.

Chromatic Components: Hue and Saturation

Brightness is aproperh. of all vision, but hue and saturation pertain only
to color vision.Together, they're known as the chromatic components of
color vision, as distinct from the achromatic component of brightness.

Hue. Definingthe word hrreasan independent component olcolor is like

w n g to describe the word "lap" while standing up There are multiple
definitions for hue--some more vague than others. We've even seen hue
defined as "a color's color."
The most precise definition is that hue is the attribute of a color by
which we perceive its dominant wavelength. All colors contain many
wavelengths, but some more than others. The wavelength that appears
most prevalent in a color sample determines its hue. We stress the word
"appears" because it may not be the actual dominant wavelength-the
color sample simply has to produce the same response in the three cone
types in our eyes as the perceived dominant wavelength would. In other
words, it produces a metameric match to a monochromatic light source
with that dominant wavelength.

Amoreuseful andequallyvalid definitionof hue is that ifs the attribute

of a color that gives it its basic name, such as red, yellow, purple, orange,
and so on.We give these names to aregion ofthe spectnun, then we refine
individual color names by adding qualifiers like bright, saturated, pale,
pure, etc. Thus, red is a hue, but pinkis not--it can be described as a pale
or desaturatedred. The set of basic names that people use is quite subjective, andvaries fromlanguageto language and culture to culture.As you'll
see later in this chapter whenwe talkabout psychological aspects of color,
this connection between hue and color names can be fairly important to
color reproduction.

htuta(ion. Saturation refers to the purity of a color, or how far it is from

neutral gray. If hue is the perceived dominant wavelength, saturation is
the extent to which that dominant wavelength seems contaminated by
other wavelengths. Color samples with awide spread ofwavelengthsproduce unsaturated colors, while those whose spectra consist of a narrow
hump appear more saturated. For example, a laser with a sharp spike at
520 n m would be a totally saturated green (see Figure 1-21].

nsumc-m loosb
Spectra and saturation

lluo samples with same hue

(peak wavelength = 520 nm)
\

400 nm

I'

500 nm
wavelength

Both sampIe8 have the s a m e p k wavelength,so they appear to be the same

hue, but thenarrow spectrum contains fewer contaminating wavelengthsand
hence appears more satllmted

36

Real World Color Managunent, kcond Wition

Representationsd hue, saturation, and brightness. Most hue diagrams

and color pickers represent hue as an angle around some color shape, and
saturation as the distance from the center. For example, the Apple color
picker accessed from most Macintosh graphics applicationsshows a disk
with neutral grays in the center and saturated pure hues at the edge. The
disk is a cross section of the full space of possible colors, so each disk
represents all the colors at a single brightness level. Brightness is represented as a third axis ranging from black to white along which all these
cross sections are piled, resulting in a color cylinder, sphere, or double
cone (seeFigure 1-22). You choose colors by first picking a brightnessvalue,

.Hue

Agun 1-99

nue, saturation, and

brightness
All the colors here arefully
saturated-at least one
primary is at level 255.

-1:1
Brightness

None of the colors here

isfully saturated.

Saturation (tint)

Chapter (:What Is Color?

37

which takes you to a certain cross section, then choosing a point on the
disk representing the hue (the angle on the disk) and the saturation (the
distance from the center of the disk).

Measuring Color
All the preceding knowledge about colorwould fall into the category that
Fred likes to call "more interesting than relevant" if we couldn't draw correlations between what we expect people to see and things we can physically measure. The whole purpose of color management is ultimately to
letus produce a stimulus (photons,whether reflected from aphotograph
or magazine page) that will evoke a known response (the sensation of a
particular color) on the part of those who will view it.
Fortunately. we areable to draw correlations between perceived color
and things we can measure in the real world, thanks to the people who
have not only figured out the complexities of human vision but have
modeled these complexities numerically. We need numerical models to
manipulate and predict color using computers, because computers are
just glorified adding machines that juggle ones and zeroes on demand.
We'll look at the numerical models in more detail in Chapter 2, Cornpute m and Color, but first we need to look at the various ways we can count
photons and relate those measurements to human perception, because
they form the foundation for the models.

We Cannot Truly Measure Color, Only Light

"hleasuring color" is really an oxymoron. We've pomted out that color
is an event with three participants-a light source, an object, and an
observer-but the color only happens in the mind of the observer. One
day we may have both the technology and a sufficient understanding of
the human nervous system to be able to identify which of the zillions of
electrochemical processes taking place in our heads corresponds to the
sensation of fire-engine red, but for now, the best we can do is measure
the stimulus-the light that enters the observer's eye and produces the
sensation of color. We can only infer the response that that stimulus will
produce. Fortunately, thanks to the work of several generations of color
scientists, we're able to draw those inferences with a reasonable degree
of certainty.

38

RuaI World Color Management, Second Edition

We use three main types of instruments ro measure the stinluli that
observers-our clients, our customers, our audience, and ourselves-will
eventually interpret as color. They all work by shining light with a known
spectral makeup onto or through a surface, then using detectors to measure thelight thatsurfacereflectsortransmits.Thedetectoris just a photon
counter-it can't determine thewavelength of the photons it is countingso the instrument must filter the light going to the detector.The differences
between the three typesof inshuments-densitometers, colorimeters, and
spectrophotometers-arethe numberand type of filters they use, and the
sensitivity of their detectors.
r Densitometers measure clensity, the degree to which reflective surfaces absorb light, or transparent surfaces allow it to pass.
Calorimeters measure colorimerricvalues, numbers that model the
response of the cones in our eyes.

Spectrophotometers measure the spectral properties of a surface; in

other words, how much light at each wavelength a surface reflects or
transmits.

Densitometry
Densrromeny plays an indirect but key rolein color management. Density
is the degree to which materials such as ink, paper, and film absorb light.
The more light one of these materials absorbs, the higher its density. We
use densitomenyas a tool for pmcessconhol,which is (to grossly oversimplify) the art and science of ensuringthat ourvarious devices are behaving
the way we want them to. In prepress, we usedensitometers to assure that
prepress film is processed correctly In the pressroom, we use densitometers to make sure that the press is laying down the correct amount of
ink-if it's too little, theprint willappearwashed out, and if it's too much,
the press isn't controllable and ink gets wasted.
Wealso use densitometers to calibratedevices-changing their hehavior to make them perform optimally, like doing a tune-up on your car.
We use densitometers to linearize imagesetters, platesetters, and proofers, ensuring that they produce the requested dot percentages accurately.

Chapter 1: Whal I s Color?

39

Some monitor calibrators are densitometers, though colorimetets and

spectrophotometers are more often used. We discuss calibration and
process control in more detail in Chapter 5. Mensrrrement, Cnlibmtiott,
and Process Control.
You may never use a densitometer-they'requite specialized,and most
of their functions can be carried out equally well by- a colorimeter or a
spectrophotometer-buL ilk helpful to understand what they do.

Reflectance (R) and transmittance 0.

Densitometers don't measure
density directly Instead, they measure the ratio between the intensity of
light shone on or through a surface, and the light that reaches the detector in the instrument. This ratio is called the refectnnce [R)or the frnnsmitrance (T),depending on whether the instrument measures reflective
materials such as ink and paper, or transmissive materials such as film.
Densitometers use filters that are matched to the color of the material
you are measuring, so that the detector sees a tlat gray. Pressroom densitometers, for example, have filters matched to the specific wavelengths
reflected by cyan, magenta, yellow, and black inks, so that they always
measure the dominant wavelength. This means first that you have to
know and tell the densitometer exactly what it is measuring, and second,
that you can only use a densitometer to measure materials for which it
has the appropriate filtet.
Density is a logarithmic function. Density i s computed from the measurement data using a logarithmic function, for several reasons. First,
as we've seen, the human eye has a non-linear, logarithmic response to
intensity,so a logarithmic density function correlates betterto how wesee
brightness. Second, it correlates better with the thickness of materials
like printing inks or film emulsions. which is one of the main functions
of densitometers. Third, a logarithmic scale avoids long numbers when
measuringvery dark materials (such as prepress films).
To see what this means, imagine a surface that reflects 100% of all the
light that strikes it-a so-called perfecr diflrser. Its reflectance, R, is 1.0,
and its density is 0. If we consider other surfaces that reflect half the light,
one-tenth the light, and so on, we derive the followingvalues:

40

Real World Color Management, Second Edition

Reflectance 07)
Densitv (Dl
1.0
0.0
0.5
0.3
0.1
1 .o
0.01
2.0
0.001
3 .0
0.0001
4.0
0.00001
5.0
Notice how a number like 5.0 is far more convenient than 0.00001?
One of the characteristicrequirements for a densitometer is that it has a
very wide dynamic range. In fact, the dynamic range of other devices (e.g.,
scanners and printers), media (prints vs. film), or even images (low-key,
high-key) is expressed in terms of density units, typically abbreviated as D
(density)or O.D. (opticaldensity).For example. thedynamic range ofa scanneris expressed in terms of the D_,, (minimum density)and Dm_(maximum
density) at which the scanner can reliably measure brightness values.

Colorimetry
C o l o r i m e t ~ ithe
s science of predicting color matches as typical humans
would perceived them. Inother words, itsgoal is to build aoumeric model
that can predict when metamerism does or does not occur. To be considered a success, a colorimetric model must do both of the following:

Where a typical human observer sees a match (in other words, metamerism) between two color samples, the colorimetric model has to
represent both samples by the same numeric values.
Where a typical human observer sees a difference between two color
samples, not only should they have different numeric representationsin
the model, but the model should also be able to compute a colordifferencenumber that predicts how different they appear lo the observer.
The cument models available aren't perfect, but thanks to the pioneering work of the CIE. they're robust enough to form the basis of all current
color management systems. If CIE colorimetry is all just so much alphabet soup to you, the one key fact you need to know is that the various
CIE models allow us to represent numerically the color that people with
normal color vision actually see. Compared to that one insight. the rest is

Chapter 1: What h Color?

41

detail, but ifyou want to really understand how color managementworks

and why it sometimes fails to work as expected, it helps to have a basic
understanding of these details. So let's look at the body ofwork that forms
the core of color management, the CIE system of colorimet~y

The CIE dorimetrlc system. Most modem colorimetry and all current
color management svstems are based on the colorimetric system of the
CIE, which we introduced at the beginning of this chapter. This system
contains several key features.
r Standard Illurninants are spectral definitions of a set of light sources
under which we do most of our color matching. We introduced you to
the Standard Illurninants A through E but in the graphic arts world.
the two most important are D50 and D65.

+ The Standard Observer represents the full histimulus response of the

typical human observer, or in plain English, all the colors we can see.
Mostcolorimetersuse the 2" (1931)Standard Observer,butthere's also
a 10' (1964) Standard Observer. The latter arose as a result of later
experiments that used larger color samples that illuminated a wider
angle of the fovea and found a slightly different tristimulus response.
Itkrare that you'll encounter the 10" observer, but we'd beremiss ifwe
didn't mention it.
The CIEXYZPrhary System is a clever definition of three imaginar).
primaries derived from the Standard Observer tristimulus response.
(Theprimaries are imaginary in thatthey don't correspond to any real
light source-it's impossible to create a real light source that stimulates only our MorS cones--but the response they model isveryreal.1
Not only does every metameric pair result in the same XYZ values.
but the primaryY doubles as the average luminance function of the
cones-so a color'sY value is also its luminance.

+ The CIE xyY diagram is a mathematical transformation of X\T that

makes a useful map of our color universe. It shows additive relationships-a straight line between two points represents the colors that
can be created by adding the two colors in various proportions (see
Figure 1-23). But it's important to note that XYZ and xyY don't factor
in the non-linearihr ofthe eve, and so the distancesare distorted.

4P

Redl World Color Managmnenl, Second Wition

b

Agunl-Pa

The xy chromaticity chart

0.8

The uniform color spaces (LAB, L W ) are two color spaces that were
defined by the CIE in an attempt to reduce the distortion in color distances. Both compute the lightnessvalue Lain exactly the same wayit's approximately the cube root of the luminance valueY (whichis a
roughapproximationof our logarithmic response to luminance).Both
attempt to create a space that is perceptually unifom-in otherwords,
distances between points in the space predict how different the two
colors will appear to a human observer. As amult, the spaces also have
features that resemble hue, saturation, and brightness, and Ci the case
of L*a*b*)our three opponent systems. LAB has largely replaced L W
in most practical applications, and while it isn't perfect (it exaggerates
differencesin yellows and underestimates them in blues, for example),
it's pretty dam useful. The quest for a perfectly uniform color space
continues, but thus far LAB has stood the test of time.

Chrptn 1: What b Color?

b

43

Color difference (AE) calculations offer an easy way to compute the

color difference between two samples. If you measure the two colors,
plot them as points in the uniform space, and then compute the distance between them, that distance will by definition, correlate well
to the difference a human observer will see. This value is called AE
(pronounced"delta-E"-delta is theGreek Ietter'D' we commonly use
to represent a difference).

Colorimeten. Colorimeters measure light through filters that mimic approxinlately the human cone response, and produce numerical results
in one of the CIE color models. Most colorimeters have user-selectable
reporting functions that let you obtain the colork values in CIE XYZ,CIE
LAB, CIE L W or other colorimetric spaces, as well as measuring the AE
value between two color samples.
While colorimeters are very llexible in their reporting functions, they
have important limitations in the colorimetric assumptions they must
make. Specifically, they're limited to a specific Standard Illurninant and
Standard Observer, though some colorimeterslet you switchbetween different illuminants (such as a D50 and a D65 option).
Colorirneters can't detect metamerism. They can tell whether or not
two samples match under the specific illuminant they use, but they have
no way of telling if that match is metameric-dependent on the illuminant--or if the samples really do have identical spectral properties that
would make them match under ail illuminants. Fortunately, for most
color-management purposes. computing a color match under a single
illuminant is enough.

Colorimetry and color management. Colorimet~is the core of color

management, because it allows 11s to define color unambiguously as
it will be seen by humans. As you'll learn in Chapter 2, Computers and
Color, the systems we use to represent color numericallyin our everyday
image, illustration,and page-layout files are fundamentally ambiguous as
to the actual color they represent. Colorimetry allows color management
systems to remove that ambiguity And in Chapter 3. ColorhlanagemenrIfow It Work you'll see how colorimetry lets color management systems
compute the numbers that we send to our various color-reproduction

44

Real World Cdor Hamgsnrmt, Second Edition

devices-monitors, desktop printers, proofing systems, and printing
presses--to ensure that they reproduce the desired colors. For now, just
accept the fact that color management systems feed on colorimetry.

Spectrophotometry
Spectrophotomeny is the science of measuring spectral reflectance, the
ratio between the intensity of each wavelength of light shone onto a surface and the light of that same wavelength reflected back to the detector in
the instrument. Spectral reflectance is similar to the reflectance (R) measured by a densitometer and then converted to density, with one important difference. Density is a single value that represents the total number
of photons reflected or transmitted. Spectral reflectance is a set of values
that represent the number of photons being reflected or transmitted at
different wavelengths (see Figure 1-24). The spectrophotometers we use
in the graphic arts typically divide the visible spectrum into 10 nm or 20
nm bands, and produce a value for each band. Research-grade spectrophotometers divide the spectrum into alarger number of narrower bands,
sometimes as narrow as 2 nm, but they're prohibitivelyexpensive for the
types of use we discuss in this book.

F ~ ~ u1-94
H

A metameric pair

I
Measumfspectra of the GATFIRHEMUght Indicatol: which contains
patches that match only underD50 lighting.

Appearance under
D50 lighting

Appearance under
06.5 lighting

Chmyer 1: What Is Color?

45

Spectmphotomdryand color managcmmt. Spectral data has direct uses

in graphic arts-such as when some press shops check incoming ink lots
for differences in specttai properties--but spectrophotometers are more
often used in color management as either densitometers or colorimeters.
The spectral data spectrophotometers capture is a richer ser of measurements than those captured by either densitometers or colorimeters.
We can compute density or colorimetricvalues from spectral data, but not
the other way around. With spectral data, we can also determine whether
or not acolor match is metameric, though color management doesn't use
this capability directly
In color management, a spectrophotometer's real value is that it can
double as a densitometer or colorimeter or both, and it is usually more
configurable than the densitometer or colorimeter at either task. Sometimes a dedicated densitometer or colorimeter may be better suited
to a specific task such as measuring the very high densities we need to
achieve on prepress film, or characterizing the very spiky response of the
red phosphor in a monitor, But in most cases, a spectrophorometer is a
versatile Swiss Army knife of color measurement.

Where the Models Fail

The CIE colorimetric models are pretty amazing, but it's important to
bear in mind that they were designed only to predict the degree to which
two solid color swatches, viewed at a specific distance. against a specific
background, under a specific illurninant, would appear to match. Color
management takes these models well beyond their design parameters.
using them to match complex color events like photographic images.
In general, color management works astonishingly well, but it's important to realize that our visual system has complex responses to complex
color events that the CIE models we currently use don't even attempt to
address. A slew of phenomena have been well documented by-and
often named for-the various color scientists who first documented
them, but they typically point to one significant fact: unlike colorimeters
and spectrophotometers, humans see color in context.
So there's one caveat we'll make repeatedly in this book: while color
management uses colorimetry for purposes of gathering data about

46

R u l World Color Management, kcond Wlllon

device behavior, the ultimate goal of color management is not to get a

colorimetric match, but rather to achieve a pleasing image. Sometimes
a colorimetric match is great, but sometimes that match comes at the
expense of other colors in an image. And sometimes two colors can differ calorimetrically and yet produce a visual match when we view them
in context.
Features like simultaneous and successive contrast (see Figures 1-12
and 1-13) and color constancy (see the next section) aren't modeled by
colorimetryand can't be measured using a colorimeter. Sometimes it's an
advantage to have an instrument that isn't distracted by these issuessuch as when you are collecting raw data on device behavior to feed to a
color management system or pro6le maker. But when it comes time to
evaluate results, don't reach for your colorimeter as a way to judge success. A well-trained eye beats a colorimeter evety time when it comes to
evaluating final results.
Here are just a few visual phenomena that color management ignores.

Color Constancy
Colorconsmncy is one of the most important features of the visual system,
and it's so ingrained a mechanism that you're rarely aware of it. Color
constancy, sometimes referred to as "discounting the illurninant," is the
tendency to perceive objects as havinga constant color, even if the lighting conditions change. In other words, even if the wavelength composition (thespectral energy1 of thelight coming from the object changes, our
visual system picks up cues from surrottnding objectsand attributes that
change to the lighting, not to the object.
What may surprise you is how basic a feature color constancy is to
the nervous system. It doesn't involve memory, much Iess higher-level thought at all, but seems to be rooted in low-level structures in our
visual system. In fact, color constancy has been verified in animals with
as simpleanervous system as goldfish (seeFigure1-25). In humans, color
constancy seems to bethe result of center-sumund fieldssimilar to those
responsible for opponency, bur instead of occuning in the second layer
of the retina, as opponency is, the center-surround fields responsible for
color constancy seem to be located in the visual cortex of the brain, and
they're farmore complex than thoseresponsible for opponency.

%urn 1-9s
Color constancy

Color constancy hru men uemomrmrtu zn

goldfrh.

Devices don't have d o r constancy. Cameras don't have color constancy. Film doesn't change its response dependingon the illumination in the
scene. This is why a photographer has to match the film response to the
lighting. Digital cameras with automatic white balance do change their
response depending on the illumination in the scene, but they don't do
so in the same way humans do. If a digital camera captures an image of
a white horse standing in the shade of a leafy tree, it will faithfully record
the greenish light that's filtered through the leaves and then reflected from
the horse, producing a picture of a green horse. But humans know there's
no such animal as a green horse, so they "discount the illuminant" and
see the horse as white.
Similarly, colorimeterscan't measure color constancy. They duplicate
the tristimulus response of the eye to isolated colors while ignoring the
surrounding colors. But even when a device l i e a scanner does measure
the colors surrounding an isolated sample, the exact nature of color constancy is so complex that we don't yet have a usable mathematical model
that would let color management compensate for it.

Color constancy and color management.While color management does

not have a model of color constancy to work with, it can do many things
without it. The important thing for color management to do is to preserve
the relationshipsbetween colors in an image. This is the difference between what we call perceptual and colorimetric renderings (see Chapter
3, ColorManagement-How It Works).When it's possible to render some
but not all of the colors with complete colorimetric accuracy, it's often

48

R u l World Color Management, Second Edition

perceptually more pleasing to render them all with the same inaccuracy
than to render some faithfully and some nor.
Color constancy is one reason why neutrals are important: neutrals-especially the highlights in a printed picture that take their color from
the paper-form the reference point for colors. If the neutrals are off, the
entire image appears off, but itb hard to pinpoint why. It takes training to
"ignore" color constancyandsay,"the neutrals are blue" when yourvisual
system is trying to say, "the neutrals are neutral."
A final point about color constancy and color management: color constancy presents an argument that the color temperature of lighting isn't as
importantassome people think It still isimponant, but not at the expense
of everything else. For example, when calibrating your monitor, you usually have the choice between setting your monitor to a D50 or D65 white
point, (Monitor calibration, and the differences between D50 and D65
white points, are described in Chapter 6,but you don't need to understand
these details to understand this point.) Many people choose D50 in order
to match the exact white point color of the viewing environment, which is
usually a D50 Lighting booth. But we, along with many other practitioners,
recommend D65 because we think you'll be happier IookIng at a bright
white D65 monitor than a dingy yellow D50 one, even if the D50 monitor
is calorimetrically closer to the D50 lighting booth. Matching brightness
levels between two viewingen\monrneots may be as important as, or even
more important than, matching the color temperature--color constancy
does a lot to adapt to slight differencesin color temperature.

PsycholoskalWcborrCdorNamesandMemoyCdorz
Now we t u m to the psychological attributes ofcolor. T h e ~ involve
e
aspects
of judgment that are notwell understood. Some of these psychological attributes may belearned. Some may even be cultural But theseattributes
relate to the way we talkabout color in our language.

Names and color reproduction. Earlier in this chapter, we defined hue as

the attributeof acolorby whichit getsits basicname.Thisconnection between hue and basic names isn't Justa philosophical nuance. but may be
one of the most important things to remember about color reproduction.
If nothing else, it sets a minimum bar for reproduction quality. We're generally fussier about discrepancies in hue than we are about discrepancies
in brightness or saturation between a target color and its reproduction,

or between a displayed color and its print. If the hue is different enough
to cross some intangible boundary between color names-such as when
your reds cross slightly into the oranges, or your sky crosses a tad into the
purples-then people notice the hue shift more, perhaps because they
now have a way to articulate it. The good news is that this is often the first
step in solving the problem-by being able to put a name to the hue shift,
you can begin to look for the source of the problem (too much yellow ink
in the reds or too much magenta in the blues).

Memory colon. One effect that all graphic artists are aware of is that
of memory colors. These are the colors such as skin tones, green grass,
or sky blue that we are very familiar with. These colors matter more than
others because we have such a strong memory of them. As our friend
(and color-management critic) Dan Margulis l i e s to say, the color on
the left of Figure 1-26 is a color, but the one on the right is a COLOR. The
evaluation of the quality of a color reproduction is weighted by the fact
that some colors are more important to get "right" than others. We put
"right" in quotes because our ingrained memory of these colors is often
quite inaccurate. Even if everything is colorimenicallyperfect, if the skin
tones aren't what the viewer expects, the image looks "wrong"; if the skin
tones fit the expectation, but a sweater went a little too red, few people
will notice, unless the picture is in a sweater catalog.
In summary, there are psychological aspects of human color perception that we can't (yet)model mathematically,so color management simply can't address them. Even the best color management must leave room
for human intervention at strategic points. And you have to be aware
of where color management needs your help, as we'll be pointing out
throughout the book.
Flgun1-M
A color and a CnmR!

50

Real World Color Management, Second Edition

for Color Management

We conclude this long chapter on color science with a summary of some
of the main lessons we should keep in mind for color management.
b

Lightingisimportant ... our eyeis best designed to workundercertain

conditions ...

... but it's not absolute. The visual system has a tremendous ability to
adapt. (See the section, "Color Constancy.")

Watch out for fluorescence with certain papers or inks, or when making scanner profiles. (Seethe section, "Fluorescence.")

Metamerism is the basis for color matching and hence the basis for
colormanagement. (See the section. "Metamerism.")

Train yourself to see yourred, green, blue response.With practice, you

can recognize all colors as mixtures of red, green, and blue. (See the
section. "Trichromacy: Red, Green, Blue.")

Train yourself also to think in blue-yellow, red-green opponencies. If

you want to reduce the yellowin an RGB image, which has no yellow
channel, you can increase its opponent color, the blue channel. (See
the section, "Opponency: Red-Green, Blue-Yellow")

Watch for hue shifts that mweacolorintoadifferent color narnecategory. Thesewill be obvious to viewers and signal a significant problem. (See
the section "Psychological Factors: Color Names and Memory Colors.")

There are colors, and there are COLORS (COLORShave different values
to us emotionally).(Again,seethe section. "PsychologicalFactors: Color
Names and Memory Colors.")

Colorimetryis the basis for the quantificationof colors. Colormanagement is based on mathematical models of color closeness provided by
colorimetry (See the section, "Colorimetry.")

Colorimetry isn't perception. Relationships behueen colors are more

important than the colors themselves. This is an important point to
understand when we look at the difference between colorimetric and
perceptual renderings. (See the section. "Color Constancy.")

and Color
Color by the Numbers

Computers know nothing about color except what we humans tell them.
They're just glorified adding machines that juggle ones andzeros to order.
One of the many ways we use numbers on the computer is to represent
color. To do that, we need some kind of mathematical model of color.
Applying mathematical models to reality is always tricky, but it's particularly so when dealing with something as slippery and subjective as
i
rIsaac Newton, made many important
color. The great mathematician, S
discoveries about color, but as far as we know he never tried to model
it mathematically. Instead he went onto simpler subjects like inventing
calculus and discovering the mechanical laws of the universe.
In Chapter 1, we explained that color is really something that only happens in our headsit's the sensation we experience in response to different
wavelengths of light. When we talk about measuring color, what we're
measuring isn't really color itself, but rather the stimulus that evokes the
sensation of color-the makeup of the light hitting our retinas. We can
correlate light measurements with the color people experience, but the
correlation isn't perfect.
In this chapter, we'll examine the various number systems we use to
represent color, explain what these numbers mean, and showhav,without
color management, the same set of numbers will produce very dierent
colors in different situations.

59

Real World Color Management.

Sccond Edition

Color bv- the Numbers

In the previous chapter, we explained how it's possible to produce all the
colors people can see using only red, green, and blue light-the "additive"
primary colors. When we reproduce color on a physical device, whether
it's a monitor, a piece of transparency film, or a printed page, we do so by
manipulating red, green, and blue light.
In the case of m e RGB devicessuch as monitors, scanners. and digital
cameras, we work with red, green, and blue light directly. With BLm and
printing, we still manipulate red, green, and blue light, but we do so indirectly, using CMY pigments to subtract these wavelengths from a white
background-cyan absorbs red light, magenta absorbs green light, and
yellow absorbs blue iight-hence the term "subtractive" primary colors.
Most digital color is encoded to represent varying amounts of either R.
G,and B or C, M, and Y,or, in commercial printing and some (but not
all) desktop printers. C.M, Y, and K (for BlacK). (See the sidebar "Why
CMYK?")
Unfortunately, these mathematical models of color are quite arnbiguous. You can think of an RGB or CMYK file as containing, not color, but
rather a recipefor color that each device interprets according to irs own
capabilities. If you give 20 cooks the same recipe, you'll almost certainly
get 20 slightly different dishes as a result. Likewise. if you send the same
RGB file to 20 different monitors, or the same CMYK file to 20 different
presses, you'll get 20 slightly (or in some cases, more than slightly) different images. You can readily see this in any store that sells television sets.
You'll see 20 televisions all lined up, of various makes and models, all
tuned to the same station, and all producing somewhat different colors.
They'rereceivingthe same recipebut theirdifferent characteristicsgenerate different visible results.This even happens within the samemake and
model of television.
The RGB and CMYK models originated in the analog rather than the
digital world. Neitherwas designed as an accurate mathematical descriplion of color: they're really controlsignnlsthatwesend to our variouscolor
devices to make them produce something that we eventually experience
as color. So you should always think of RGB or CMYK numhers as tuned
for a specific device.

Chapter P: Computers and Color

Why CMYK?
\Vhy CM\X rather than CMY? In

theory, pure cyan absorbs 100% of

red light just as pure magenta and
yellow fully absorb green and blue
light respectively. A combination
of perfectly pure cyan. magenta, and yellow colorants would
absorb all light, which people
would see as black (if they could
see it all-theonly perfectly black
objectswe're aware of in this universe are black holes, which we
can't see directiyl. When one or
more colorants aren't 100% pure.
some light is reflected instead
of being absorbed. This is why
many toner-based devices have
a greenish three-color black.
and why three-color black on a
printing press is usually a muddy
brown. The colorants are simply
not perfect. Photographic dyes

come close, but inks and toners.

for example, have to satisfy many
diRerent physical requirements
besides color, such as adhering
to the paper and each other, drying in a reasonable length of time.
being fade-resistant, and being
affordable.This almost invariably
involves compromising the color
purity. So m geta better black that
will absorb as much light as possible, as neutrally as possible, we
use black ink. Another good reason for printing with black ink is
that black-only objects such as
text are a lot easier to print when
you don't have to perfectly align
the cyan, magenta, and yellow
versions.
You may also wonder why it's
CMYK rather than CMYB. There's
general agreement that refer-

ring to black as "B" wc~uldlead

to confusion with Blue. Press operators often refer to cyan and
magenta as blue and red, which
is, of course. incorrect and not
something we personally encourage. But it's an ingrained, timehonored practice, and trying to
change it is, as our friend and colleague Herb Paynter would say, "a
hill that aln'tworth dyin' on."
There are various theories as
Lo why " K was chosen, hut the
likeliest, in our vieh,, is that it refers to "key"-the master plate
to which the other three colors
are registered. Since black is the
darkest color, it's u s d y used as
the key because irk the easiest to
see. Whateverthereason, we print
with CMYK,not with CMYB.

Analog Origins
The numbers in RCB and CMYKfiles don't really represent color. Instead.
they represent the amounts of colorartrsthe things our devices use to
make a color. Both RGB and CMYK were used in the analog world long
before they were translated to the digital world.
CMYK printing has been around as a mass-market commercial process since the early 1920s, a n d until pre-press went digital in the 1970s,
CMYKseparations were made optically by photographing the original art
through C, M.Y,and neutral-density (for the black plate) filters. The earliest scanners used analog RGB signals. The scanners' RGB signals were
Qpically converted directly to analog CMYK,which was used to expose
film fromwhich printing platesweremade.When westarted making color
digitally, we simply used digitalRGB anddigital CMYK to mimic their analog predecessors. In short, it was the easiest way to make the transition to
digital color, but not necessarily the best way.

MonlW RGB. When we display color on a monitor, we do so by spraying

streams of electrons that strike phosphors. Phosphors are chemical and
mineral compounds that emit light when they're struck (the technical
term is excited) by a beam of electrons.Color monitors use three different
phosphorspainted on the inside of the faceplatethat emit red, green, and
blue light, respectively. By varying the strength of the electron beam, we
can make the phosphors emit more or less red, green, and blue light, and
hence produce different colors (see Figure 2-1).

Monitor phosphors
red light
green light

blue light
electron beam

I //

~1

red, gemgrerm
and blue

phosphors

But the precise color that the monitor produces depends on the type
of phosphors used, their age, the specific circuitry and other characteristics of the monitor, and even the strength of the magnetic field in which
the monitor is located. All monitors' phosphors produce something we
recognize as red, green, and blue, but there are at least five quite different
phosphor sets in common use, and the phosphors can vary substantially
even in a single manufacturing batch. Factor in individual preferences
for brightness and contrast settings, and it's highly unlikely that any two
monitors will produce the same color fromthe same signal, even ifthey're
two apparently identical monitors bought on the same day.

Scanner RGB.Whenwe capture colorwith ascanner or digital camera, we

do so using monochromatic light-sensitive sensors and red, green, and
blue filters. Each sensor puts out a voltage proportional to the amount

Chaptw P: Cornputen and Color

55

of light that reaches it through the filters, and we encode those analog
voltages as digital values of R, G, and B. The precise digital values a scanner or camera creates from a given color sample depend on the makeup
of the light source and the transmission characteristics of the filters. As
with monitor phosphors,scanner and camera filters vary fromvendor to
vendor, and they also change with age. Scanner lamps also vary both from
vendor to vendor and with age, and the light source in a digital camera
capture can range from carefullycontrolled studiolighting to daylight that
varies from exposure to exposure, or even, with scanning-back cameras,
over the course of a single exposure. So it's very unlikely that two capture
devices will produce the same RGB values from the same color sample.

Prlnkf urn.When we print images on paper, we usually do so by laying

down dots of cyan,magenta, yellow, and black ink. In traditionalhalftone
screens, the spacingis constant fromthe center of one dot to the next, but
the dots vary in size to produce the various shades or tints. Many desktop
printen andsome commercial press jobs use differenttypes of screening,
variously known as error d i f i i o n or stochastic screens, where each dot
is the same size, and the color is varied by printing a greater or smaller
number of dots in a given area (see Figure 2-2, and the sidebar, "Pixels,
Dots, and Dithers." on the next page).

CMYK halftone

In a wnventional
CMYKhalfone, dot
placement is regular;
and the illusion of
different tonal
values is produced by
varying the size
of the dots.

In a CMYKdiffuson
dither, thesize of
the dots is regular;
and the illusion
of different tonal
values is produced
by varying the dot
placement.

56

Real World Color Management, Second Wltlon

Pixels, Dots, and Dithers

Wherever possible we tr). to avoid
nitpickingabout terminology but
the practice of using the terms
"dpi" (dots per inch) and "ppi"
(pixels per Inch) interchangeably
inevitably leads to confusion, because dots and pixels are distinct
entities with different properties. Some digital mavens insist
on using the term %pi" (samples
per inch) rather than dpi when
discussing scanner resolutionthat's nitpicky.
Pixels represent varying levels
of density. A single pixel can not
only be red, green, and blue at the
same time, but have different lntensities of red, green, and blue,
rather than just on or off. This is
whereweget theterm continuous
rone. Amonitor is an example ofa
continuous-tone device.
But most digital hard-copyoutput devices are not cootiououstone. Instead of pixels, they print
dots of Ink or toner, which are

eitheron or off, present or absent

Digital printers' resolution is expressed in dots per inch, which
describes the number of locations
per inch in which the printer can
either print a dot, or not print a
dot. We can't vary the density of
ink, nor can we vary the size of
the dot. Rather. we can only tell
the output device whether or not
to print a dot at each location. A
GOO-dpi laser printer can print, or
not print, 600 dots per linear inch,
while a 2400-dpi imagesetter has
to make that same decision 2400
times per linear inch. Each dot
has the same dens@-the only
thing we can control is the location of the dots.
We make digital printers produce the illusion of continuous
tone by arranging these equally
sized, constant-density dots
using some kind of ditherinway of arrangingthe dots In a partern that isnP obvious to the eye.

Back in the analog days, prepress

folks converted continuouc-tone
originals to halffonesa kind of
dither where the doaare coostantdensity and equally spaced. varying in size to produce the illusion
of darker or lighter shades-11).
projecting the original onto plate
material through color filters and
a screen like the ones you find
on a screen door. The holes in the
screenactedas pinholelenses,producinglarge dots in the dark areas
and small ones in the light arras.
Most presses still use this kind
of dithering, but to make a digital halftone on an imagesetter or
platesetter, we arrange its prinlabledots into larger groups called
hnlfrone cells or halftone spots.
We simulate the traditional analog halftone by turning dots on
and off in acell.This type ofdither
is also known as "AM (Amplitude
Modulation) screening," or sun.
ply as a "conventional halftone."

But the precise color that the printer produces depends o n the color of
the inks, pigments, or dyes, t h e color of the paper stock, and the way the
colorants interact with the paper, both chemically and physically. Inkjet
printers commonly showcolorshifts overtime (most obviousin neutrals)
when ink and paper aren't appropriately matched. Color laser printers
a n d color copiers are very susceptible to humidity change. On a commernial press, the color can vary with temperature, humidity, prevailing
wind, and the state of the press operator's diet a n d marriage, but that's
another story! So it's very unlikely that two different printing devices will
produce the same color from the same set of CMYK values.

Chap-

Halftoningconvertscontinuous
toneimayes, suchasdigital filesor
scans, frompixeis into dots so the
image can be reproduced on an
inkjet printer or a printing press.
forexample.The best-knowntype
of halftoning is sometimes called
ordered ditlrer or "conventional
halftune dot" in addition to the
aforementioned "AM screening':
A different kind of halftoning,
errorrlifusion, also goes hy names
such as "FM screening" and "stochastic screening."
AM screening varies the size of
Lhe dots, but not their localion.
Darker areas havelarger dots and
lighter areas have smaller dots (or
no dots). This type of halftoning
is common for printing presses
as a carryover frommakingpiates
through a "screen."
Becauseofthe pattern generated with AMscreening,undesirable
artifacrs can occur with multiple
inks when their "screens" conflict

with each other. Combating this

necessitates rotating each ink's
screen to avoid these conflicts.
hence the term "screen angles."
The other kind of dithering.
error dilbsion or FM screening, is
used on most inkjet printer%and
occasionally on presses too. FM
screening varies the location of
the dots, but not their size. Darker
areas have more dots closer together while the lighter m a s have
fewer dots dispersed farther apart.
The more random nature of FM
screening gives Ngh-resolution
inkier output the appearance of
continuous tone. FM screening is
sometimesusedon press, but since
small variations in press behavior
are exaggerated much more by FM
screeningthanbyAM screening,it's
usually limited to premium jobs in
shops that have gained considerable experience using FM screens.
Dye sublimation and photographic output methods are con-

9: Computers and Color

57

sidered continuous tone because

the colorants' density can be controlled, so dots, and therefore halftoning, aren't needed.
The importance of this lo color
management is that ICC-based
color management only works on
pixels, not dots, but the final output is more often than not created
with dots. The effect of screening
algorithms can avect what we
see compared to what was
measuted and predicted by the
color managementsystem (CMSI.
So wr need to take the screening algorithm into account when
we color-manage a device, by en.
wring that we print the profiling
target using the samescreeningwe
plan touscforpmduction, because
ditierent screening algorithms will
producedifferenr tonal renderings.
(See "Tone Reproduction Characteristics," later in this chapter.)

Digital Evolutions
The poinlofthe previoussectionis that RGBand CMYKare fundamentally analog concepts-they represent some amount of colornms: the dyes.
inks, phosphors, or filters we use to control the wavelengths of light. RGB
devices such as televisions, monitors, scanners, and digital cameras to this
day, and for the forseeable future, all have analog components- things
that work m terms of continuous voltages: magnets, lenses, mirrors, and
phosphors and filters baked i n chemlcal labs. (=M\X printers still deal
with the idiospcraciesof chemical inks, dyes, and pigments o n sheets of
mashed wood pulp that we call "paper."

58

Real World Color Mimagcmmt, kcond Wltlon

However, RGB and CMYK numbers are nevertheless ... numbers. This
has made them ripe for adoption into the digital age where numbers
themselves take the form of bits and bytes (see"How the Numbers Work,"
on the next page).
Over theyears we've seen more and more analog components replaced
by digital. The Darwinian force that drives this evolution is, veery simply,
money. Digital components are faster, cheaper, and (most importantly
From the point ofview of color management) repeatable and predictable.
AU of these benefits translate directly into monetary savings.
But keep in mind two things about this evolution. First, it's incremental. Companies usually produce products that are small improvements
over previous technologies-despite what their marketing brochures say
New products and subcomponents of products must coexist with old.
and new technologies must be usable by people who have worked with
old components foryears. Second. because of this incremental evolution.
digital RGB and CMYKare often designed tomimic theiranalogpredecessors. The msult has been that digital color-reproduction equipment often
has odd little idiosyncracies that might not be there if the things had been
designed from the ground up to be digital, or mostly digital, devices.
An example is the evolution of the imagesetter, the output device that
generates the film used to image plates for offset lithography. lmagesetters evolved from the analog methods of imaging film using photographic
techniques-for example, projecting a photographic negative through a
fine screen to produce a halftoneimage (hence the terms "screening" and
"screen frequency"). This analog photographic process was replaced by
computers controlling lasers that precisely exposed the film microdot by
microdot, but these imagesetters stillneeded traditionalanalog darkroom
equipment, chemistry, and photographically skilled technicians to develop the film. However, bit by bit, even the darkroom processing was
replaced by digitally controlled processor units that control all aspects of
film development.
But why did we need film at all? Because extremely expensive (and
therefore not expendable) printing presses had almost as expensive
platemakers that required film for platemaking, and because creating
an analog proof from the film was the only affordable method in place
for creating a contracr between the print client and printer (premium
jobs sometimes use actual press proofs-in effect, separate press runs
for proofing only-but they're bn~tallyexpensive). Nowadays, however,

Chrptn P:Compukn and Color

59

as platemaking and digital proofing technologies become more reliable.

film itself is being skipped (with obvious cost benefits), and the digital
process is being extended from the computer tight up to the platesetter
itself, even including digital platemakers that image the plates right on
the press rollers.
What does all this mean for digital RGB and CMYKnumbers and color
management? It means that all the digitalcomputation and control of the
numbers exercised by color management are only as goodas their ability
to model the behavior of analog components. Digital color management
alters the nirnabers to compensate for the behavior of the various analog
components.As such, the strengths and weaknesses of color management
lie entirely in how well our digital manipulations model the behavior of
the analog parts, including that most important analog "device" of all,
the viewer's eye.
In a moment we'll examine the key parameters that describe the analog
behavior of color-reproduction devices. But fint, we should examine the
digital part of digital color-the numbers-a bit more closely.

How the Numbers Work

Let\ pause for a mument and examine the systems we use for representing-or, more accurately, encorling--colors as numbers in a compurer.
We'll take this the opportunity to clarify a few points that often confuse
people about the basics of digital color, which propagates into confusion
about color management. Even if you're extremely familiar with the basics of bits, bytes, tones, and colors, this section is worth reviewing as we
make a few key points about the difference between colors-as-numbers
and colors as "Real World" experiences.
The system computers use for encoding colors as numbers is actually quite simple: colors are comprised of channels and each channel is
subdivided into tone lewls. That's it!We start with a simple model of color
perception-the fact that colors are mixtures of red, green, and blue in
this model for efficient storage.
various i n t e n s i t i e ~ a n dthen we a d a p ~
computation, and transponation on computers.The number ofchannels
in our encoding system is usually three, to correspond to our basic threeprimary way of seeing colors.The number of tone levels in our encoding
system is usually 256,to correspond to the minimum number of tone levels

60

RfaI World Color Management, kcond Wltion

we need to create the illusion of continuous t o n e t o avoid the artifacts

known as banding or posterization, where a viewer can see noticeable
jumps between one tone level and the next (see Figure 2-3).

With only 128 shades

of gray. some tonal
derail is lost, and we
start to see hints of
oosterization.

With only 64
shades ofgray. we
see banding in the
gradients and
in the sky.

Chapter P:Computm and Color

61

Why 956 Levels?

This number, 256, seems arbitrary and mysterious to some people, but
it crops up so many times in computers and color that it's worth making
your peacewithit. lt'snotthat mysterious. Wewant to beable to represent
enough tone levels so that the step from one tone level to the next is not
visible to theviewer. It turns out that the number of tone levels needed to
produce the effect of a smooth gradient is about 200 for most people. So
why not encode only 200 levels? Why 256? For two reasons.

Headroom. It's useful-in fact, essential For color management-to have

some extra tone levels in our data so that the inevitable losses of tone
levels at each stage of production (scanning,display, editing, conversion,
computation, printing) don't reintroduce banding.

Bits. The second reason is just that we use bits to represent these tone
level numbers. Seven bits would let us encode only 128 tone levels e 7 ) ,
which would be a surefire way to get banding in our skies and blotches
on the cheeks of our fashion models. Eight bits lets us encode 256 tone
levels (29, which gives us just enough, plus a little headroom. The third
reason we go with eight bits is that computer storage is already organized
in terms of bytes, where a byte is a unit of exactlveight bits.This quantity
of eight bits is already so useful-for example, it's perfect for storing a
character of type, which can be any of 256letters and punctuation marks
in a western alphabet-that it seems a cosmic coincidence that a byte is
also the perfect amount of memory to encode tone levels for the human
visual system. Engineers love those cosmic coincidences!

Millions of Colors
So &bit encoding, with its 256 tone levels per channel, is the minimum

number of bits we want to stote per channel. With RGB images, storing
eight bits for each of the three channels gives us24 bits total (whichis why
many people use the terms %bit color" and "24-bit color" interchangeably to mean the same thing). The number of colors encodable with 256
tonelevels ineachofthree channels is256 x 256 x 256,or (if you pull out
your calculator) about 16.8 million colors! Quite a lot of encodable colors
for our 24 bits (or three little bytes) of storage!

69

Red World Color Management, keond Edition

Although this basic 3-channel. &bit encoding is the most common because it's based on human capabilities,we can easily expand it as needed to
encode more colors for devices other than the human eve, either by adding
channels or by increasing the number of bits we store for each channel.
For example, when we're preparing an image for a C m printer, we increase the number of channels from 3-channel to 4-channel enmdh7g, not
because we need more encodable colors (in fact, we need fewer) but
because irk natural to dedicatea channel to each of the four inks.
Similarly, we often go from &bit to 16-bit encoding when saving
images captured with a scanner capable of discerning more than 256
levels of RGB (the so-called "10-bit," "12-bit," and "14-bit" scannersalthough, becausewe storefiles inwholebytes, thereareno 10-bit. 12-bit,
or 14-bit files, only %bit or 16-bit files).
A key point to remember is that this is all talking about the number
of encodings, the set of numeric color definitions we have available. But
just as in the San Francisco Bay Area there are far more telephone numbers thanthereare actual telephones, with computer color the number of
encoclnblecolorsfar exceeds the number of reproduciblecolors.In fact, it
even ifwe makedevices
far exceedsthe numberof perceit~ablecolors.And
such as high-end scanners that can "perceive" more tone levels than the
human eye, we can always expand our encoding model to handle it. All
that matters is that each perceivable color has a unique encoding, so there
are always more encodable colors than we need-just as the phone company must ensure that eveq telephone has a unique telephone number.
so it had better have more telephone numbers than it really needs.
We make this point because it's a key step to understanding the difference between colors as abstract numbers and how those numbers are
actually rendered as colors by "Real World devicesprinten, monitors,
scanners, etc.When youlookat how thosenumbersareactually interpreted
by a device, the number of actual "Real World" colors, drops dmmatically!
(See the sidebar "Color Definitions and Colors.")
So while it's useful to understand how the numbers work-why we see
numbers like 256 or 16.8 million crop up everywhere-don't forget that
they're just numbers. .. until they're inferpreted by a color device as colors.
In the next section we'll look at what gives the numbers a precise interpretation as colors.These are the analog parts of our color devices, the
things that color management systems need to measure to know how to
turn number management into color management.

Chapter P:Compulrn and Cdor

Lots 01 people confuse the

number of color definitions
with the number of colors. For
example, we mentioned that
color7 intended for a CMYK
printer are naturally encoded In
four channels. Does 8-bit CMYK
really represent 256 X 256 x
256 x 256 or 4.3 billion encodable colors?! Theoretically, yes.
Four arbitrary channels produce
4.3 billion encodings, but when
we assign the interpretations C.
M. Y, and K to those four channels, we realize that the fourth
channel (K) doesn't seem to add
many colors. In fact. many of the
CMYK encodings represent the
same color. For example, 50C,
50M. 50Y. OK theoretically encodes the same shade of gray as
OC, OM. OY.50K, so there's a lot of

rrdundancy. Nowone could argue

that the extra Kvalues contribute
additional tonal levels to rheCMY
channels, but thisstarts to get way
more complicated than we need
to get here. Let3 just say that the
totd number of encodable colors
in &bit CMYK is far less than 4.3
billion.
As another example, we mentioned that there are scanners
that claim to be able to see far
more than the 256 tone levels
encodable in eight bits. These
tout 10-bit. 12-bit. or even 14-bit
capability. Many people confuse
this with the scanner's djmmic
range, the range from brightest
white to darkest dark in which
the scanner can distinguish tonal
variation reliably. Some scanner vendors may actually claim

63

that these "high-bit" scanners

give a larger dynamic range than
8-bit ones. This is nonsense--the
dynamic range is an analog
limitation of the capture device
and has nothing whatsoever to
do with hit depth. Higher bit
depths simply allow us more
editing flexibility by slicing the
device's dynamic range Into more
discrete steps. You can think of
dynamic range as the height of
a staircase, and bit depth as the
number of steps that staircase
contains. Obviously, If we want
to keep the steps as small as
possible (which we do, to avoid
posterization or banding), a
higher dynamic range needs
more steps than a smaller one.
but there's no direct relationship
between the two.

the Numbers Vary

In the cotningchaptrr~\*e'll lookat thespecificsof measuringthe behavior
of display devices (monitors), input devices [scanners and digital c a n eras), and output devices (printers and proofingsystems), but here w e n
look briefly a t the basic parameters that vary from device to device.
All devicesvary incertain basic parameters. Theseare the things you'll
measure if you are making your own profiles, a n d which must remain
stable for your color management system to work effectively.
The three main variables are
b

The color and brightness of the colorants (primaries).

The color and brightness of the white point and black point.

The tone reproduction characteristics of the colorants.

64

Real World Color Management, Second Wltlon

These concepts aren't new to color management or unique to digital

devices. They're all variables introduced by analog components like inks
on paper, phosphors and analog voltages in monitors, and filteredsensors
in scanners. While digital components rarely vary much, analog components vary a great deal in design, manufacturing, and condition.

Colorants (Primaries)
The first, most obviousfactorthat affects the color adevice can reproduce
are the colorants it uses to do so. On a monitor, the primaries are the phosphors. In a scanner or digital camera, the primaries are the filters through
which the sensors see the image. In a printer, the primaries are the process
inks, toners, or dyes laid down on the paper, but, because the subtractive
color in CMYK printers is a bit more complicated than the additive color
in RGB monitors, we usually supplement measurements of the primaries
with measurements of the secondaries (the overprints-Magenta+YeIlow,
Cyan+Yellow,and Cyan+Magenta)as well (see Figure 2-41,
The exact color of the colorants determines the range of colors the device can reproduce. This is called the color gamut of the device. We care
not only about the precise color of the primaries, but also howbright they
are. In technical terms we oftenrefer to the densityof the primaries,which
is simply their ability to absorb light.

Subtractive primary and

secondary colors

Chapter P:Cornputen and Color

65

White Point and Black Point

Besides the primaries, the other two points that define the gamut and
hence need to be measured and monitored in adevice are the tuhitepoinr
and black point. Books (and people) often talk about the white point and
black point in very different terms: with the white point, they're usually concerned with the color of white, while with the black point they're
more concerned with the rlerzsity (thedarkness) of black. In fact, we can
talk about both color and density of either the white point or the black
point-the difference is only amatter ofemphasis.b'ith the white point,
the color is more important than its density and with the black point. the
density is more important than its color.
The color of the white point ismore important than its density because
the eye uses thecolorof this white as a reference forall other colors.When
you view images, the color of white on the monitor or the color of the
white paper on a printed page affects your perception for all the other
colors in the scene. This white point adaptation is an instantaneous and
involuntary task performed by your eye. so the color of whiteis vital.This
is why, as we'll see in Chapter6 we often sacrifice some brightness during
monitor calibration to get the color of the white point c o m c t Similarly.
when lookingat a printed page. ~t'simportant to rememberthat the color
of the white point is determined as much by the light that illuminates the
page as by the paper color itself.
With black point the emphasis shifts toward density as a more important variable than color. This is because the density of black determines
the limit of the dynamic range, the range of brightness levels that the device can reproduce. Getting as much dynamic range as possihle is always
important, as this determines the capacity of the device to render detail.
the subtle changes in brightness levels that make the difference between
a rich, satisfying image and a flat, uninteresting"misrake."
On amonitor, we try to calibrate so that we get just enough brightness
differences near the bottom end to squeeze some extra detail out of our
displayed shadows.
On sprinter, we can improve both the color and the densityof the black
point by adding our friend K to the CMY colorants. Adding K lets us producearnoreneutral blackpoint than we could with our somewhat impure
C, M, andYinks,and usingfour inks rather than three gets us much denser
(darker) blacks than we could using only C. M, andY.

66

R u l World Color Managemat, Second Ldllion

On a scanner, the softwarelets you change the density of the white and
black points from scan to scan either manually or automatically, but for
color management we need to use a fixed dynamic range,so we generally
set the black and white points to the widest dynamic range the scanner
cancapture.
Measuringthe color and densityofthewhiteand blackpointsisusually
one of the steps in preparing to use any device in a color management
system. You also need to watch out for changes in thesevalues over time.
so that you know when to adjust either the device itself, or pour color
managemenl system, accordingly.

Tone Reproduction Characteristics

Measuring the precise color and density oftheprimaries, white point. and
black point is essential, but these points only represent the extremes of
the device: the most saturatedcolors, the brightestwhites, and the darkest
darks.To complete the description of a device,the color management system also needs to know what happens to the "colors between the colors"
(to paraphrase a well-known desktop printer commercial).
There are several ways to measure and model devices' tone-reproduction
characteristics.Thesimplest,called a ronerepmcluctionnme(TRC1,defines
the relationship between input values and resulting brightnessvalues in a
device. Most analog devices have similar curves that show gain (increase)
in the darkness levels that Mect the midtones most and taper off in the
hightlights and shadows. In monitors, scanners, and digital cameras this
is called a gamma curve. Printers exhibit a slightly different dotgain curve.
but the two are similar (see Figure 2-51.
Some printers havemuch mnrecomplicated tonal responses that can't
be represented adequately by a simple curve. In these cases, we use a
lookup table (LUT), which records representative tonal values from light
to dark.
When you take measurements inthe process of calibrating or profiling
a device for color management, you're measuring the tone-reproduction
characteristics of the device as well as the primaries and blacklwhite
points.You should try to get a feel for the things that can alter these tonal
characteristicssuch as changing to a diffemnt paper stock,oradjusting the
contrast knob on your monitor-becausewhenadevice'stone-reproduction
characteristicschange. you'll have to adjust your color management system
to reflect the changes

Chapter P:(omputen and (dor

67

Flgun 9-5

Tone reproduction curves

Both diagrams plot the
input value (uertical axis)
against the output ualue
Olorizontal axis).

gamma curve

dotgain curve

Device-Specific
Color Models
We call RGB and CMYK device-specificor device-dependent color models,
because the actual color we get from agiven set of RGB or CMYKnumbers
is specific to (it depends on) the device that's producing the color. Put
simply, this means two things:
F The same set of RGB or CMYKnumberswill produce different colors
on different devices (or on the same device with different paper, if it's
aprinter or a press--see Figure 2-61,
F To produce the same color on diierent devices, we need to change the
RGB or CMYK numbers we send to each device--see Figure 2-7.

The problems we face as a consequence are

F How do we know what color the numbers in an RGB or CMYK file are
supposed to represent? In other words, what does "255,0,OVmean?
Yes, it means "red," but precisely what red? The red of your monitor,
or Chris's or Fred's? The red sensor in your scanner or BN& digital
camera?
F How do we know what RGB or CMYK numbers to send to a device to
make it produce a desired color?In other words, even if we know pre-

cisely what "red" we're talking about, what RGB numbers do we send
to Chris's monitor, or what CMYK percentages do we send to Fred's
color laser printer to reproduce that precise red ... if it's possible to
reproduce it at all?

68

Real World Color knagclmnt, Second Edition

The images contain

identical numbers ...

Same numbers,
differentcolor

II
!

I
I
I

...

buthave verydiffemnt
appearances. This is why we need
color management-one image's
appearance is much more desimbk
than theother;but withoutcolor
management you mn't reN
what color the numbers represent.

CI
The images contain

Same color,
differentnumbers

... but have identical

appearance. This is how w use
color management-we change the
numbers so that each of our devices
produces thedesiredappearance. &=

- --.- -

Chaplet P: Compulm and Color

69

Color management systems allowus to solve both problems by attaching absolute color meanings to our RGB and CMYK numbers. By doing
so, the numbers cease to be ambiguous. Color management allows us to
determine the actual color meaning of a set of RGB or CMYK numbers,
and also lets us reproduce that actual color on another device by changing
the numbers we send to it. But to do so, color management has to rely on
a different kind of numerical model of color, one that's based on human
perception rather than device colorants.

Device-Independent Color Models

Fortunately, we have several numerical models of color that are deuiceindependent. Instead of using the numbers required to drive a particular
device to produce color, device-independent color modelsattemptto use
numbers to model human color perceplion directly.
The device-independent color models in current use are based on
the groundbreaking work we mentioned in Chapter 1 by a body of color
scientists and technicians known as the Commission lntenmtionole iie
I'Eclairoge,or CIE-in English, the name means "International Commission on Illumination"-and the CIE is the international standards body
engaged in producing standards for all aspects of light, including color.
In 1931, the CIE produced a mathematical model of color with the
formidable-sounding name CLE XYZ (1931).This model was unique in
(hat it tried to represent mathematically the sensationof color that people
with normal color vision would experience when they were fed a precisely
defined stimulus under precisely defined viewing conditions. Since that
original workwas done, the CIE hasproducedawildalphabet soupofcolor
models with equally opaque names--CIE LCh. CIELW CIE xyY, CIELAB.
and so on, all of which are mathematicalvariants of CIEXrZ
You don't need to know the differences between the various models in
order to use color management effectively. In fact, outside of measuring
the colors a device produces as part of the profiling process, you needn't
deal with any of the CIEmodelsdirectly. But it isimportant to understand
the distinction between device-dependent models l i e RGB and CMYK,
and device-independent models like CIE XYZ and CIELAB.

70

Real World Color Manasement, Second Edition

RGB and CMYK just tell machines how much colorant to use: they tell
us nothing about the actual color the machines will produce in response.
The CIE models describe the specific color that someone with normal
colorvision wouldsee under veryprecisely describedviewingconditions.
bul tell us nothing about what we need to do to make a particular monitor, scanner, or printer produce that color. To manage color in the real
world, we need to use both device-independent and device-specificcolor
models.

CIE LAB
The ClE color model you're mast likely to interact with is CIE LAB (LAB).
You can actually save images in the LAB model, and edit them in Adobe
Photoshop. Heidelberg's LinoColor, and several other applications. LAB
also plays a central role in color management, as you'll learn in the next
chapter.
If you've ever tried editing a LAB file in Photoshop, you've probably
concluded that LAB is not the most intuitive color space around. It is.
however, based on the way our minds seem to judge color. It uses three
primaries, calledLL[pronouncedUL-star"),
a', and b*. La represents lightness, a* represents how red or green a color is, and b* represents how
blue or yellow it is. (Remember, red-green and blue-yellow are opponent
colon-they're mutually exclusive. There's no such thing as a greenishred or a bluish-yellow)
LAB, by definition, represents all the colors we can see. It's designed
to be perceptually uniform, meaningthat changing any of the primaries
by the same increment will produce the same degree of visual change. In
practice, it's not perfect, but it ispretty darn good, and more to the point.
nobody has as pet presented an alternative that is both a clear improvement and can be implemented using the computing power available on
today's desktop. Many of the problems we have with LAB stem from the
fact that we use it to do things for which it was never intended (see the
sidebar, "LAB Limitations").
Despite its flaws, CIE IAR allows usto control our color as it passes from
one deviceto another by correlatingthe device-specific RGB or CMYKvalues with the perceptu* based LAB values that they produce on a given
device. LAB acts as a form of universal translation language between devices, or, as Bruce is wont to say, "a Rosetta stone for color." It allows us to
express unambiguously the meaningof the colon we're after.

Chapter 9: (omputen and Color

7l

LAB Limitations
At some point in your color management travails, you'll almost
certainly run into a situation
where CIE colorimetry says that
two colors should match, but you
see them as being clearly different. LAB does have some inherent flaws. It's not as perceptually
uniform as it's supposed to be. It
also assumes that colors along a
straight hue-angle line will produce constant hues, changing
only in saturation. This assumption has proved false, particularly in the blue region, where a
constant hue angle actually
shifts the hue toward purple as
blue becomes less saturated. But
it's also helpful to bear in mind
the purpose for which LAB was
designed.
The design goal for LAB was to
predict the degree to which two
solid color samples of a specific
size, on a specific background

LAB imperfections

color, under very specific Lighting, at a specific viewing distance and angle, would appear
to match to someone with normal color vision. It was never
designed to take into account
many of the perceptual phenomenawe coveredin Chapter 1, such
as the influence of surround colors. Nor was it designed to make
cross-media comparisons such
as comparing color displayed on
a monitor with color on reflective
hard copy.
Yet color management systems try to make LAB do all these
things and more. When we colormanage images, we do so pixel
by pixel, without any reference to
the surrounding pixels [the context) or the medium in which the
pixels are finally expressed (dots
of ink on paper or glowing pixels on a monitor). So it's not entirely surprising that the model

occasionally breaks down. If

anything, it's surprising that it
works as well as it does, considering the limited purpose for which
it was designed.
The good news is that the theory and practice of color management are not dependent on LAB,
even though most color management systems today use LAB as
the computation space. The basic theory stays the same and the
practicethe"RealWor1d" part of
this book-does not depend too
muchon the computation model.
As developers come up with fixes
and workarounds to LABk flaws,
as well as alternative color models, they can just swap out LAB like
an old trusty car engine that
has sewed us well. In the meantime, LAB continues to be the
worthy workhorse of the colormanagement industry.

Thisfigure shows the a,b

plane of LAB, where hue
is shown by the angle, and
saturation by the distance
from the center; which
represents neutral gray
(a and b both equal 0).
In theory, the dashed
lines should represent
constant hues, but actual
constant perceived hues
produce the solid l i 7 1 ~ ~
shown in thefigure.
The hue ship is most
pronounced in the
red and blue regions.

72

Real World Color Management, Second Ldltlon

Alphabet Soup
Accord~ngto the citations we've
been able to turn up, the official
names are CIE X I 7 (19311, abbreviated as CIE m.CIE L'a'h*
(19761, abbreviated as CIELAB,
and CIE L*uV (1976). abhreviated as CIELUV. We haven't
found a canonical abbreviation
for CIE LCh (the primaries are

referred to as L* Cuvhuvl or CIE

CIE LCh, and CIE xyY,hut us-

ing rhe " superscript or the uv

subscript is excessively fiddly
If you want to be pedantic, (and is also technically incorrect)
CIE Lxa"b' and CIE L*u8v*are unless y0ut.e using the long name
incorrect. No one seems to have of the space, or referring to the inanv serious objection to using dividual primaries. While we're at
CIE Lab and CIE Luv. which are it, how many angels can danceon
mbre consistent with CIE XYZ, the head of a pin?
xyY, so we abbreviate them asCIE
LCh and CIE xyY.

But there's another aspect to the device-dependence problem that

color management addresses.

Mismatching-the Limits of the Possible

We use color management to reproduce faithfully the colors in our source
file on one or another target device--a monitor, a printer, a film recorder,
or a commercial press. But it's often physically impossible to do so, because each of our devices is limited by the laws of physics as to the range
of tone and color it can reproduce.

Device Limitations-Gamut and Dynamic Range

All our ourput devices (both prmters and monitors) have a fixed range of
color and tone that they can reproduce, which we call the devices' color
gamut. The gamut is naturally limited by the most saturated colors the
device has to workwith, its primaries. You can't get a more saturated red
on amonitor than the red of themonitor's phosphor.You can't get a more
saturated green on a printer than you can gel by combining solid cyan
and yellow ink.
Output devices (printers and monitors) also have a finite dvnamic
mnge-ther rangeof brightness differences they can reproduce. On amonitor the darkest black you can display is the black that results if you send
RGB value 0.0,Oto the monitor (which, if you have turned the brightness
control too high, may be quite a bit brighter than all the phosphors completely off). The brightest white on a monitor is the white you get when
all three phosphors [R,G, and Bl areglowing at maximum--althoughyou

Chapter 9: Computm and Color

73

usually have to calibrate your monitor to a less-bright white point in order to get a more color-accurate white. On a printer, the brightest white
you can render is the whiteness of the paper, and the darkest black is the
highest percentages of the four inks you can print on top of each other
without resulting in a soggy mess (usuallyconsiderably less than all four
inks at 100%).
Input devices (scanners and digital cameras) don't have a color gamut
because there is no sharp boundary between colors that thev can "see"
and colors that they can't-no matter what you put in front of them,
they're going to see something. Instead, we say thev have a color mLxing
fitmion, the unique mixture of red, green, and blue values that they will
produce for each color sample.This leads to aproblem h o w n as scanner
metamerism, which we described in Chapter 1, hut is not the same as the
gamut issue we are describing here.
However, although scanners don't have a specific gamut, we can often think of the effective "gamut" of the materials-usually photographic
prints or transparenciesthat you scan with the scanner, and this effectivegamut isusually much wider than any output device you will be using
to reproduce these scans. Digital cameras don't have a fixed gamut since
they capture color directly from the real world and have to cope with it in
all its multi-hued glory.This makes them tricky to profile.
Although input devices don't have a tixed gamut, they do have a fixed
dynamic range, the range of brightness levels in which the scanner or
digital camera can operate and still tell brightness differences. Below a
certain level of darkness--or densitv--a scanner or digital camera can no
longer distinguish between brightnesslevels, and just returns the value 0,
meaning"man, that's dark!" Similarly. above a certain level of brightness.
the device can't capture differences in brighmes+rarely a problem with
scanners, but all too common with digital cameras. Inputdevices typically
have a wider dynamic range than we can reproduce in our output.
This difference between device gamuts and dynamic ranges leads to a
problem. Ifour original image has awider dynamic range or a wider color
gamut than our output device, we obviously can't reproduce the original
exactly on that output device.
There's no single "correct" solution to this problem of variable gamuts
and dynamic ranges. You "can't get theah fiom heah," as they say in New
England. so you have to go somewhere else instead.

74

Real World Color Managrmmt, b o n d EdiHon

Tone and Gamut Mapping

The dynamic range ot our prmters, be they desktop color printers, film
recorders, or printing presses, is limited by the brightness of the paper
at the highlight end, and by the darkest black the inks, dyes, or pigments
can produce on that paper at theshadow end. A h recorder has a wider
dynamic range than an inkjet printer, which in turn has a wider dynamic
range than a printing press. But none of these output processes comes
close to the dynamic range of a high-end digital camera. Even a film
recorder can't quitematch the dynamic range of film exposed ina camera.
So some kind of tonnl compressionis almost always necessary.
Similarly, a film recorder has a larger gamut than an inkjet printer,
which again has a larger gamut than a printing press, and the gamuts of
all three are smaller than the gamut of film, so we need some strategy for
handling the out-of-gamut colors.
Gamut size isn't the only problem. We typically think of a monitor as
having a larger gamut than CMYK print, and this is true, but monitor
gamuts don't wholly contain the gamut of CMYK print. Even though the
monitor has a larger gamut, there are some colors we can produce with
CMYKink on paper that monitorssimply can't display, particularly in the
saturated cyans and the blues and greens that lie adjacent to cyan. So the
mismatches between various device gamuts can be as much due to the
shapes of those gamuts as to their size-see Figure 2-9.
Color management systems use various gamut-mapping strategies
that let us reconcile the differing gamuts of our capture, display, and output devices. but it's important to recognize that not only is there no "correct" way to handle out-of-gamut colors, it's also pretty unlikely that any
gamut-mapping strategy, or any automatic method of compressing tone
and color, will do equal justice to all images. So colormanagement doesn't
remove the need for color correction, or the need for skilled humans to
make the necessary decisions about color reproduction. As we've said
before, perception of color is uniquely human, and its judgment is decidedly human aswell.What color management does do is to let usview color
accurately and communicate it unambiguously, so that we have a sound
basis on which to make these iudgements.

[email protected] Gamut plots,gamut sizes,and gamut mismatches

There are various ways to illustrate and comparegamuts, some of which are more complete than others.
Two-dimensional gamut representations are invariably incomplete or misleadinethey show either a slice
of thegamut at a single 1ightm.s level or they show a projection of maximum saturation at all levels.
XI0
9 ,

-1

Lab

XI0

Two-dimensionalxyYplots exaggerate some

color regions and downplay others, because
xyYisn't perceptually uniform.

Two-dimensional LAE plots, while perceptually

uniform, are incomplete because they can only
show a single lightness level.
In reality, gamuts are complex
three-dimensional shapes. You
can tell a great deal morefrom
a three-dimensional gamut plot,
such as the oneshown at left,
than you canfrom any of the
m-dimensional gamutplotting techniques.

In this three-dimensional LAB gamut plot, the Lightness axis is

vertical, the horizontal a axis runsfrom red to green, and the
horizontal b axis runsfrom yellow to blue.

76

Real World Color MalugafIIWIl, Second Edition

Agure 9-9

Gamut plots, gamut sizes, and gamut mismatches, continued

When we use plots to compare two deuices'gamuts, the differences in plotting techniques become more
critical. All three plot types illustrated below show us that the typical monitor has a larger gamut than a
sheerfed press, containing many unprintable colors, particularly in the reds, but that the press can also
reproduce some colors that the monitorcannot.
vrl
-1
XI0

ThexyYplotatrightshows us that thepress

gamut conrnins a cyan-green region and a
smalleryellow-orange region that the
monitor can't display, but it exagerates
the size of both while minimizing the
monitor's unarintable reds. v

2
r Io

1
8

T y p i i l Monito!

S~~

PWS

7
6

The LAB plot at left provides a more accurate

depiction of the relativesizes of both the
gamuts themselves and of the regions where
they mismatch, but still doesn't tell the
whole story.

The three-dimensional LAB plots reveal the exact nature of the gamut mismatch, showing the monitor's
unprintable blues and the press's undisplayable light yellows, lef, and the monitor's unprintable reds and
light greens, and the press's undisplayable darkgreens and cyans, right.

Chaplei P:Cornputem and Cdor

Flgun 9-9

77

Gamut plots, gamut sizes,and gamut mismatches, continued

In some cases, such as the one shown below, two-dimensional plots may simplygiue the wrong
answer;as in this comparison between a sheetfed press and an Epson Stylus Photo 2200 printer
(you might want to make this kind of comparison ifyou were considering using the Epson inkjet
as a proofingdevice, for example).
XI0

- 1

Vn,

Lab
X10

Both 2-D plots indicate that the Epson inkjet'sgamut completely encompasses that of the press.

A 3-D comparison tells a rather different story.While the inkjet has a good overall match to the sheetfed press,
it can't reproducesome of the dark, saturnred colors of which the press is capable--which would be important
ifyou were printings shoe catalog,for example.When you really know your way aroundgamut plots, you'll
also conclllde that a whiter paper stock in the inkjet will provide a closer highlight match to the press.,

78

Real World Color Hanagrmmt, Second Wition

Agood manytools are available for visualizing gamuts in both two and
three dimensions, but the one we always keep coming back to is Chromix
ColorThink,developed by our friend and colleagueSteve Uptnn. (We used
it to make most of the gamut plots that appear in this book.) If you're the
type who finds a picture worth a thousand words, and you'd like to be
able to see representations of device gamuts and color conversions, we
think you'll find that ColorThink's graphing capabilities alone are worlh
the price of admission (though it does many other useful things too). It's
available at www.chromix.com.
One final point about gamuts: a device's color gamut isn't the same
thing as a device's color space. The gamut simply represents the l i m i t s
the whitest white, the blackest black, and the most saturated colors of
which the device is capable. A device's color space includes not only the
gamut boundar)., but also the tonal information that tells you what goes
on inside that boundary. For example, a newspaper press will likely have
a fairly large discontinuity between paper white and the lightest actual
color it can lay down, because newsprint can't holdsmall dots of ink.That
fact isn't conveyed by the gamut, but is a property of the device space. So
the gamut is one of theimportant properties of adevice's colorspace. but
not the only important property.

Color Is Complex
If all this sounds dauntingly complicated, that's because we're laying out
the complexity of the issues color management must address. Actually
tisinga color management system isn't that complicated, despite what
sometimes seems l i e aconspiracy on the part of applications vendors to
make it appear that way. But if you don't understand the underpinnings
of color management. it may seem like magic. It isn't magic, it's just some
pretty cool technology,and in the next chapter we'll look in detail at how
color management actually performs the complex tasks of changing the
numbers in our files to make the color consistent.

Color
Management
How It Works

For 1400 years, reading Egyptian hieroglyphics was a lost art. In 1799, a
chance discovery by a soldier in Napoleon's conquering army changed
all that. The Rosetta stone, as his find was named, allowed the brilliant
French linguist Jean Franqois Champollion to unlock the secret of hieroglyphics because it contained the same text in three different scripts,
hieroglyphics, demotic script, and Greek, the last two of which were
already known.
Depending on your background and training, RGB or CMYK numbers
may seem about as comprehensible as Egyptian hieroglyphics:fortunately, thanks to the work of the CIE that we introduced in Chapter 1, color
management is able to use the perceptually based CIE LAB and CIE XYZ
color spaces as aRosettastone for color, lettingus translate our color from
one set of device-specificRGB or CMYK numbers to another.
In Chapter 2 we broke thesad news that the numerical systems wemost
often use for representing color on our computers-RGB and CMYI-are
fundamentally ambiguous. They aren't descriptions of color. Instead,
they're control signals, or instructions, that make devices like monitors
and printers produce something that we can experience as color. In this
chapter we'll discuss how a color management system (CMS) works to
reconcile the RGB and CMYK control signals with the perceptually based
CIE numbers.

80

Real World Color Manag.mrm, Second Ldltlon

Color management systems have to perform two critical tasks:
b

They have to figure out what perceived colors our RGR and CMYK
numbms represent.

They have to keep those colors consistent as we go from device to

device.

Most discussions of color management focus on the second task, but

it\ important to realize that until you do the first. the second is impossible--you can't match a color until you knowwhat it is! It's also useful to
set realistic expectations-see the sidebar, "The WYSIWYG Myth." later
in this chapter.
The details of implementing color management and using it in our
applications can be insanely complex, since each application vendor
seems to insist on developing its own unique interface and terminology.
but the fundamental workings of color management systems are relatively simple. Color management systems really do only two things:
b

They attach a specific color meaning to our RGB or CMYK numbers,

makingthem unambiguous.~thcolormanagement,we always know
what color a given set of numbers represents.

They change the RGB or CMYK numbers that we send to our various
d e v i c e s a monitor display, an inkjet printer, an offset press-so that
each produces the same colors.

Some CMS implementations may appear to do more complicated

things, but on closer examination, anything a CMS does nlzooys boils
down to a combination of these two tasks. In Part 111 of this book we'll
discuss the many ways you can use a CMS, but in thls chapter we'll look
at what a CMS actually does.
Once you understand what a CMS really does, you'll find that it's a lot
easier to wade through all the various menus and dialog boxes that applications use to control color management. First, though, let's look briefly
a t why we need color management in the first place.

In the old days, life was a lot simpler. We didn't need color management
in what we might call the one-input-one-ourput workflow. AU images
were scanned by a professionaloperator using a single scanner producing
CMYKtuned to a single output device. Spot colors were handled either by
mixing spot inks or by using standard CMYK formulas in swatch books.
Accurate monitor display was unheard of. The system worked because
the CMYKvalues the scanner produced were tunedforthe output device,
forming a closed loop that dealt with one set of numbers.
Fast-fomd to the third millennium. For input, we now have not only
high-end drum scanners, but also high-end flatbed scanners, desktop
flatbeds, desktop slide scanners, and digital cameras. On the output end,
we not only have more diverse web and sheetfed presses with waterless
inks, soy inks, direct-to-plate printing, and HiFi color, but also digital
proofers, flexography, film recorders, silk screeners, color copiers, laser
printers, inkjets,and even monitors as final output devices.This diversity
breaks the old closed-loopworkflow into a zillion pieces.
The result is a huge number of possible conversions from input to output devices (see Figure 3-1). Instead of one input and one output device,
today we have to deal with a very large number (which we'll arbitrarily
label m)of input devices and an equally large number (which we'll arbitrarily label n) of output devices. With an m-inputln-output workflow,
you need n x rn different conversions from input to output. We'll let you
do the math, but if you have more than a handfut of inputs or outputs,
which most of us do, things quickly become unmanageable.
Rgun M

nxminputto
output conversions

m input devices

n output devices

89

Real World Color Management, Second Edltlon

The ingenious solution provided by color management is to introduce

an intermediate representation of the desired colors called the profile
connection space, or PCS. The role of the PCS is to serve as the hub for all
our device-to-devicetransformations. It's like the hub city for an airline.
Rather than have separate flights from 10 western cities (San Francisco,
Seattle, Los Angeles, etc.1 to 10 eastern cities (Boston,Atlanta, Miami, etc.)
for a total of a hundred flights, the airline can route all flights through a
major city in the middle, say Dallas, for a total of only 20 flights.
The beauty of the PCS is that it reduces the m x n link problem to m +n
links. We have mlinks from input devices to the PCS, and nlinks from the
PCS to the various output devices (see Figure 3-21, We need just one link
for each device.
figure 3-9
ncm input to output

conversions

rn input devices

&

n output devrces

Each link effectively describes the color reproduction behavior of a

device. This link is called a device profile. Device profiles and the PCS are
two of the four key components in all color management systems.

Chapter ):Color Management

83

1 The WYSIWYG Myth

While the desire for WYSIWYG monitor andwhatvou get on your
(What You See Is What You Get) printer.Wepoint thisout tocounwas a drivingforce behind the de- ter both the overzealous hype of
velopment of color management. color-management advocates
colormanagement isn'tjustabout who may claim that color manachieving WYSWG, which in its agement gives you true WYSIextreme sense is an impossible WYG, and the detractors who
goal. While color management claim that color management is a
goes a long way toward makinga failure because it hasn't delivered
monitor simulate the limitations on its promise ofWYSMNG.
In fact, we question whether
of a printer, monitors and printany
system has ever achieved
ers have fundamentally different
viewing environment issues. so the extreme interpretation of
there will always be differences WYSIWG. A traditional filmbetween what you see on your based proof certainly produces

a close visual match to h e press

sheet. but we've never seen one
that was identical to the press
sheet, Our belief is that m y
proofing system requires some
interpretation on the part of the
user. Color management won't
make your monitor and your
printer produce identical color,
but it u ~ t l lproduce a close and
consistent visual match that, with
very little learning, you can use as
a pretty accurate predictor of the
final output

The Components of Color Management

All ICC-based color management systems use iuur basic components:
b

PCS. The profile connection space allows us to give a color a n unambiguous numerical value in CIE XYZ or CIE LAB that doesn't depend
on the quirks of the various devices we use to reproduce that color, hut
instead defines the color a s we actually see it.

Pmfiles. Aprofile describes the relationship between adevice's RGR or

CMYEl control signals a n d the actual color that those signals produce.
Specifically, it defines the CIEXYZ or CIE LAB values that correspond
to a given set of RGB o r CMYK numbers.

CMM. The CMM (Color Management Module), often called the

engine, is the piece of software that performs all the calculations
needed to convert the RGB or CMYKvalues.The CMM works with the
color data containedin the profiles.

Rendering intents. The ICC specification includes four different rendering intents, which are simply different ways of dealing with -outof-gamut" colors--colols that are present in the source space that the
output device is physically incapable of reproducing.

84

Real World Color Management, kcond Edition

The PCS
The PCS is the yardstick we use to measure and define color. As we hinted
earlier in this chapter, the ICC specification actually uses two different
spaces. CIE XYZ and CIE LAB, as the PCS for different profile types. But
unless you're planning on writing your own CMS, or your own profilegeneration software (in which case you'll need to learn a great deal more
than the contents of this book) you needn't concern yourself greatlywith
the differencesbetween the two. The key featureof both CIEXYZ and CIE
LAB is that they represent perceiwdcolor.
It's this property that makes it possible for color management systems
to use CIE XYZ and CIE LAB as the "hub" through which all color conversions travel. When a color is defined by XYZ or LAB values, we know how
humans with normal color vision will see it.

Profiles
Profiles are conceptually quite simple, though their anatomy can be
complex. We'll look at different kinds of profiles in much more detail in
(orfor a truly geek-level look at the contents
Chapter 4, A1IAborrtPro~les.
of profiles, see Appendix A, Anatomy of a Profile).For now, though, we'll
concentrate on their function.
A profile can describe a single device, such as an individual scanner,
monitor, or printer; a class of devices, such as Apple Cinema Displays.
Epson Stylus Photo 1280 printers, or SWOP presses; or an abstract color space. such as Adobe RGB (19981 or CIE LAB. But no matter what it
describes, a profile is essentiany a lookup table, with one set of entries
that contains device control signal valuesRGB or CMYK numbers-and
another set that contains the actual colors, expressed in the PCS, that
those control signals produce (see Figure 3-3).
A profile gives RGB or CMYK values meaning. Raw RGB or CMYK values
are ambiguoustheyproduce diffemnt colors when we send them to differentdevices.Apmfile,by itself, doesn't change the RGB or CMYK n u m b e r s
it simply gives them a specific meaning, saying, in effect, that these RGB or
CMYK numbers represent this specific color (definedin XYZ or IAB).
By the same token, a profile doesn't alter a device's behavior-it just
describes that behavior. We'll stress this point in Chapter 5 when we discuss the difference between calibration (which alters the behavior of a
device) and profiling (which only describes the behavior of a device), but
it's a sufficiently important point that it bears repeating.

Chaptar 3: Color Management

85

The ICC-Some Historical Perspective

Apple Computer recognized
In the late 1980s and early 1990s,
many companies-most notably, that this problem of incomAdobe, Agfa, Electronics for Im- patibility of profiles had to be
aging, Hewlett-Packard, Kodak, addressed at the operating system
Linotype-Hell. Pantone, Tektro- level, and in 1993 it introduced
nix, and Xerox--developed color ColorSync, a color-management
management systems that used architecture built into the Macinprofiles to solve the device-to- tosh operating system. Apple also
device color-matching prob- started the ColorSync Consorlem. However, the profiles from tium, which consisted of compaone company's solution weren't nies using the ColorSync profile
usable by others, and consum- format as well as shaping its deers were limited to the profiles velopment.
This consortium of compacreated for a specific color mannies later became known as the
agement system.

International Color Consortium,

or ICC. One of the consortium's
goals was extending this profile
architecture from the Macintosh
to Windows and Unix computers. The central document of the
ICC is the ICC Profile Format
Specification, which describes an
open profile format that all vendors can use. By defining aformat
that allowed consumers to mix
and match profiles created by different vendors, the ICC standardized the concepts of profile-based
color management.

Converting colors always takes two profiles, a source a n d a destination. The source profile tells the CMS what actual colors the document
contains, a n d the destination profile tells the CMS what newset of control
signals is required to reproduce those actual colors o n the destination device.You can think of the source profile as telling the CMS where the color

came from, a n d the destination profile a s where the color is going to.
hpur deuicek space

Input profile

Output deuicek space

Output profile

A profile contains two sets of values, RGB or CMYKdevice control values, and the
corresponding CIEXYZor CIE LAB values that they produce.

86

Real World Color Management, krond WlHon

The CMM
The Color Management Module, or CMM, is the software "engine" that
does the job of converting the RGB or CMYK values using the color data
in the profiles [see the sidebar, "What Does CMM Stand For?"). A profile
can't contain the PCS definition for every possible combination of RGB
or CMYK numbers--if it did, it would be over a gigabpe in size--so the
CMM has to calculate the intermediatevalues. (See thesidebar, "Why Do
We Need CMMs At AN?")
The CMM provides the method that the color management system
uses to convert values from source color spaces to the PCS and from the
PCS to any destination spaces. It uses the profiles to define the colors that
need to be matched in the source, and the RGB or CMYK values needed
to match those colors in the destination, but the CMM is the workhorse
that actuallv performs the conversions.

When do p u care about the CMM? You rarely have to interact with
the CMM-it just lurks in the background and does its thing. But if you
have multiple CMMs--Bruce's Mac has CMMs loaded from Adobe, Agfa,
Apple. Heidelberg, Kodak, and X-Rite, for example--it's often useful to
know which one is being used for any given operation.
Whv should you care? Well, ICC-compliant CMMs are designed to be
interoperable and interchangeable. but they differ in their precision and
their calculations of white point adaptation and interpolation (using the
profile points as guides), and some profiles contain a "secret sauce" tailored for a particular CMM.
Thedifferencesin precision tend to be subtleand often profile-specific.
lfyou havemultiple CMMs installed, and youget strangeresultsfrom one.
it's always worth trying another,

What Does CMM Stand For?

The acronym CMM has been
know to stand for several differ-

ail the names mean the same in a modular color management

thing. Our preference is Color system. But it doesn't really matent things-Color Mnnngement Managentent Module, as this terwhat it5 called aslong as there
Module, Color Mnrciting Merhod, captures the essence of what the is agreement on what it is.
Color Manipzflation ModeCbut engine i s - a drop-in component

Chapts 1:Color Manrgemmt

87

Why Do We Need CMMs At All?

A prnfile for an input device can't

contain the PCS definition for

every possible combination of
RGB values, nor can an output
profile list the CMYK values for
every possible PCS color. For example. as each of the three channels of an RGB device contains
256 discrete values (0-2551,an
RGB space contains about 16.7
million possible RGB values. If
an RGB device profile contained
the PCS definition of each of the
16.7 million RGB values, and if
you dedicated only three bytes
for each of thosePCS definitions,
the single RGB pmfile would occupy more than 48 megabytes!
Profiles of this size would quickly
become prohibitively large when
you started embedding profiles in
images and in page-layout documents that contain color from
multiple sources.

So rather than vtoring a point

for every possible RGR value (in
an inputprofile) orevery possible
PCS color (in an output profile).
the CMM describes a way ot
computing points from existing
p o i n t s t h e process called interpolation. This reduces dramatically the number of points that
need to be stored in the profiles.
The simplest CMM assumes
that wedo this interpolationinCIE
LAB. ln fact, if we use LAB as the
PCS Itself, then the CMM Would
seem to havelitdc to do at all. lust
take the IAB values produced by
the input profiles. find the nearest PCS value Found in the output
pmfile, and look up the output
value. However, in practice this
has not produced stellar results.
LAB is simply not as perceptually
uniform as we need For this type
of computation. So different ven-

dors have different mechanisms

for doing this interpolation.
Some supplement LAB with
some clever math to compensate for Its known inaccuracies.
Others avoid LAB altogether and
convert to some other space
for purposes of interpolation.
Still others create profiles that
support additional, private
conversion spaces in addition
to supporting one of the two
required conversion spaces (XYZ
or LAB).
The continuing battle over
which oneofthese strategiesproduces thebest color management
results is one of the reasons why
so many vendors hawk their own
CMM, claiming that theirs produces results superior to those
produced by competing ones.

Thedifferent math for white point adaptation can be moreobvious. As

we explained in Chapter 2, Computers and Color. white comes in many
different flavors, and our eye adapts automatically to the flavor of white
with which it's currently confronted, judging all other colors in reference t o that white. So we usually convert the white of the source space to
the white af the destination space when we d o color conversions. Some
CMMs have d i i c u l t y doing this with some profiles, translating white to
a "scum dot" with o n e percent of ink in one or more channels, instead of
paper white. Switching to a diierent CMM will oftencure this problem.
The differences in math for interpolation can range from subtle to
dramatic. Many CMMs have fixed notorious problems with LAB, such as
skies that seemed to turn noticeably purple, by implementing clever tricks
in the interpolation mathematics.

88

Rral World Color Management, kcond Edition

The practice of building "secret sauce" into profiles directly conaavenes the goal of an open, interchangeable profile format. Kodak is the
main offender in this case. If you use a profiling tool that builds in customkations for aparticularCMM,you may getslightly betterresultsusing
the preferred CMM, but in our experience, the differences have been so
slight that we question thevalue of the practice.

HOW the CMM i s chosen. Profiles contain a tag that lets them request a
preferred CMM when it's available, though they must be able to use any
ICC-compliant CMM if the preferred one is unavailable. This becomes
an issue on Mac OS if you set the choice of CMM to Automatic in the
ColorSync control panel. Doing so allows each profile to select its preferred CMM. It also means that unless you do a lot of detective work, you'll
have no idea which CMM is in use at any given moment.
The Macintosh and Windows operating systems, and almost all colormanaged applications, let you override the profiles' preferred CMM and
choose a specific CMM for all color-management tasks. We recommend
that you choose one CMM and stick to it, experimenting with others
only if you run into a specific problem or if you're trying to get a specific
advantage touted by a CMM vendor.

Rendering Intents
There's one more piece to the color-management puzzle. As we explained
in Chapter 2, each device has a fixed range of color that it can reproduce,
dictated by the laws of physics. Your monitor can't reproduce a more saturated red than the red produced by the monitor's red phosphor. Your
printer can't reproduce a cyan moresaturated than the printer's cyan ink.
The range of color a device can reproduce is called the colorgamut.
Colors present in the source space that aren't reproducible in the
destination space are called out-of-gamut colors. Since we can't reproduce those colors in the destination space, we have to replace them with
some other colors, or since, as our kiends from NewEngland are wont to
remark, "you can't get theah 60m heah," you have to go somewhere else.
Rendering intents let you specify that somewhere else.
The ICC profile specification includes four different methods for
handling out-of-gamut colors, caned rendering intents (see Figure 3-4).
Percephraland sottlmtionrenderingsuse gamut compression, desaturating all the colors in the source space so that they fit into the destination

Chapto

+ Color bnagemeni

89

gamut. Relariveand n6solutecolorime~icrenderingsuse gamut clipping,

where all out-of-gamut colors simply get clipped to the closest reproducible hue.
Perceptual tries to preserve the overall color appearance by changing
all the colors in the source space so that they fit inside the destination space while preserving the overall color relationships, because
our eyes are much more sensitive to the relationships between colors
than they are to absolute color values. It's a good choice for images
that contain significant out-of-gamut colors.
Saturation just tries to produce vivid colors, without concerning itself
with accuracy, by convertingsaturated colorsin the source to saturated colors in the destination. It's good for pie charts and other business
graphics, or forelevation maps where saturation differences in greens,
browns, or blues show different altitudes or depths, but it's typically
less useful when the goal is accurate color reproduction.
Relative colorimetric takes account of the fact that our eyes always
adapt to the white of the medium we're viewing. It maps white in the
source to white in the destination, so that whiteon output is the white
of the paper rather than the white of the source space. It then reproduces all the in-gamut colors exactly,and clips out-of-gamutcolors to
the closest reproducible hue. It's ohen a better choice fot images than
perceptual since it preserves more ofthe original colors.
Absolnte colorimetric differs from relative colorimetric in that it
doesn't map source white to destination white. Absolute colorimetric rendering from a source with a bluish white to a destination with
yellowish-white paper puts cyan ink in the white areas to simulate
the white of the otiginal. Absolute colorimetric is designed mainly
for proofing, where the goal is to simulate the output of one printer
(includingits white point1 on a second device.
When you use a CMS to convert data from one color space to another.
you need to supply the source profile and the destination profile so ir
h o w s where the color comes from and where the color is going. In most
cases, you also specify arendering intent, which is how you want the color to get there. When you aren't given a choice, the application chooses
the profile's default rendering intent, which is set by the profile-building
application and is usually the perceptual rendering intent.

90

Real World Color Manag.IMot Second Wition

Figure 3-4

Rendering intents

Thisfigure shows theMacBeth Color Checker target rendered to the CMYKprofilewe used to make
this book, using each of the four rendering intents. Each rendering is accompanied by a plot in the
a,b plane of LAB space showing the original colors in green, and their rendered equivalents in red.
(We didn't plot theflaays,
- - because they tend to ship only in luminance, which is the thirdaxis ofLAB
and is not ihown on a,b plots.)

original
rendered

Perceptual rendering
Perceptual rendering tries to compress
thesourcegamut into the destination
gamut in such a way that overall color
relationships are preserved.

1
8

Saturation rendering
Saturation rendering mapsfully saturated
colors in the source tofully saturated colors in
the destination without concernfor accuracy.

original
rendered

Chapter 3: Color Managemml

Figure 3-4

9l

Rendering intents, continued

1 original
1

rendered

Absolute colorimetric rendering

Absolute colorimem'c rendering attempts to
reproduce all source colors, including white,
as exactly as possible.
out-ofgamut colors

These colors are largely within gamut but

get different treatmentfron each rendering
intent.
These colors are desaturated by perceptual
rendering, but not by the colorimetric
rendering intents.

original
rendered

I
Relative colorimetric rendering
Relatiue colorimerric renderingscales the source
white to the destination white and
adjusts all other colors proportionally.

99

R u l World Color Management, Iscond Wltlon

Figure 3-4 Rendering intents, continued

In most cases, the differences between perceptual, relative colorimenic,and saturation rendering are
quite subtle. Absolute colorimenic rendering produces uery different resultsfrom the other three since it
doesn'tperform white point scaling, and hence is usually used only for proofing. Notice the difference in
the saturated reds between the perceptual and relative colorimetric neatments of the image below, and
the hue shift of the reds in the saturation rendering.

Perceptual rendering

Saturation re.

~g

Relative colorimetricrendering

Chapta 3: Color Management

93

Color Management in Action

Let's look at how the differenr color-management components interact
We've told you that a CMS does only two things:
b

Assign a specific color meaning to RGB or CMYK numbers.

Change the RGB or CMYK numbers as our color goes from device to
device, so that the actual color remains consistent.

We accomplish the first by assigning or embedding a profile in our

document.We accomplish the second by asking the CMS to convert from
the assigned or embedded source profile to a chosen output profile.

Assigning and Embedding Profiles

hlosr color-managed applications let you assign a profile to images and
other colored objects. For example, Photoshop allows you to assign a
profile to an image. When you do so, you're defining the meaning of the
RGB or CMYK values in the image by assigning the image a profile that
describes where it camefrom, such as a scanner or digital camera. Apagelayout document may have multiple images or illustrations in a single
documentand will allow you to assign a profile to each one. For example.
you may have some scanned images and some digital camera captures.
You'd want to assign the scanner profile to the scans, and the digital
camera protile to the camera captures, so that theCMS knows what actual
colors the scanner RGB and digital camera RGB numbers represent.
Most color-management-enabled applications also let you embed
profiles inside documents such as images or page-layout files when you
save them. Doing so lets you transfer documents between applications or
computer systems without losing themeaning of the RGB or CMYKvalues
used in those documents (see Figure 3-51.
Note that assigning or embedding a profile in a document doesn't
change the RGB or CMYK numbers; it simply applies a specific interpretation to them. For example, if we embed the profile for Bruce's hnacon
scanner in an RGB document, what we're doing is telling the CMS that
the RGB numbers in the document represent the color that the Imaoon
scanner saw when it recorded these RCB numbers.

94

Real World Color Managcnrmt, Second Edition

rigurn 3.5
Prome embedding

An embedded profile doesn't change the

values in the image; it simply tells the
CMS what colors those values represent.

PmjiIe embedded
in an image

Some people 6nd it counterintuitive that if we assign a different

profile-for example, the Adobe RGB (1998) working space-the RGB
numbers in the image don't change, but the image's appearance does. It
does so because we've changed the meaning of the RGB numbers-the
actual colors those numbers represent.
Assigning or embedding a profile is a necessary first step before you
can convert the color for output on another device. This may be done
automatically by scanner or camera software, it may be done explicitly
by the user, or it may be done implicitly by a color-managed applicationmost color-managedapplications let you specify default RGB and CMYK
profiles, which they automaticallyassign to color elements that don't already have aprofile embedded.This assigned or embedded profile is used
as the source profile when you ask the CMS to perform a conversion

Converting with Profiles

To convert animagefrom one profile's space to another's-actually changingthe RGB or CMYKnumbers-we have to specify two profiles, a source
profile and a target or destination profile. The source profile tells the CMS
where the numbers in the document come from so that it can figure out
what actual colors they represent. The destination profile tells the CMS
where the document is goingtoso that it can figure out the newset of RGB
or CMYK numbers needed to represent those same actual colors on the
destination device (see Figure 3-6).

When you mnuert usingprofiles,you must always supply the CMS with a
source (input) profile and a destination (output) profile.

For instance,if Chris shoots an image with adigitalcamera, he ends up

with an RGB image+but not just any RGB image, rather a "Chris's Nikon
D l RGB" image. So he needs a profile for the digital camera in order to
tell the CMS how Chris's NikonD1sees color. That is the source profile--it
describes where the color numbers came from and what perceived colors
they represent. Chris wants to convert theimage to CMYK so it can go into
a magazine article-but not just any CMYK image, rather a "Color Geek
Monthly CMYK" image. So he needs a profile for Color Geek Month+
press to tell the CMS how the press reproduces colors. Now the CMS has
the necessaryinformationto figure out which perceived colors the original image RGB numbers represent and which CMYKvalues it needs to
send to the press to reproduce those colors.
It may seem counterintuitive, but converting colors from one profile's
space to another's doesn't change the color appearance-preservingthe
color appearance is the whole point of making the conversion.
The usefulness of converting colors becomes clearerwhen youconsiderwhy youuse conventionalcolor-correction techniques.Whenyou print
out an image on an inkjet printer and it comes out too green overall, with
an excessively yellow gray balance in quarter-tones and too much cyan
and yellowin shadows, what do you do?Youcolor-correct it using curves,
huelsaturation, replace color, and any other tools you feel comfortable
with, until it prints out the way you want.You're changingthe numbers in

96

Real World Color Management, kc&

Wlllon

the image to produce the color appearance you want. Conversions made
with profiles find the proper device values for you, automatically.
Color management doesn't take a bad image and make it lookgood on
output. Instead, it makes theoutput faithfully reflectall theshortcomings
of the original. So color management doesn't eliminate the need for color
correction. Instead, it simply ensures that once you've corrected the image so that it looks good, your corrections are translated faithfully to the
output device.

How Conversions Work

First, the conversion requires you to select four ingredients:

The sour# profile. This may be already embedded in the documenr, it

may be applied by the user, or it may be supplied by a default setting in
the operating system or application.

The destination prolk.This may be selected by a default setting in the

operatingsystem or application, or selected by the userat the time ofconversion (forexample. choosing a printer profile at the time you print).

The CMM. This may be chosen automatically as the destination profile's

"preferred" CMM, or selected by the user either at the time of conversion.
or as a default setting in the operating system or application.

A mderlng intent. This may be selected by the user at the time of conversion,or as a default setting in the operating system or application. Failing that, the "default" rendering intent in the destination profile is used.
Thecolor management system then performs the series of stepsshown
in Figure 3-7.
That's really all that a CMS ever does. Some CMS implementations let
you do multiple conversions as a single operation-Photoshop, for example, lets you print an inkjet proof that simulatesapress from an image
that's in an RGB editing space. But when you breakit down, all that's happening is conversions from one profile's space to anotherk in sequence.
with different rendering intents at each step.

Chapter % Color Manag.ment

97

Flgun 3-7 A color space conversion

1. The CMS looksat thesource
profihand builds a table
that correlates source RGB
lor CMYX-), with PCS values,
wing relative colorimetric
rendering.
2. The CMS looks at thedestination

profile and builds a table that

correlates PCS values with
destination CMYK(orRGB)
values, wing the selected
rendering intent, or, if none is
selected, the profile's default
rendering intent.

PCS

3. Using the interpolation algorithm

defined in the CMM, the CMS

connects the two tables together
through the common PCS values
and buildsa table thatgoes directly
from source to destination.

4. The CMS uasses each uixel

in the source image, or

color in the source artwork, through the table,
convertimthe valuesfrom
source to destination.

/
'

Conversions and Data Loss

It may seem from the foregoing that you can simply convert your color
from device space to device space as needed. Unfortunately, that's not
the case: if you're dealing with typical 0-hit per channel files, where each
channel contains 256 possible levels, each conversion you put the file
through will lose some of these levels due to rounding error.
Rounding errors happen when you convert integers between different
scales-for example, 55 degrees Fahrenheit and 56 degrees Fahrenheit
both convert to a brisk 13 degrees Centigrade as a product of rounding

98

Real World Color Managmnt, Second Witlon

to the nearest integet value. The same kinds of errors occur when you
convert 8-bit RGB or CMYK channels to PCS values.
To put this in context, we should hasten to point out that almost any
kind of useful editing you perform on files made up of A-bit channels will
result in some levels getting lost. So it's not something to obsess about.
But it is something to bear in mind.

The High-Bit Advantage

ll you start with 16 bits of data per channel, your color suffers much less
from rounding error, because instead of having onlv 256 possible input
values per channel, you have at least 32.769, and possibly as many as
65.536! (Photoshop's implementation of 16-bit color uses values from 0
to 32.768. To those who would say that this is 15-bit color. we point out
that it takes 16 bits to represent, and it hasan unambiguous center value.
16,384. which is useful in many operations.)
In either implementation, you'll still get roundingerrors, but you have
so much more data to start with Lhat they become insignificant. There's
no such thing as a free lunch, but high-bit conversions come pretty close.
However, just converting an existing 8-bit per channel file to 16 bits won't
buy you much-you have to start out with real high-bit data to get real
benefits.
Rounding error is simply something you need to keep in mind when
constructing color-managed worktlow--you generally want to put your
color through as fewconversions as possible.

Color Management I s Simple

\Ve'\ e dehberately kept thmgs conceptual in this chapter. and we've left
out a lot of derails. But until you understand the basic things that a CMS
does, the details won't make any sense. Keep in mind the fundamental
concepts behind color management-an assigned or embedded profile
describes the actual colors represented by RGB or CMYKnumbers, and a
conversion from one profile to another keeps those actual colors consistent
by changing the numbers that get sent to the output-and the details,
which we'll continue to feed you throughout this book, will make a great
deal more sense.

All About
Profiles
Describing Devices

If color management is as simple as we say it is in Chapter 3, Color

Management, you may be wondering why you even need this book. The
answer is that, while color management is simple, color reproduction is
not. The various machines we use to reproduce color are prey to a host
of factors that affect their behavior, and color management is blissfully
unaware of them-unless they are captured in a profile.
Color management systems know nothing about the various devices
they drive except for the information that's recorded in the profiles for
those devices. All the intelligence in color management systems-rendering intents, previewing capabilities, simulations-is built into the device
profiles. So it's essential that you obtain or create accurate profiles for your
various devices. We tell you how to create device profiles later in the book,
but to make good ones, you need to know a little about what comprises
them and how they work.

What Are Device Profiles?

Contrary to prepress legend, profiles aren't magic, nor are they made by
elves during the full moon in Orthutanga.

106

Real World Cdor Management, kcond Edition

A profile is just a file that correlates device color values with corre-

spondingdevice-independent color values that represent the actual color

people see. The device values are expressed as the control signals-usually RCB or --that
we send to our devices to make them reproduce color.The corresponding device-independent values that represent
the color they reproduce are expressed in the profile connection space
(PCS)--either CIEXYZ or CIELAB. (Refer to Chapter 3 for more on the
PCS.)
The device profile contains information about the three main variables
that describe how a device is behaving
b

Gamut-the color and hrighmess of the colorants (primaries)

Dynamicrange--thecolorandbrightnessofthewhite point and black

point

Tone-reproduction characteristics of the cnlonnts.

[We tell you more about these variables in Chapter 2, Computers mad
Color, in the section. "Why the NumhersVary.")
Some kinds of profiles may contain additional information, such
as instructions for handling out-of-gamut colors, more detailed tonereproduction information, or special "secret sauce" information that's
only used with a particular CMM.

Profile Classes
Profiles come in a few varieties, or classes: inputprofilesdescribe scanners
and digital cameras: display profilesdescribemonitors andLCD displays;
and orifpltr profilesdescribe printers and presses.
Some people confuse the term inpiitprofilewith source pmfile, and outpul profilewith destination profile. Input and output profiles refer to distinct
Lypes of devices that these profiles represent, whereas source and destinalion profiles refer to temporaryrolestwo profiles take at the moment a color
conversion happens.We describe source and destination profiles in "Using
Profiles-Source and Destination,"later in this chapter.
One factor that differentiatesthe three types of profilesiswhetherthey
are one-way or two-way-that is. whether they allow the CMS to convert
from device space to PCS and horn PCS to device space. Input profiles
only have to define the conversion from the input device's color space to
the PCS.You can't view or output color on a scanner or digital camera, so

C
h
a
w 5: All About P d i h

10l

there's really no need to convert color to scanner or digital camera space.

Input profiles simply tell the CMS how people would see the color the
scanner or camera captures (see Figure 4- 1).

One-way and
two-way prnfles

Communication between the CMSanddisplay profiles is two-waxso

the CMS can send the rightRGB values to the monitor to reproduce
a requested color;and can interpret monitorRGB ualues to reproduce
them on outputdevices.

Profiles let the CMS

translate between
different devices'RGB
and CMYK v a l w
using the PCS as a
common ianguage.

Communication between
the CMS and input
profiles is one-way,
because the CMS simply
needs to know what
colors the inputRGB
numbers represent.

Communication between the CMS and output devices must also be

two-way so the CMS can translate output CMYKor RGB to monitor
RGB forsofr-proofing, and can translate between different output
RGB or CMYKso that it can, for example, reproduce press CMYKon
an RGB inkjet.

On the other hand, display profiles must be two-way because your

display acts as both an input and an output device. If you create or edit
a color based on its appearance on your monitor, you're using the monitor as an input device: the CMS needs to know what color you're seeing
on the monitor so that it can reproduce it on a printer, press, or another

109

Real World (dor Y b n w n t , Second EdlUon

display. The CMS looks at the monitor RGB values being displayed and
uses the profile to calculate what actual color those RGB values represent.
Conversely,when you display an image on your screen, the monitor is the
output device: the CMS evaluates the image's embedded profile to determine what actual colors are represented by the numbers in the file, and
then uses the monitor profile to calculate the monitor RGB values needed
to display those colors accurately.
Output profiles are always two-way profiles, too. We use them not only
to convert from the PCS to the output color space for printing, but also to
display files already converted to output space on the monitor, or to convert a press CMYKimageto some other output device's space for proofing.
When you displaya CMYKimage on your RGB monitor, the CMS relies on
the CMYK output profile to tell it how to convert the numbers back to the
PCS and, ultimately, to monitor RGB.
The device-to-PCS transform is known as the backwards transform,
and the tables that specify it are known as the AtoB tables, while the PCSto-device transform is known as theforward transforna,and the tables that
specify it are known as the BtoA tables (see Pigure 4-2).

U ~ 4M
4

M a e and
table-based pmliles
Matrix profiles contain
tags that M b e the
CIEXYZ values of the
primaries, whichform
the matrix,plus tags
that ddbe the
tone reproduction
characteristics of
each colorant.

(AtoB),and benueen LAB

Chapter 5: All A b o a Proflles

103

Matrix-Based Vs. Table-Based ICC Profiles

ProEles can be either matrixbased or table-based. Both types
include the dev~ce'swhite point,
but they differ in the way they
represent the device's tonereproduction attributes. This is
why w m e profiles are tiny while
others can weigh in at over a
megabyte, and why some profiles
seem to produce identical results
no matter which rendering intent
)wu choose while others produce
very different results with each
rendering intent.
The simplesi way to store
a conversion from one three
channel space to another-such
as from KGB to XYZ-ls to use a
mathenlatical model known as a
3 X 3 nlorrix. A 3 X 3 matrix is an
array of nine numbers that can
convert any triplet of numbers.
such as (20. 30. 401, to another
triplet, such as (1.83. 2.0. 3.67).
In ICC profiles, the 3 X 3 matrix
consists of the XYZ values for
each of the three colorants
(primaries) of the device. The
profile also contains one or more
numbers defining the tone curve
of each colorant. The device
numbers are passed through the
tone curves before conversion
using the 3 x 3 matrix. Matrix
ptufiiesalways use CIEXnas the
profile connection space.
The other way to store the
conversion is in a lookup idle, or
LUT, which is a table of numbers
that lets you look up any input

value and fmd its corresponding

output value. The profiling package decides the number of points
in the LUT, and vendors often
compete on the trade-offbehveen
the quality afforded by more
sample points and the Increased
size and decreased speed of
the profile. LUT-based profiles
always use CIE LAB as the profile
connection space.
Matrices are much smaller
than lookup tables because you
always store only nine numbers
for the matrix whereas the size of
a LUT depends on the number of
sample points it contains, which
can number into the hundreds
or thousands. Matrices are reversible-the same matrix lets
you convert numbers in either
direction. But matrices are only
good for devices with fairly
simple tone curves, like scanners and basic CRT-based rnonitors. For more complex devices.
like printers, matrices just don't
contain enough information.
Matrix-based profiles only
contain information for a single rendering intent, which is
assumed to be relative colorimetric. even if the user interface
of the application calling the
convenion lets you choose
perceptual or saturation.
Matrix profiles do support
absolute colorimetric rendering
in addition to relative colorimetric; absolute colorimetric doesn't

require additional tables because

it's computed from the relative
colorimetric intent using the
white point value.
Lookup tables can represent
extremely complex devices just
by adding more points in b e
table. They also have an advantage in that they aren't limited
to three-channel devices, so
they can describe four-channel
devices such a s CMYK printers. or even printers with higher
numbers of inks. But lookup
tables get very large when you
add more data points, and since
the tables are unidirectional, you
need a separate table for each
direction.
Input and display profiles can
be matrix-based or table-based,
which is why you niay find that
profiles made by one vendor are
much larger than those from
another. Display profiles are
always bidirectional. Some input
profiles are also bidirectional,
though it seems a little pointless
10 convert color to a scanner or
digital camera space.
Output profiles are the largest profiles by far: they musL be
rable-based and must store two
tables (one for each conversion
direction) for each of the three
rendering intents, perceptual, relativecolorimetric, and saturation.
So a typical output profile has six
lookup tables!

Real World Color Management, kcond Edition

Stimulus and Response

One easy way tovisualize the process of making profiles is to thinkin terms
of stimulus and response. Empirical science often boils down to poking
things and seeing what happens. Profilers work by sending a stimulus to
the device, obtaining a response, and then comparing and correlating
the two. In other words, they poke the device with known RGB or CMYK
values, and then measure what the device produces.
For a display device, the profile maker sends a stimulus in the form of
known RGB values and then measures the colors those RGB values cause
the monitor to display. The measurements are made with an instrument
that captures PCS (XYZ or LAB) values. Then it constructs a profile that
correlates the RGB values and their PCS equivalents. Hence the profile can
tell the CMS what actual color will result from a given set of RGB values.
The profile can also tell the CMM what RGB values are needed to display
a specific color (see Figure 4-3).
Monitor profiling and
calibration packages send
known RGB ualues to the
monitor and measure
the color that results with
an instrument like theones
shown here. They use the
first Set of measurements
to calibrate the monitor;
adjusting its behavior;and
a second set to create the
monitor profile.

Chapter 5: A11 About ProRks

105

For an output device such as a printer, the profile maker sends known
RGB or CMYKvaluesto the printer and then measures the printed output
Again, it builds a profile that correlates the stimulus and the response, so
that the CMS can tell from the profile what actual color will result from
specific RGB or CMYKvalues,and what RGB or CMYK values are needed
to print a specific color (see Figure 4-41,
Figure 4-4

The IT8.713 target

The 1111.713 target is a standard target for projiling CMYKourputdevices. It contains parches with 096,
10% 20% 4053, 70% and 100%of each ink in all possible combinations, plus 13-step ramps of each ink
In addition, rows Kand L contain near-neutral patches usingfiner increments. hfllingpackages
use this target to model all the colors CMYKdeuicescan produce.

106

Redl World Cdor Management, W n d Wltlon

For an input device such as a scanner, the process differs only in that
the measurements are usually already done for you. The stimulus in this
case is a scanning target that's either supplied with the profiling package or obtained from a third-party vendor. Scanner targets are always
accompanied by a data file that records the LAB or XYZ values of the color
patches. You scan the target, and then feed the scanner profiler the scan
and the target data tile. The profiler compares the RGB values in the scan
and the LAB or XYZ values in the target data file and builds a profile that
tells the CMS how the scanner sees color (see Figure 4-5).
Figure 4-5

The HCT scanner target

and referencedata file

Chaprw 4: An About Profilea

107

rhere arc t h ~ e elimitations in this process of creating profiles that you

need to keep in mind.
First, the profile maker can't cover the entire set of possible device signals. For example, a printer profiler can't send every possible combination
of C M X values to the printer to see what results-at one percent increments, doing so w o ~ ~require
ld
a target with 100,000.000 patches! Even if
you had the equipment (and the patience) to measure all these patches.
storingthat manv sample points in a lookup table would produce output
profiles in the gigabytes.
So interpolation is unavoidable, both at the time the profle is created
and when it is used by the color management system. We mention this
becauseithelpsexplain what theCMM does, and why some color conversions areslower than others. Many people assume that with a big enough
lookup table there would be no need for a CMM because there would
be no interpolation, or at the least all CMMs should produce the same
results with the same profiles. While in theory this is true, in practice, interpolation is unavoidable,which is why some vendors try to find better
interpolation systemswith different CMMs.
The secondlimitation is that profilescan't make devicesdothings they
cannot do. If a printer can't produce acertain shade of red, a profile won't
somehow make it magically produce that color.'I'he profile describes the
gamut of the device, but it can't enlarge the gamut.
The third limitation is that the profile is only as accurate as the measurements on which it's based. A profile is a snapshot of the way the
device was behaving when the measurements were captured. The majority of color devices drift over time, so youneed to take steps to ensure that
the device is behaving the way you want it to before you profile, and you
need to takesteps to keep itbehaving thatway afteryou profile; otherwise
:he profile will no longer provide an accurate description of the device's
behavior, and you won't get the color you want. We discuss this topic in
more detail in Chapter 5, Mensurernent, Calibration, andProcess Control
and in the following three chapters that deal with the specifics of building
display, input, and outpul profiles, respectively.

108

Real World Color Management, k o d Wltlon

Using Profiles-Source and Destination

By itself, a profile does nothing at dl-even when a profile is embedded

in a file. It's only when you're converting colors from one device to another that a profile becomes active. At that time, the color management
system doing the conversion needs to know where the colors came from
and where they are going-and thus requires hcwprofiles to do the color
conversion.
In a color conversion, the profile you're convertingfrom is the source
profile, and the profile you're converting to is the destination profile.
Source and destination profiles arenvttypes or classes of profile--they're
just temporary rolesthat two profiles play when a color management system usesthem to convertcoloisfiomone device to another. Input profiles
are almost invariably used only as source profiles-it makes much more
sense to convert colorfrom a scanner or camera space than to convert it
toa scanner or camera space--but output and displayprofiles are equally
at home as source and destination profiles. For example, if you have an
image from an RGB scanner that you want to print on a press, you'd convert it using the scanner RGB profile as the source and the press CMYK
profile as the destination. But if you then wanted to proof the converted
CMYKimage on an RGB inkjet printer, you'd perform another conversion
using the CMYK press profile as the source and the RGB inkjet profile as
the destination.

Generic Profiles
l u s ~about every color device sold nowadays comes with one or more
generic KC profiles made by the device manufacturer and describing the
typical behavior of the device model rather than the specific behavior of
the individual unit. Some of the more rabid color-management enthusiasts will tell you that all such profiles are useless. While it's probably true
that the inappropriate orover-optimisticuseof generic profiles-profiles
that were made from something other than the measuted behavior of
your particular unit-have caused far more than their fair share of colormanagement headaches, some generic profiles are agreat deal more useful than others.

Chaptn 5: bll About Profilu

109

Generic profiles are useful when they're for very stable devices that
display little or no variation from unit to unit in manufacturing, or little
or no variation from batch to batch in consumables. The profiles that accompany many inkjet printers, for example, fall into this class: you map
get slightly better results from a custom profile, but the generic ones are
nevertheless quite useful. Generic profiles are also useful in representing standard conditions, such as SWOP (Specifications for Web Offset
Publications), that define how a press should behave, or far common
mainstream proofing systems.
~
say worse than useless; Chris
Generic profiles are useless ( B N c would
would say somewhere between lethal and catastrophic) for unstable
devices. or for ones [hat display significant unit-to-unit variation, such
as CRT displays. CRTs not only have a lot of unit-to-unit variation fresh
from the factory, they drift over time and they offer controls that let you
change their behavior radically, making generic monitor profiles a cruel
joke at best.

Color Space
Profiles
-

Thus far in this chapter,we've conlinedourdiscussion of profilesto device

profiles-profiles that describe the behavior of an actual physical device
or a class of physical devices. But there are also prafiles that describe
device-independent color spaces such as CIE LAB or CIEXYi-these are
known, logically enough, as color space profiles.
In recent years, another type of profile has become commonplacedevice-independent RGB profiles. Technically. these aren't color space
profiles as defined in the ICC spec-they're built l i e display profilesand
appear to the CMS as display profiles--but in practice they behave a lot
more like color space profiles than device profiles. They don't represent
devices, so they aren't subject to variability, and they're always accurate.
We'll discuss these profiles more fully in Chapter 10. Color Managernent
Workflolu.

110

Real World Color Management, Second Edltlon

Good Profiles
One entirely reasonable way to decide whether your profiles are accurate
enough is simply to look at the results you're getting and decide whether
or not you're happy with them. But if you bought this book, chances are
that you're less than satisfied with what color management is currently
doing fot you. In the next four chapters, we look at building profiles, and
reveal our techniques for making sure that your profiles give you the best
results from your devices and that your devices keep behaving the way
your profiles think they do.

Building and
rning Profiles

This page lefi blank. ..

Ileasurement,
Calibration;
ocess Control
"The Map Is Not the Territory"

Whether you attribute the above quotation to Britishartist RalphRumney

and the Situationist International movement, or to Alfred Knrzybski, the
father of general semantics, you probably didn't picture either of them
slavingover a hot spectrophotometer or wrestling with rendering intents.
As far as weknow, bothwere blissfully unaware of color management. But
the quotation contains an important lesson forwould-be color managers:
just as the map is not the territory, in color management, the profile is
not the device. If your map doesn't reflect the territory accurately, you'll
get lost, and ifyour profiles don't provide an accurate description of your
device's behavior, your color output will be unpredictable.
In Chapter 4, we explained that color management lives or dies by
the accuracy of the profiles used. Profiles are essent~allysnapshots of a
device's behavior,so if the profiled device's behavior changes, the profiles
are no longer accurate and you won't get the color you expected. In this
chapter, we'll talk about keeping our devices' behavior in sync with the
profiles that describe them. Itb the part of color management that most
often gets overlooked because it's something you need to do outside of
the color management system.

114

Real World Color Hanagetnml, Second Edltlon

Keeping Track of Device Drift

Prciczss rorinnl

is (he art and science of tracking and compensating for

variability In the realm of color management, it means keeping track of
the way our devices are behavlng and either calibmtingthem (adjusting
their behavior to keep it in sync with our profiles),or characterizing them
[building new profiles that reflect their current behavior).
b

Calibration is the act of changing a device's behavior to achieve some

desired state.We calibrateformanyreasons, but for color management,
the most important reason isto make the device behave consistentlyso
that the profile that describes it remains accurate.

ProJling, which is often called characterization, is the process by

which we record a device's behavior into a profile. It doesn't change
the device's behavior; it just records how this device makes color and
what colors it can (and cannot) reproduce.

Calibration is like performing a tune-up on your car. You're actually

changing the behavior of the device (in this case, your car's engine) by
adjusting something. Profiling. on the other hand, is like getting a printout o i the mechanic's report-the car's current levels of hydrocarbons
and carbon monoxide emissions, RPM measurements of idle speed, and
so forth-that describe the car's current behavior. lust as you can get an
emissions report with or without a tune-up first, with color management
you can profile (characterize)a device with or without calibrating it first.
Calibration actually changes the way a device behaves, whereas profiling
simply describes how the device was behaving at the time the profile was
created.
If a device isn't stable--if its behavior changes over time--then simply
profiling it, as Bruce likes to point out, is like trying to measure a moving
target with a rubber ruler.Thepoint is not that it's impossible; it's just much
harderthan measuringa stationarytargetwith arigid ruler. Ideally, process
control is accomplished through calibrationalone.It's easier to calibrate an
unstable device to make sure that it's behaving the way the profile says it
is. In some cases, though, we simply have to accept a certain amount ofinstability. In those cases, process control will require either more hequent
profiling or basing the profile on a device's average behavior.

Chapler 5: Mcarurummt, Callbratlon, and Pro-

Control

115

Stimulus and Response

Calibration and profiling are otten confused because the steps involved
seem very sbnilar on the surface. Both processes involve sending a target of known color signals-the stirnulu.+to a device, measuring the
color that results--the resporzseand then feeding the results into some
software. With some kinds of devices, such as monitors, calibration and
profiling are performed at the same time by thr same pieces of software
or hardware.
But calibration and profiling are separate processes with different
goals. When you profile, you simply record the response and encode it
in a profile. When you calibrate, you use the stimulus and response to
actually change the device's response. For example. when you set a
monitor's white point, the calibration software sends a stimulus-RGB
255, 255,255-to the monitor, examines the response that's derected by
the measuring instrument, and keeps adjusting either the monitor guns
or the lookup table in thevideo card until the response matches the white
point that you requested.
Some devices simplycan'tbe calibrated because they don't have anything
you can adjust. Desktop scanners. for example, generally have no useraccessible calibration features. The same holds true for non-PostScript
inkjet printers. In these cases, you just have to keep track of the device's
behavior and reprofile if it drifts too far (which is relatively rare).
In all cases. though, vou need to control all the variables that can make
thedevice produce a different response to stimuli than the one the profile
predicts.We'l1 look indetail at the kinds of things that introduce variability
for different device types in the followingchapters that deal with building
profiles, but here's a quick rundown.

Controlling Variables
Some sources of variability are inherent in the devices themselves, and
we have to account for those either with calibration or by reprofiling, but
others are controllable with nothing more than a little common sense.
Here are a few sources of variability that may or may not be obvious but
that nonetheless need to be kept under control.

116

Real World Color Management, kcond CdHlon

Software dtlngs. Input, display, and output devices are all subject to
having their behavior controlled by software. For example, scanners often
have autoexposure routines, and some printers have automatic colorcorrection routines that do different things depending on image content.
These features defeat color management before it starts because they
make the device respond inconsistently-they'll typically tr).to increase
the contrast on a flat image, for example. The color management breaks
because it expects the device to always produce the same response to a
given stimulus, so you need to turn these features off. Figure 5-1 shows
the kinds of things that happen when you fail to do so.
Slightly less obviously, different resolution settings may change the
color rendition-more so on display and output devices than on scanners and digital cameras. So you need to keep your resolution settings
consistent or build profiles for each setting.
Often, something as seemingly innocuous as a software update may
change the device's behavior. So be vigilant, ensure that vou keep all your
software settings consistent, and update them only when necessary.
Media and consumables. Probablythe biggest variables that affect output devices are the inks, toners, dyes. or waxes that form the color, and
the paper or other substrate on which they're laid down.
Thesame inks can produce radically different colors when you change
paper stocks, for example. So, if you switch to a different brand of ink [or
toner. etc.) or paper, you'll almost certainly need to reprofile the device.
We could write a 2,000-word section on this topic alone. but it wouldn't
change the reality: even if you buy paper or inks consistently horn the
same supplier, watch out for manufacturingvariatians, or unannounced
manufacturing changes. Ifyourdevicebehaviorsuddenlyseems tochange
immediately afteryou'veloadednewconsumables, thenewconsumables
are probably the culprit.
A related issue is that many inkjet printers use media settings to control ink limits and black generation. We always make a point of doublechecking these settings before pressing theprint button, because if they're
wrong they have a dramatic impact on the appearance of the print, and
more often than we'd like, media settings in the printer driver that appear
to be sticky, aren't. Figure 5-2 shows how the wrong media settings can
ruin a print.

Chapter 5: Musunmcnt, Callbratlon, and Proccrr Control

Flgure 5-1

The effectof autoexposure routines

The image at right shows the resu1t.sof

applyinga pmfile, created with autoexposure
turned off while scanning the profiling target,
to a scan also made with autoexposure turned
ofi Itproduces an image with clean neutrals
andgood saturation.

Theimugeatleftmu1t.s
from neglecting to
turn offautoexposure
while scanning the
profiling target,
producinga rather
useless profile that
imposes poor gray
balance and noticeably
reduced saturation.

.,

$ling

and scanning

Autoexposure on while profiling, off

while scanning
The image at right is the result offailing to
defeat autoexposure while scanning both
the profiling target and the image.
In this case it produced a heavy bluepurple color cast. On another image, the
results would be different,but almost
certainly just ar bad because
autoexposure randomizes the
scanner's response.
Autoexposureon while profiling
and scanning

117

118

Real World Cdor Managcnrmt, a n d Wltlon

Figure CO

The importance of correct media settings

Matte media
settings on matte
paper

Glossy media
settings on matte
Paper

Matte media
settings on glossy
paper

Chapter R Murummnt, Calibrath, and P

m Control

119

Hardran settings. Ifthe device in question has knobs, sliders, orswitches, they need to be set correctly In particular, the brightness and contrast
controls on CRT monitors have a radical effect on the tone and color the
monitor produces. Once they're set correctly, makesure that they stay set
that way by taping them down or using the monitor's setup features.

Controlling Variability
Even if you control all the aforementioned variables, at least some of
your devices will still exhibit some kind of variability. CRT monitors drift
over time, color laser printers react to changes in temperature and humidity, inkjet printers almost invariably need some drying time for the
ink to reach its final color, and printing presses are subject to so many
mechanical. chemical. and physical variables that a slew of books have
been written on controlling them. For color management to work, you
need to account for the inherent variability in your devices. There are
t h e ways to do this:
b

Calibrate the device. changing its behavior back to the state in which
ir was profiled.

Make a new profile that describes its new state.

Collect data that describes the range of variability you deem acceptable, average that dara, and build a profile from the average.

?S.pically, you need to use a combination of all three methods. But

to control variability, you need to know that it's happening. You could
just wait until you see that something's wrong. but a much more reliable
method is to use measurements lo track the variation.

The Multiple Roles of Measurement

Measurement plays a key rule in both calibration and profiling. If you're
serious about building your own profiles, you need some kind of measuring
instrument. We'll discuss the different types of instruments in Chapter 6.
BftildingD~pla.yProflles,
and Chapter 8, BftildingOurputProjilps,but for
now, we'll point out the many useful things measuring capability brings
outside of the profiling process.

190

Real World Color Managcnunl, kcond Edition

You need to make measurements to build display and output profiles.

(The only reason you don't need to make measurements to build input
profiles is that the target you use to do so has already been measured for
you.) But measurement plays other, equally important roles.
We use measuring instruments to calibrate our various devices, or at
least those that can be calibrated, but how do we knowhow often weneed
to calibrate a particular device? Measurement provides the answer. How
about those devices that can't be calibrated? Keeping a log of measurements lets you decide when thedevice has changed sufficiently to require
reprofiling.
We don't mean to suggest that you go hog-wild and measure everything
in sight every 10 minutes, but investinga modest amount of time making
some carefully planned measurements both before and after profiling will
save you time, frustration, and premature hair loss.

Evaluating Devices
The first role of measurement is to erfaluatethe device--to find our just
what kind of a beast we're dealingwith. If it's a monitor, can It still reach a
high enough brightness level while maintaining a decent black? If it's an
output device, doesit produce clean neutrals and reasonably even tonal
gradations whiie achieving the best color gamut of which it's capable?
Measurement can help you answer all these questions. When the answers aren't to your liking, measurement can play other roles:
b

You canuse measurements to calibrate the device to a better-behaved

state.

You can use measurements to arrive at better driver settings.

You can use measurements to decide that the device in question simply isn't a candidate for color management.

We'll go into much more detail on the kinds of issues that affect different types of devices in the next three chapters. For now, we just want to
give you the Big Picture.

Chapter B Hrasurmrml, CaUbrallon, and ProcessControl

141

Callbratlng Devlm
As we said earlier, calibration means adjusting a device's behavior so that
it produces a specific, known response to stimuli-in plain English, it always produces the same color in response to a given set of numbers. But
we don't just calibrate our devices to any old state. In fact, there are three
possible goals for calibration, which we list here in order of importance
for color management.
b

Stability. Profiling is much more likely to be successful when the device we're profiling is stable, so that the same stimulus willalways provoke the same respons+the same set of RGB or CMYK numbers will
always make the device produce the same color.
When the device's response drifts over time, the profile becomes progressively less accurate and the colors get further and further away
from the desired ones. To keep the color right, you must calibrate the
device often enough to bring its response back to the desired aim
point. If the device has no user-accessible calibration features, you
need to track the changes so that you knowwhen it's time to reprofile
(see"Monitoring Devices,"later in this chapter).
Optimization Oiearization). Once we've made the device stable, a
second objective is to make the device perform optimally, so that we're
using as much of the device's dynamic range and color gamut as possible, while still delivering smooth and predictable tonal gradations
(see Figure 5-3).

Before and after

linearization

199

Real World Cola Management, Second Edkion

Simulation. If we've managed to make the device both stable and
optimal, we may want to take the extra step of simulating another
device. Forexample, we may adjust the white luminance of a monitor
to match the brightness of white paper in a Light box. or adjust the ink
curves of a proofer to simulate a certain press. As long as we can do this
without compromising thedevice's behavior, it can make less work for
the color management system.
These three goals sometimes compete. We might be able to tweak a
press to produce denser blacks by increasing ink densities, but doing so
may make it less stable. In this case we'd almost certainly choose stability
over dynamic range.
Once you've got the device behaving the way you want it to, it's time
to profile it, which again uses measurement. But the role of measurement
doesn't end when you build the profile. Most devices are subject to some
kind of variation over time, and if you don't account for that variation your
profile will become progressively less accurate. So the last role of measurement is to track your device's behavior over time and make sure that
it's still behaving the way the profile says it does.

Monitoring Devices
The snnplest way to check for device drift is to wait until you notice that
the color is wrong. Unfortunately, it's also the most expensive way, particularly if you notice that the color is wrong on a press run costing tens
or hundreds of thousands of dollars. It's also inefficient: most color outcomes involve at least two, andsometimes more than two profiles. If you
wait until something has gonewrong, all that you know is that something
has gonewrong-you don't know which of the possible suspects is really
the culprit.
Afewwell-plannedmeasurements,on theotherhand, can tell you when
a device is drifting too far and when it's time to recalibrate or reprofile.
Measurement can help you determine the drying time your inks need on
your inkjet paper, or how often you need to recalibrate your color laser
printer, and can alert you to unannounced manufacturing changes in
consumables, saving you a ton of time and frustration This is what good
process control is all about.

ChapIer 5: McaPurrmcn(, Callbration, and P r o m Contml

123

Working Outside the System

The big lesson you need to take away from this chapter is that color management IS the map, not the territory, and all too often the territory is
shiftingsand. No matter howgoodyourprofiles are, they only reflect what
your devices were doing when you captured the input target, printed the
output target, or measured thecolor swatches on your display. Unless you
take positive steps to make sure that your various devices keep behaving
the way the profiles say they do, yourbest effortswill be doomed to failure
and frustration.
If,ontheotherhand,youspendamodestamountoftimebefore profiling
to make sure that the device is behaving properly, and after profiling to
make sure that it keeps behaving thatway. you'll create the necessaryconditions for the color management system to do its job-namely, helping
you produce great color with no surprises. In the next three chapters, we'll
lookindetailat allthethings youneed to do before, during, and after building profiles for your display, input, and output devices.

This page lefi blank. ..

Building
Display
Profiles
Your Window to Color

Who said "the eyes are the windows to the soul"?Well, Guillaume de Salluste
Du Bartas, the 16* century Frenchpoet and diplomat,did pen the line,"These
lovely lamps, these windows of the soul," and Shakespeare's Richard El did
say, "Ere I let fall the windows of mine eyes." But the line everyone remembers
seems to stem from the 1955 movie, "The Lady!dlers," where Alec Guinness'
Professor Marcus asks, "And didn't someone say, 'The eyes are the windows of
the soul'?" and Katie Johnson's Mrs. Wilberforce replies, "I don't really know.
But, oh, it's such a charming thought, I do hope someone expressed it." Of
course, since then many people have.
Whether or not you happen to be one of them, one thing is certain: your
monitor is the window to the world of digital color. Monitor profiles-or
more accurately, display profilesare key to the big payoff of color management: the instantaneous preview of all your color-managed materials.
We can't stress enough the importance of good monitor calibration and
profiling to a well-oiled color management system.You maybe able to get
by withvendor-supplied profiles for input and output devices, but effective
color management really demands a custom profile for each display.

146

Real World Color Management, Second Edition

As with any other type of profiling, buildingdisplay profiles is a process
of comparing known values with measured ones-in this case, the profiling software displays aseries of color patches with known RGB values, and
compares them with measurements from a calorimeter or spectrophotometer.And aswith any other type of profiling. you want to get thedevice
behaving properly before you profile it. The difference between monitor
profiling and other types is that monitor profiling packages actually help
you get the device behaving well before you profile it, because they combine the functions of calibration and profiling.
This occasionallygives rise to some confusion-see the sidebar,"Muddying thewaters," later in this chapter-but generallyspeaking, it'seasier
to calibrate and profile displays than it is to calibrateand prolile just about
any other type of device.
Note that we say calibrating and profiling displays not monitors.\'ou're
really calibrating and profiling the combination of the video driver, the
video card, and the actual monitor-inshort, the whole display system.
So it'svetyhelpful to understand just how monitorcalibmtion isachieved.
Profiling the calibrated result is really the trivial part of the exercise.

Display Calibration
-

\{"hen you calibrate a display you (oryourcalibrationsohvare)adjust up

to four things, depending on the displaytechnology
b

The luminance ofmonitorwhite, expressed either in foot-lamberts or

in candelas per meter squared (cdlm2).

The color of monitor white, usually expressed in Kelvins.

The tone response curve of the display system, expressed as a gamma

value.

+ Optionally, the Iuminance of monitor black, expressed either in footlamberts or in candelas per meter squared (cdlrn",
Where you make these adjustments depends on the capabilities ofvolu
monitor.

Chapter 6: Building Diglay Pmfllu

197

What to Adjust
Thereare two ways to change the behaviorofadisplayYoucan adjust the
controls on the monitor itself, or you can adjust the signals that get sent
to the monitor by tweaking the d u e s in the video card's lookup table, or
videoLUT. It's always preferable to make the adjustments to the monitor
itself, because when you tweak the videoLUT, it's essentially the same as
editing an 8-bit channe&you start out with 256 levels, but you end up
with a smaller number.
However, the only way to adjust the display gamma is in thevideoLUT,
so you'll always be doing some tweaking t h e e t h e goal is to keep the
videoLUT tweaks to a minimum. The degree to which you can do so is
dictated by the controls your monitor offers, and, to some extent. the
controls offered depend on the type of monitor you're using.
r Theworkhorse ofcomputerdisplays is still the CRT(cathoderaytube).
Acathode rav isa stream of electrons that are fired from one end ofan
enclosed glass tube to the other.The electrons originate horn elecnon
guns that spray a beam of electrons onto the inside face of the tube,
which is coated with phosphorschemical compounds that kick off
photons (light) of specific wavelengths when they're strudc by electrons. Color monitors use three phosphors in the mixture coating the
inside of the CRT-the red phosphor, the green phosphor and (vou
guessed it) the blue phosphor. The voltages sent from the video card
control the bursts of electrons from the electron gun, and hence the
amount of light emitted by the phosphors.
b

The new kid on the block, and gaining fast in the world of computer
displays, is theLCD (liquidcrystal display).Theliquidcrystals thatgive
this type of monitor its name have the peculiar property of changing
shape in response to electrical currents. When sandwiched between
layers of polarized glass or plexiglass. they act as filters that modulate
the imcklight, a lluorescent light behind a diffuser that produces all
the light the display emits.

monltol CofItrdr. Every CRT we've ever seen bas controls for contrast and brightness (whichreally control white luminance and black luminance, respectively). Most offer some degree of control over the color
of white, either as a series ofpresets. as a continuously variable control.

198

Real World Color Management, L w n d Edition

or as separate gain controls for the red, green, and blue channels. A few
high-end monitors also offer red. green, and blue biascontrols, which set
the black level For each channel.

LCD monitor controls. Becauseofthe way theywork, the only real control
possible on most current LCD monitors is the brightness control, which
controls the brightness of the backlight. Some LCDs, particularly those
with analog rather than digital interfaces, supply software controls that
mimic those found in CUTS, but they're just tweaking the videoLUT, not
adjusting the behavior of the monitor itself.
The upshot of all this is that with a CUT monitor, you may be able to
adjust the white luminance, black luminance, and color temperature in
the monitor itself, while with an LCD monitor, the only thing you can
really adjust in the monitoris the whitelurninance.This hassome implications for your choice of calibration aim points See "Choosing Calibration
Settings,"later in this chapter.
However, LCD display technology continues to evolve, and during this
book's lifespan, we expect to see LCD displays that use red, green, and blue
LEDs to produce the backlight. In theory, such displays would allow you to
adjust the color temperature of monitor white in addition to luminance.
Also, we know of at least two L C D s t h e EIZO ColorEdge CG18 and CG21that have their own 10-bitlookup tables built into the display. EIZO's calibration software uses these 10-bit tables to adjust color temperature and
gamma, so the calibrated display is still capable of producing the full 256
shadesperchannel. Note, however, that LCD displays havea fixed contrast
rdtio-unlike CFtT displays, they offer no way to adjust black luminance
separately from white luminance, so as you increase the white luminance,
the black luminance is increased too.

Calibration and Profiling

- Tools
There are essentially four differen1 t p e s of calibration and profiling
packages.
b

Packages that are bundled with a monitor

Standalone insmment-and-software bundles

Muddying the Waters

We've mentioned several rimes
in this book the rule that proFJes

merely describ-a
single profile
doesn't change anything. But for
every rule, there's an exception.
and the exception to this particular rule is most display pmfiles. Display profiling blurs the
distinction between calibration
and profiling in w o ways:

The instrument and software

packages we use to calibrate
and profile monitors usually
perform both tasks as a single
operation.

Many monitor profiles contain

calibration information that
actually changes the behavior
of the display.

Fred remembers his jaw dropping in shocked dismay the first

time he selected a new monitor

profile in a beta version of what

was then called the Monitors &
Sound Control Panel and saw
his monitor flash to a new white
point.
This shredded the simple rule
that a profile merely described
device behavior but didn't change
device behavior. The culprit,
if you wish to call it that, is the
"vcgt" tag found hmany monitor
profiles,which stores the calibra.
tion data that gets downloaded to
the videoLUT-the gamma and.
possibly, white point corrections.
When you load a display
profile that has a vcgt tag, it
downloads the contents of this
tag to the videoLUT, and hence
changes the behavior of the
display. We confess that we have
mired feelings about this, and we
could bash out all the arguments.
pm and con, but the bottom line

is that it's a done deal-monitor

profiles can, and often do,

change the behavior of the
display system.
There are still a few monltor profiling packages that don't
write a vcgt tag into the profile.
Instead, they use a startup application, which downloads the
calibration data to the videoLUX The major downside lo this
approach is that if you choose a
display profile written by one of
these packages after startup, the
matching calibration data probably won't get downloaded to
the videoLU7, with Lhe unhappy
result that the display won't behave the way the profile thinks it
does.
We suspect ha1 this approach's
days are numbered for that very
reason. But if you do use such a
package, be careful.

Standalone s o h a r e packages that support different instruments

Visual calibrators
Let's deal with the last firm. If you're at all serious about color management, we don't recommend visual monitor calibration. The very same
adaptability that makes our eyes incredibly useful organs fot living on Planet Earth makes them pretty useless as calibration devices-the goal of calibration is to return the device to a lolown, predictable state, and while our
eyeballs areverygood at making comparisons, they'relousy a t determining
absolutes. If your viewing environment is completely stable--essentially,
a windowless cave-you m q b e able to get by with a visual calibrator, but
in our opinion, you'd be penny-wise and pound-foolish to do so. But if you
must, see the sidebar, "Visual Calibration," later in this chapler.

130

Real World Color Management, Second Edltlon

Monitorlblibrator bundb. Monitors that come with a bundled calibrator are, thus far, invariably CRTs. These are "smart" packages in the sense
that the communication between monitor and host is two-way, allowing the
calibration software to adjust the monitor's internal controls in response to
feedback from the measuring instrument Early models used a serial cable
to communicate with the host CPU, but the current ones all use USB. (This
means that they occupy two USB ports, one for communication between
monitor and host, the other for the measuring instrument.)
Examples include the Sony Artisan Color Reference System, the LaCie
Electron/BlueEve line, the Mitsubishi Spectraview line, and the Barco
Calibrator V line. The less-expensive packages typically use off-the-shelf
general-purpose calorimeters that will work with any CRT monitor.
The Sony Artisan uses a colorimeter with filters specifically designed
forthe phosphor set used in the Artisan monitor-so itwon't work as well
with other monitors-whiletheBarco ReferenceCalibratorVactuallyuses
a monochmmeluminance meter-a simplephoton counter-and gets its
color information from factory measurements of the phosphors in each
specific unit The phosphors don't change color over time, though they
lose their ability to emit photons, so this approach actually results in very
accurate calibration and profiling.
One huge advantage of these bundled systems is convenience. They
automatically adjust the monitor's brightness, contrast, gain, and bias
controls to reach the specified luminance and color temperature, and
tweak the videoLUT only to achieve the requested gamma. Calibration
and profiling with these packages is typically a one-button operationyou attach the measuring instrument to the screen, press the button. and
walk away. It's almost impossible to make mistakes with these systems.
Asecond less-obvious advantage of thesesystemsis that they can often
make finer adjustments than are available through the display's frontpanel controls, so they can create amore accuratecalibration thanispossible with a human operator making adjustments through the display's
controls (see Figure 6-1).
With the advent of the DDC2 (Display Data Channel) standard, which
is incorporated in just about every CRTmonitor currently being sold, any
DDC2-compliant monitor, when equipped with a suitable cable, has the
monitor.Thusfar, the only standalone propotential to becomea"smartW
filing package to exploit the DDC2 standard is Integrated Color Solutions'
BasICColor Display-we're still waiting for others to follow suit.

Chaptu h Building Display Profiles

Standalone Instrument-and-software bundles. At the time of writing,

the prominent contenders in this category are GretagMacbeth's EyeOne
Display, Monaco Systems' MonacoOPTMXR, and ColorVision's Monitor
Spyder with either PhotoCAL or the fuller-featured OptiCAL.
The packages differmainly in instmment support. ColorVision'sPhotoCAL and OptiCAL support several third-party instruments, while the GretagMacbeth and Monaco solutions only work with their own instruments,
the EyeOne Display and the Monaco OPT% akathe X-Rite DTP 94, respectively.While all the aforementionedinstruments are calorimeters, EyeOne
Monitor also supports the EyeOne spectrophotometer. (Arguments about
whether a colorirneteror a spectrophotometer is the better instrument for
measuring monitors tendquicklyto headinto angels-dancing-on-pinheads
territory. In practice, we can't say thatwe'vefound that one instrument type
is invariablysuperior to the other, and have had great success with both.)
The only real advantage we've seen from standalone bundles over separately purchased instruments and software is one-stop shopping, both for
the initial purchase and for after-sales support.

Stanrblono software and imtrummt+ Most monitor calibration packages

try to support the widest range of instruments possible.

UP

Real World Color Manylcfinnl, Second Edition

Monitor calibration and profiling is a pretty straightforward process.

We find that the main differences between packages are the list of supported instruments, and, mainly, the licensing terms. The latter isn't an
issue if you're only worried about calibrating a single monitor, but if you
need to calibrate all the monitors in a shop, the varying license policies
can add up to significant differencesin cost.

Instruments
Once known affectionately as hockey pucks, the instruments that you
attach to the center of your monitor now come in so many shapes and
forms that the epithet rarely applies. The puck stops here, so to speak. (We
couldn't resist.) See Figure 6-2.

Warning: If you're calibrating an LCD monitor, make sure that youuse an

instrument expressly designed for LCDs. The suction cups used to attach
calibration instruments to CRTs will rip the coating right off the front of
an LCD monitor, rendering it useless. (Dead LCD monitors aren't heavy
enough to make good doorstops.)
Some instruments are more influenced by ambientlight than0thers.A
monitor hood can help shield the face of the monitor from ambient light,
so that less of it is nflected into the instrument With CHT monitors, we
generally recommend working in a fairly low ambient light anyway, so if
you find that ambient light is making your instrument produce inconsistent results, you may want to consider paying some attention to your
viewing environment. See "ViewingE!nvimnment" in Chapter 9, Evaluating and Editing Profiles.

Chapter 6: Bulldlng Dlrpky Protikr

n3

Choosing Calibration Settings

To agreat extent, thewhole nohon of inonitorcalibrationstandardsstems
from the days before the displav was taken out of the color management
chain, and represented an attempt to make all monitors behave identically by calibrating them to the same white point and gamma. Ofcourse,
this never worked, and has been supplanted by the current approach of
color-managing the data that gets sent to thevideo card, using the profile
that represents the behavior of the specific display in question.
So fmm the narrow color-management point of view, it doesn't really
marter what calibration settings you choose--all that matters is that your
display profile represents the behavior of your display as accurately as
possible. But other considerations come into play For example, if some
of your work lies outside the color-management realm, you may want to
calibrate your monitor to a recognized standard. Andeven if all ynur work
is color-managed, yuur display will behave better at some settings than
others. The one thing to avoid is making major corrections in the videoLUT,
because doing so reduces the number of discrete levels your display can
reproduce-the bigger the correction, the fewer levels.
Adjustments you make to the monitor itself, on the other hand, are
"free"-there's no loss-so a major consideration in choosing calibration
settings is the means availahleto achieve them. AU calibrators ask you for
two target values:
r white point
b

gamma

For CRT displays the three of us, by separate paths, have come to h e
same recommended calibration settings: calibrate your monitor to a
white point of 6500 K and a gamma of 2.2. You can just take our word for
e white point-6500 K.
it, but we think it's useful ro understand u h ~ w like
and gamma-2.2, so we'll discuss that in the next section.
For LCD displays, the only thing you can typically adjust on the monitor
itself is the brightness of the backlight. Changes to the color temperature
are carried out in the videoLUT, so we recommend using the native white
point for LCDs whenever the software allows us to do so. If native white
point isn't an option, we use D65 because the color temperature of most
LCD backlights is closer to D65 than to anything else.

134

Real World Color Msnagemcnl, kcond CdlHon

Target White Point

The while pomt of the monitor plays an important role for your eye. As
described In Chapter 1, your eye judges all colors relative to some neutral
baseline that it considers white.The three of us unanimously recommend
that you calibrate your monitor to 6500 K even though many people think
of 5000 K as the standard viewingwhite point in graphic arts. Here's why.
The eye has a tremendous ability to adapt to different white-point environments. Howevec the eye works best when it's operating in a white
point closest to that environment most familiar to it through millions of
years of evolution-namelydaylight. So the discussion quickly reduces to
which of the two most commonly used daylight standards. DSO (5000 K
correlated color temperature) or D65 (6500 K correlated color temperature), is best (See Chapter 1 if you're unfamiliar with these terms.) Ifyou
have a D50 viewingbooth, this might seem to tip the scales toward setting
your monitor to 5000 K so that your monitor and print-viewing environment haveidentical white points. But longexperience bas toldus that this
doesn't work the way the tbeorywould seem to predict--see "Monitor-toprint comparisons." in Chapter 9.Evaluatingatrd EditingProfiles.
Asecond factor is that many uncalibrated CRTs, especially older models, are pretty darned blue, with a color temperature closer to 9300 K,
though the better current CRTs have a native white point closer to 6500
K. In either case, to move the white point to 5000 K.we have to limit the
output of the display's blue channel, lowering the overall brightness and
dynamic range. This is why many people (includingus) often find a 5000
Kmonitor to be a bit too dim, dingy. and well too darned yellow.
So instead, it's worth remembering the sentence that started off this
explanation.Theeye has a tremendous abilityto adapt to different whitepoint environments. The eye takes a little bit of time to adjust to a change
in brightness, but it has little trouble looking at a color image in a 6500
K monitor and then moving to view the same image printed out and
mounted in a viewing booth. It's the relationships within the image or
page that you're evaluating. As long as you give the eye a good adaptation
environment, and both environments are of approximately equivalent
brightness, then you should have no problems.
If yoursoftware has more than just two choices forwhite point. you have
another option-you can perform iterative calibrations to find the white
point setting that best approximatesa sheet of paper in yourviewing booth.

...

...

Some people in extremely color critical environments, with control over

all of their equipment, including the paper being used, have found this to
be useful. We, however, prefer to deal with the color of paper white in the
printer profile, and concentrate on matchingthe brightness of the monitor
and the viewing booth.

Target Gamma
The recommended setting for gamma depends somewhat on how much
work you do outside the color-managed environment. Without color
management, the general practice has been to choose a gamma that best
simulatedthemid-tone characteristicsof the target medium-in English,
if you'regoing to print, you want your monitor to have similar tone reproduction characteristics to your printer.
This, supposedly, is why Macintosh displays have traditionally been
assumed to have a gamma of l.&the default table in Macintosh video
cards isset to produce an approximate display gamma of 1.8, and calibrationinstructions have longtouted 1.8 as the "standard Macintosh gamma"
when calibratingyourmonitor.Why1.8?Amnnitorgamma of 1.8 roughly
approximated the dot-gain curve of the Apple Laserwriter. This served
as a poor-man's color management (actually a grayscale management,
as this was well before the days of color management, since the displays
being used at the time were monochrome). If an image looked good on
a monitor calibrated to gamma 1.8, then it would look reasonable when
printed without much adjustment.
But with color management, simulation is now the job of the color
management system, and specifically, of the printer profile. So instead,
we recommend that you calibrate your monitor to a gamma of 2.2, for
the simple reason that, in all our testing, we've found that calibrating to
around gamma 2.2 produces the smoothest display of gradients, with
little or no visible banding or posterization.
If you're picky, like us, you may want to go further, and use the techniques described in "Checking the Display," in Chapter 9. Evahratingand
Editing Profiles, to lind the ideal gamma for your particular display. On
the other hand, if you've been happily calibrating your monitor to D50.
to gamma 1.8, or to both, don't feel you have to change it just because we
say so--as long as you have an accurate monitor profile, the differences
between the various calibration aim points are pretty darned subtle.

l36

Real Workl Color Management, S m n d Edition

How Often Should I Calibrate?

With CRT monitors, we recommend calibratingweekly. At minimum, you
should calibrate monthly-some of the high-end systems like the Barco
CalibratorVand the Sony Artisan are pretty insistent that you recalibrate
after 30 days. We know people who calibrate their monitors every day.
which seems excessive, but harmless.
With LCD monitors, the jury is still out.The dyes in LCD cells wear at a
much slower rate than the phosphors in CRTs-itS unlikely to be an issue
in a human 1ifespan.The backlight. however, will decay slowly over time.
We calibrate our LCDs weekly to be on the safe side, but you can probably
get by with less-Erequent calibration than with CRTs.

Before Calibration
The tirst thing to do before calibmling your monitor is to make sure that
it's worth calibrating. On a CRT, hrm the contrast control all the way up;
on an LCD, turn the brightness control all the way up. If the result issomething other than a display that's at leasta hair brighter than you'd like, the
monitor is probably a candidate for replacement rather than calibrationcalibrators work by turning things down, so if the display, running wide
open, isn't as bright as you'd like, it'll be even worse after calibration.

Warm-up Time
Before calibrating a CRT, make sure it's been on and in use for at least 30
minutes, and preferably for an hour so that it's reached its stable operating temperature. If you have any energy-saving software that turns the
monitor off when not being used, make sure this doesn't kick in during
the warm-up period. (Ifyou use this as your excuse for a littleweb surfing
every morning, we won't tell anyone.) Regular screensavers (those floatie
fishies or flying toasters) are fine, as are the simpler ones that just display
a black screen. as long as they don't actually turn the monitor off.
Since we wrote the first edition of this book, we've realized that LCD
displays also need at least 30 minutes warm-up time, maybe more.Various
objective evaluationsof LW)display warn-up time range from a low of 30
minutes to ahighof90minutes forthe fluorescent backlight to reach stable
output. We've also seen some work that suggests the LW)filters themselves
may he influenced by temperature, but the effects are much smaller.

Chapter 6: Bulldlng Display Profiles

137

Resolution, Refresh, and Other Monitor Settings

hlakesure that all the other monitor settings are finalized before you calibrate. Just changing the resolution from say 1280 1024 to 1152.864, or
changing the refresh rate from 85 to 75 Hz, can affect the overall brightness output of the monitor. So can changing the geometry settings on the
display-reducingthe size of the displayed image on the monitor targets
the electron gunson asmaller aren of phosphors, which slightly increases
the brightness of the image.

Tip: Extend the Life 01 Old Monitors. If you followed the instructions
on evaluating your monitor and didn't achieve satisfactory brightness, it
may be time for a new monitor. However. to try and squeeze some extra
brightness back into the monitor, you can:
F

use the geometry controls on the front panel of the monitor to reduce
the horizontal and vertical size of the displayed image

reduce the refresh frequency from, say, 85 Hz to 75 Hz

reduce the resolution from, say 1280.1024 to 1152.864 or wen
800x600

The refresh frequency and resolution settings can be found in the Display Settings or Control Panel on your system, but be careful that you
select a combination that supports true 24-bit color (millionsof colors).
It's OK to change these settings, but you must do so beforeyou calibrate.
and if you change any of these parameters, you need to recalibrate for each
change. Also, it should go without saying-but we will anyway-that you
should set your monitor to display true 24-bit color (millionsof colors).

Cleaning the Screen

It's important to make sure your screen is free of dust and fingerprints
before calibration, because they can introduce inaccuracies into the readings. In the case ofCRTmonitors, they can also cause the sensor to fall off
halfway through the calibration process, which can be a major frustration
and a hazard for your keyboard and measuring instrument
Use a weak solution of mild detergent or a cleaning product specifically
formulated for monitors. Don'tuse glass cleaners containing ammonia or
other hash solvents-most high-end CRTsuseantiglaecoatings that household glass cleaners wreck: the coatings on LCDs are even more fragile.

138

Real World Color Iknlgcment, Second Edition

Precalibration-Setting the Analog Controls

-

In addition to specihing thecolor temperatureand gammaof the displa!:

calibrators ask you to set the dynamic range by choosing white and black
luminance. In some packages, this step-setting the analog controls on
the monitor-is handled as part of the calibration process. Others, such
as ColorVision's OptiCAL and PhotoCAL, use a separate application, PreCAI., to step you through optimizing the monitor's contrast, brightness,
and color temperature settings. Smart monitors with bundled calibrators.
such as the LaCie ElectronlBlueEyeand the Sony Artisan Color Reference
System, handle this task automaticdy. and usually more accurately.

Setting White Luminance

Some packages allow you to specify a value for white lumin,anc+how
bright you want the monitor to be. Others simply adjust the monitor to an
internal [and sometimes undocumented) preset value. With CRT monitors, when given a choice, we opt for avalue somewhere between 85 and
95 cdlm', or 24 and 28 foot-lamberts.
Most of today's CRTs can reach a higher luminance--sometimesmuch
higher-for a while. It's always temptingto crank up the whiteluminance
to get the maximum possible contrast ratio, but when you do so, you
shorten the monitor's useful life. A decent CRT should be able to achieve
our recommended luminance for at least three years, often for longer.
The default brightness of LCD displays is usually so high it needs reducing, sometimes by as much as half. Good calibration packages guide you
in setting brightness. This is important because we find that the half-life
of the backlight isabout 18 months--ifyou drive the display at full brightness, it will be half as bright in 18 months and there won't be anything
you can do about it. Droppingthe brightness from the get go will give you
headroom to increase white luminance as the backlight ages.
Unless we're using a calibrator that adjusts the monitor a~~tomatically,
we start out by turning the contrast all the way up and the brightness all
the way down-most calibrators tell you explicitly to do so, anyway. The
non-automatic varieties generally display the target luminance and the
current measured luminance. You then adjust the contrast control (on
CRTs) or the brightness control (on LCDs) to make the two match. Figure
6-3 shows some typical examples.

Chapter k Bufldlng Dbphy Pmfikr

Figure 6-3

Setting white luminance

ColorViiion's PreCAL lets you take a memurement and reports the measured
luminance.You adjust the contrast control and repeat the process until you
achieve the desired luminance.

GretagMacbeth'sEyeOne Match, left,

measures continuously.Youadjust the
contrastcontrol until the white indicator
lines up with rhe black one.

140

Rsal Wotld Color Manrg.~nt,Second Wltion

Setting Black Level

Setting the black level is quite a bit trickier than setting the white. If you
set it too low, you'll clip some shadow levels, and if you set it too high,
your blacks will be washed out. To further complicate matters, most of
the inshmments we use to calibrate monitors have diminishing accuracy
as the sample gets darker, and the light output of CRT monitors gets progressively less stable the closer it gets to black
This is where systems that adjust the monitor's controls automatically
have a huge advantage. They take hundreds of measurements and average
them. Calibratorsthat step you through these same adjustmentsmanually also take many measurementsofblack, but the problem then becomes
how to present these measurements to the user in a useful fonn.
On CRTs (and on avery few LCDs),you adjust the blacklevel using the
brightness control (whichisreally what the tecbiescall an offset).On most
LCDs, the blacklevelisn't controllableseparately fromwhiteluminance-the only real control is the brightness of the backlight-so you simply have
to accept what the calibrator gives you. We've seen many cases where the
indicator of the current measured black level jumps around, making it
quite hard to determine when you've arrived at an optimal setting. In this
situation, the key ispatience. Don't expect the indicator to magically stop
moving around-it won't. Instead, just keep making smalladjustments to
the black level. Gradually, the results will converge on the target point.
Some calibrators let you set the blacklevel visually instead-they usually show a dark gray patch against a black one. With this approach, it's
veryeasy to set the blacklevel toolow. which results in clipping potentially
useful shadow values.The goal is to set the black level a hair (since you're
working visually, more technical definitions don't apply) higher than no
signal at all. Figure 6-4 shows some different approaches to setting the
black level.
Setting the black level is by far the most difficult part of monitor calibration. It's worth learning the quirks of your chosen calibration system
and working with them to get the most consistent results possible. See
"Black-point check" in Chapter 9, Evaluating and Editing Profiles, for
some techniques that can help you refine the black-point setting.

Figure 6 4
Settingblack level

ColorViiwn's
OptiCAL, Wlefl.
uses a visual target
to help you set the
black leveL

GretagMacbeth's EyeOne Match, above leA andMonaco OptixPro. above right, both measure black
continuowly and provide feedback via a slidez When it lines up with the target, you've set the bhck
level correctly.

149

Real World Color Management, Second Edllion

Setting the Color Temperature

In CRT monitors, the color temperature is adjusted by changing the individual gains on the red, green, and blue guns. With LCDs, it can only
be accomplished by filteringthe backlight, so we generally leave LCDs at
their native white point, which is usually close to 6500 K anyway
Rut the level of control that CRTs offer varies From model to model.
'IS.pically CHTs offer one (or more) of the following:
Several preset color temperature settings (typically 5,000 K. 6500 K.
and 9300 K, though some offer additional presets)
Continuously variable color temperature
Gain controls for two of the three guns
b

Gain controls for all three guns

So how you get to the desired color temperature depends on the monitor controls. Again, with automaticsystems, thesoftware takescare of the
necessaryadjustments automatically Many of the manual calibrators tell
you to skip this adjustment if your monitor doesn't have individual R. G,
and B gains, but we find it useful on all CRTs (see Figure 6-51.

CRTs with pmotr. We always start out with the preset closest to our desired color temperature, but the presets are often off by a considerable
amount, almost invariably givingalower color temperature than the preset value. In that case. we may t~ going to the next-higher preset. For
example, if the 6500 K preset measures as 6100 K, and the monitor offers
a 7500 K preset. we'll use that instead. Once we've determined the best
preset to use, we'll skip this step in future calibrations.
CRTs with continuously variable color temperature. Again, the colortemperature labels on the color-temperature control are often significantly off from the real, measured color temperature.We adjust the color
temperature to get as close as possible to the desired result.
CRTs with gain controls. We find that it doesn't make a whole lot of difference whether the monitor offers control over two or threeof the channels-we rarely adjust more than two, anyway. Most calibrators show you

b
u
n6-5

Setting white point

Clockwisefrom top: Colorlriion'sPreCAL, Monaco Optix Pro, and

GrefagMacbeth'sEyeOne match all prouideagraphic display of the
currentRGB leversand how they need to be adjusted to achieve
the requested white point.

where the guns are now, and where they need to be. Make the adjustments
until you get as close as possible. (How close this is varies considerably
with diffemnt monitors, dierent instruments, and different soafter you've gone through the process afew times you'll develop a sense of
how far you can go before you reach the point of diminishing returns.)

144

Real World Color Management, Second Wlllon

Calibration and Profiling

-

At this point, you've done all the difficult stuff-setting the analog con-

trols is really the only part of calibrating and profiling the monitor that
requires user intervention. The software displays patches on the screen.
looks at the measurements from the instruments. and makes the necessary tweaks to thevideoLUT.Then it displays some mote patches. looks at
the measurements, and builds a profile. There are, however, a few things
that can go wrong.
Make sure that some screensaver software doesn't kick in during calibration. Any good calibrationsoftware should detect when this happens,
and, in fact, most have code that prevents it. Nevertheless, if your floatie
fishies appear during calibration,you mayhave to adjust the screensaver's
settings, or disable it altogether. And of course, you'll have to restart the
calibration process.
It's possible to throw off acalibration by allowing yout mouse pointer to
move across the measurementarea duringa measurement. Most calibrator
programs nowadays are able to hide the pointer if it enters the measurement area, or at the least can detect when this happens and beep angrily at
you, but there are still exceptions. It's just good advice to keep the pointer
safely offto the side and don't touch anything during calibration.
If you use the calibration process as an opportunity to take whatever
kind of break you need, check that the calibrator is still attached to the
monitor when you come back The industrial design of instruments intended to attach to the monitorhas greatly improved since the early days,
but it's still by no means unheard-of for the instrument to fall off before
the calibration is done. In that case, you need to start over.

Saving the Monitor Profile

Regardless of what type of calibrator you are using, the last thing it does
seems like such a tiny afterthought, but is the key to getting this laboriously calibrated monitor into your color management system-saving
and naming the monitor profile.
Some people create a new profile each time by including the date in
the filename. We think this is silly-the condition that the old monitor

Chapta 6: Building Display Profiles

145

Visual Calibration
We're tempted to say that visual
calibrationisan oworon--talibration means bringhg a device
back to a known state, and the
adaptive nature ofour eyes makes
itjust about certain that you won't
be able to hit the same aim point
twice by visual methods alone.
We don't recommend visual calibration, but if you must, hen: are
some pointers that you may find
helpful.

hll the directions about warmup time and resolution settings

apply to visual calibration just
as they do to instrumented calibration.
fi

The viewing environment

is much more critical lor
visual calibration than for
instrumented calibration.
The ideal is a darkened room.
but if that's impractical, at
lmst try to make sure that you
always calibrate under thesame
lighting conditions--if the

ambient light is dramatically

different from one calibration
to the next, the results almost
certainlywill be, too.
t

NI visual calibrators start out

with an existing profile. Try to
choose one that bears a general
resemblance to the monitor
you're calibrating. Ideally, you
should start out with a profile
describing a monitor that uses
the same phosphors as yours.
Failing that, if your monitor is
an aperture-grille type (a TriniIron or Diamondtronl, choose
a profile that describes that
type of monitor, and if it's a
shadow-mask monitor, use a
shadow-mask profile. Sony and
Mitsubishi are the main pmvidersof the former type: I-litachiis
the main provider of the latter.
If the calibrator offers a choice

benveenasinglegammaandindividual gammas for red, green,

and blue, choose thelatter.

Gamma adiustments are

almost always carried out hy
having you move a slider until
a solid color marches a dither
pattern. It's easy to do this with
redand green. but it'sjustabout
impossible to do with blue.
So when you adjust the blue
gamma, don't look at the target-look at your neutral gray
desklopinstead Lofcourse. your
desktop pattern isa neutral gray
rather than pinkmarbleor your
fawrite picture of your cat). It's
very easy to see what's happening to the neutralsaim for the
blue setting that gives you a
neutral gray.

w \'lsual calibrators an, designed

with CRTs in mind. if you try to

use them with LCD monitors.
all bets are off, but we can just
about guarantee that you'll be
dkappointed with the results!

profile described n o longer exists since you've recalibrated the monitor.

We recommend that you just overwrite the previous profile--theold ones
are useless and just clutter your system.
Most profilers automatically save the profile in the correct location a n d
load it a s the profile that describes the monitor, even in dual-monitor
systems. Unless you really know what you're doing, let the software save
the profile m its preferred location, which varies on different operating
systems. On Mac OS 9, save the profile in the ColorSync Profiles folder in
thesystem Folder (somecalibrators save the profilein the Displays folder
inside the ColorSync folder-either way is fine).

146

Real World Color Margrmml, Suond Edition

On Mac OSX the situation is a little trickier. Some calibrators save the
profile inthe ILibrarylColorSynclPnfileslDisplaysfolderifyou're logged
in as an Admin user. themby making the profile available to all users.
Otherssave it in the -1Libra~lColorS~clProfileslDisplaysfolder-that's
your "Home" Library, which is unavailable to other users. Eventually, we
hope vendors will make up their minds whether monitor profiles should
go with the user or with the monitor, but meanwh~lewe advise saving
it in ILibrarylColorSynclProfiles whenever possible, so all users have
access to it.
On Windows 98,98SE, and ME, save the monitor profile in Windows\
Svstem\Color. On Windows 2000. and Windows XP, save the monitor
On Windows NT, save the
profile in WinNT\System\Spool\Drivers\Color.
profile in WinNT\System32\Color.

Piece of Cake
Monitor calibration and profiling is really very straightforward, which is
just as well, because it's the lynchpin of a color-managed workflow. Most
people say they want to trust their monitor, and calibration and profiling is the first step in attaining that goal. Often, though, it's an iterative
process. See "Checkingthe Display,"in Chapter 9. Eualuatingand Editing
Pmfiles, for techniques to help you evaluate and improve your display
calibration and profiling.

Profiles
Starting Out Right

To get the color you want, you first have to knowwhat that color is. The
fundamental role that input profilesplayis to tell the CMS what that color
is. We must stress upfront that input profiles don't automatically give you
great color or remove the need for color correction-they just tell the CMS
what colors your capture device sees. So a good input profilewill faithfully
render the dark and murky appearance of an underexposed image or the
washed-out appearance of an overexposed one.
Moreover, some types of input are difficult-to-impossible to profile.
Digital cameras shot in the field rather than the studio are difficult to
profile because the light source is all over the place. Color negative scans
are basically impossible to profile for a raft of reasons, the main ones
being:
b

Color-negative scanning targets are hard to find-Fuji is the only

source of which we're aware.

Unless you like orange, inverted images, you don't want to reproduce
what's on the film.

The orange mask on negatives varies so much with exposure that the
profile would only work reliably if your negatives were exposed very
close to the same way the target was.

148

Real World Color Managcnwnt, Second Edition

However, if you're scanning prints or reversal (positive)film, or you're

shooting digital captures under controlled lighting, good input profiles
can be significant time-savers. Scanner profiles are so easy to build that
there's relatively IittIe reason not to do so. Digital camera profiles, however, are quite a bit harder, for reasons that we'll examine a little later.
First, let's look at the similarities behveenpmflingscannersand profling
digital cameras.

lnput Profile Basics

1he process of building input profiles is the same as building other types

of profiles in that we compare device values-almost always RGB in the

case of input devices-with measurements in a device-independent color space such as CIE XYZ or CIE LAB. The difference is that with input
profiles, the measurements have, in the vast majority of cases, already
been done for you.

lnput Profiling Targets

lnput profiling targets always have two components:
b

The physical target that you scan or shoot

A target description file (TDF) that contains measurements of the
color patches on the target

The main differencebetween inexpensiveand expensive input profiling

targets is the accuracyof the measurements in the target description file.
Inexpensivetargets are manufactured in fairly large batches. and the measurements arecarried out o n a small sample of the total batch. As a result,
some targets'TDFs in the batch are moreaccurate than others. Expensive
targets are manufactured in smaller batches, and measured individually,
so the TDFs are generally a lot more accurate.
Of course, if you have an instrument capable of doing so, you can always measure the profilingtarget yourself and plug the measurements
into the target description file that accompanied the target. This is easy
to do with reflective targets, quite difficult with large-format transparency
targets [you need a transmissive spectrophotometer, and the affordable
ones force you to position each patch manually). and extremely difficult

Chmpkr Z Building input P d k r

149

with 35mm transparencies (you need a very specialized transmissive

spectrophotometer with a very small aperture, and they're both rare and
expensive).

kanna T a r n . By far the most common scanner targets in use are the
IT8.711 (transmissive)and IT8.712 (reflective)targets, which are available
from several different vendors on several different film stocks. Kodak's
version, the Q-60, follows the IT8 standard (see Figure 7-1).

b u n 7-1
The FT8.711 and

150

Real World Color Management, Second Wltlon

One frequently asked question is, do I need to make different profiles

for different film stocks?Wehesitate to give a definitive verdict, but based
on our experience,we don't thinkso.The image-forming dyes used in film
stocks as different as Kodak Ektachrome and Fuji Velvia are almost identical-the differences come from the image-capture dyes, and a profile
simply records how the capture device sees color. The colors may be quite
different on a Kodak and a Fuji IT8 target, but as long as the TDF is accurate, we've found that they produce extremely similar profiles.
We've built profiles using Don Hutcheson's excellent hand-measured
HCT targets (you can find out all about them at www.hutchcolor.com),using both the Kodak and Fuji versions, and we've found that they produce
almost identical profiles that are capable of characterizing scans from
either film stock very well. Of this much we're certain. You're definitely
better off buying one individually measured target than buying batcbmeasured targets on different film stocks.
We prefer the HCT target to the IT8 not only for the accuracy of theTDE
but also because it contains a better sampling of colors, particularlydark,
saturated colors, and we find that it simply produces a better profile than
the IT8 (see Figure 7-21,

Chapter Z Bulldlng Input Profiles

15l

Digital umcra t
a
mThere are only two digital camera targets that have
gained any significant degree of support in profiling packages: the 24-patch
Macbeth ColorChecker, and the brand-new (at time of writingl 140-patch
GretagMacbeth ColorCheckerSG, which replaces the now-deprecated ColorcheckerDCThetargetis specificaUydesignedforprofiling digitalcameras,
and includes a series of white, gray and black patches around the perimeter
designed to let the profiling software compensate for uneven lighting, along
with the 24 patches horn the original Color Checker (seeFigure7-3).
F&ure 7-3

Macbeth ColorCheckei

154

Real World Color Managemenl, Second Edition

The ColorChecker DC was a great example of good intentions gone

awry. The glossy patches had a tendency to turn black in cross-polarized
lighting setups, and the high patch count tended to make profiles that
were notable for their lack of smoothness.
In many cases, we still prefer the old 24-patch ColorChecker. One simation where we prefer the ColorCheckerSG (forSemi Gloss) is for profiling
cameras in the field under available (that is. uncontrollable) light, for the
simple reason that the black, white, andgray patches around the perimeter
of the target help you determine whether or not the target was evenly lit.
We do suggest that you measure the 24-patch ColorChecker yourself.
The formulation has changed at least three times over theyears. and while
the patches fromeach formulationappear similar to the human eye, they
appear rather differently to digital cameras.

Input Device Variables

As with any olher type of device, it'svital that you control an the variables

that can affect your inpol device's performance before you profile it. There
are really only three factors that affect capture devices such as scanners
and digital cameras:
b

The light source

The color filters

The software settings

In scanners, the light source is stable in all but the least expensive
models, and is almost always compensated for by an internal calibration
before each scan. In digital cameras, however, the light source is a huge
variable--this is why scanners are easier to profile than digital cameras.
How you account for differences in the light source depends on the capabilities of the camera.
The color filters on scanners and cameras change over the Life of the
device. but so slowly that it'll probably take five years or more before you
need to reprofile.
With scanners, the biggest variable by far is the software settings. Softwaresettings arealso importantwith digital cameras, but in rather different ways, so from hete on it makes more sense to deal with scanners and
cameras separately. We'll start with the easier option, scanners, first.

Chapter R Buildlng Input Pmfila

la

Building Scanner Profiles

Huilding scanner profiles is relatively easp-the key to doing so successfully lies in setting up the scanner software correcflp, then keeping it set
that way. The first goal is to get the scanner behaving consistently. Once
you've achieved that, the second goal is to get the scanner behaving
optimally

Stabilizing the Scanner's Response

Ib make a good scanner profile. the first thing you need to do is to make
sure that the scanner responds the same way to every image you scan. If the
scanner uses features that tailor its response to the image at hand, you're
back to chasing the proverbial moving target with the rubber ruler!
So you need to turn off any features that set black or white points,
remove color casts, or do anything else automatically in response to the
image content. Less obviously, you need to turn off any sharpening in
the scanner-sharpeningis also a feature that produces different results
depending on image content.
There are, however, hvo automatic Features that we've found are benign
with regards to color management [and are extremely useful in their own
right). Applied Science Fiction's Digital ICE surface-defect removal and
GEM (Grain Enhancement Module) noise-reduction technologies don't,
in our experience, interfere with color management, so ifyour scanner is
equipped with them, and you like what they do. by all means leave them
active. (Athird ASF technology Digital ROC, is inimical to color management, but since it's designedto work on faded or color-casted originalswhere you don't want to reproduce what's on the film-it's unlikely that
you'd wantto use color management and Digital ROC at the same time.)
A very few scanner drivers make it impossible to defeat the auto settings-the ones we run into over and over again are those for the Polaroid Sprintscan series. Not all models are affected, but some members of
the family have auto-exposure hardwired into the scanner firmware. You
can't really profile these scanners. If you're using a scanner whose auto
features are undefeatable, or if you're scanning color negatives, use lhe
techniques described in "Bringing Color into a Color-Managed Environment" in Chapter 10. Color-MnnngeemenrWorkflow, Instead.

154

Real World Color Management, kcoml Wition

Optimizing the Scanner's Response

Once the scanner is stable-you've turned off auto-everything-the next
task is to make sure that it's behaving optimally. The default tone curves
on most scanners, whether a lowly desktop flatbed or a Heidelberg drum
scanner, produce images that are pleasantly contrasty and saturated, but
tend to compress shadow detail and distort darker colors.
If your scanner software allows you to set the output gamma, we recommend setting it somewhere in the range of 2.63.0.See 'DbjectiveTest
ForInput Profiles" in Chapter 9.Emlrmh'ngandEditingPmfiles,
for atechnique that allows you to determine the scanner's optimum gamma. For
further reading,we recommend DonHutcheson'sexcellentpaperentitled
"Scanning Guide," which is available for download in PDF format from
www.hutchcolor.com.
Our usual approach to color-managed scanning is to run the scanner
"wide open" with no black-point or white-point correction, and capture
the maximum bit depth the scanner allows. We then do any necessary
corrections, including optimizations for differentoutput processes,in an
image editor such as Adobe Photoshop.
However, if vout scanner allows it, you can achieve significant productivity gains by performing color correction and conversion to output
space in the scanner software. The downside is that your corrections will
be based on whatever size prescan the scanner software provides. But
whether you bring raw scans onto animage editor or do conversions and
corrections on the scanner. the procedure for building the profile is the
same.

Scanning the Target

Thr target scan doesn't have to be particularly high-resolution-weusually aim for a file in the 5-12 MB range for 24-bit scans. 10-24 MB for 48bit scans. Make sure that you've set all the scanner parameters correctly.
and try to mount the target as straight as possible. If you're profiling a
flatbed scanner. it's a good idea to mask the rest of the bed-the area
that isn't covered by the target-to minimize distortions fmm unwanted
reflections.
We usually scan the target as a high-bit TIFF file. If your profilingsoftware doesn't accept high-bit TIFFS,you can downsample the target scan

to eight bits per channel in Photoshop, or just scan in 24-bit RGB. We

always open the target scan in Photoshop before feeding to the profiler,
and perform any necessary rotation and cropping (you set the final crop
in the profiling package-cropping in Photoshop just minimizes the
amount of data you're slinging around).We also check each patch for dust
and scratches, and spot out any defects with Photoshop's clone tool (see
Figure 7-41,

Tip: C h e d ~Your Polida Befort Opening the Targel Scan. When you're
building input profiles, it's vitally important that you don't change the RGB
values in the target scan, so make sure that you don't have any automated
conversions specified in your image editor. We also recommend that you
make a point of never, ever, embeddingaprofile in any color-management
target, including scans or captures of input profile targets.

Massaging the target scan

Straightening the target scan is easy in Photoshop.We use the measure tool to
measure the angle of one of the lines that should be horizontal, then choose
Arbitrary Rotate. The opposite angle is automatically entered, so clicking OK
results in a straightened scan. Make sure that the interpolation method is set
to Nearest Neighbor to preserve the integrity of the color patches.

156

Real World Cdw Management, Second Edition

Building the Scanner Profile

Once you've got the target scan in the best shape possible, all that remains
is to build the profile.The user interface may vary From profiler to profiler,
but they all work in fundamentally the same way.
First, you're asked to open the target reference file, which contains the
measured color data for the target. Then you're asked to open the target
scan (see Figure 7-5).
nnurc
- 7-5
Cropping the target scan

Profiling took such as GretagMacbeth's F'mfileMaker Pro,

aboue, let you crop the target exactly to the index marks so
that the sofnvare can identify the color patches correctly

The profiler now has all the information it needs to build the profile.
Some packages offer choices ofprofile size, gamut mapping, or gray tracking. We haven't found any consistency as to which options work best-it
varies from device to device and from profiler to profiler. Since the actual
profile-buildingprocess is so quick, we suggest you simply try each option
and then use the techniques described in "Input Profiles" in Chapter 9,
Evaluatingand Editinghfiles, to determine which options work best for
the device at hand.
Most profilers ask you to name and save the profile before calculating
it, and all reasonably recent ones default to the correct location for saving.
Once you've entered the name and clicked the Save (or Build, or Calculate)
button, the profiler builds the profile. On a reasonably current computer,
this process usually takes well under a minute.

Chapter 'l:Bullding Input Pmflkc

1 9

Building- Digital
Camera Profiles
The actual profile-building process for digital camerasis pretty much identical to that for scanners-you provide the profiling package \nth a target
reference file anda capture of the target, and then it cr~mchesthe numbers
and builds a profile. What makes digital cameras much harder to profile
than scanners is the inherent differences in the ways they're used.
Where scanners have a fixed and reasonably stable light source, digital
cameras do not. Moreover, scanners, for the most part. capture originals
that have already been reduced to three or four colors of dye or ink, while
digital cameras have to capture red-world objects in all their diversity.
So metamerism-where the capture device sees colors differently from
the way wedo-ismuch more commonly a problemwith digitalcameras
than it is with scanners.

Controlling the Camera's Response

With digital cameras, the strategy that works for scanners-locking the
device's response--is totally useless.You need the camera to react differently to different scenes so that it can accommodate lighting conditions
that vary in both dynamic range and color temperature.
Controlling the camera's response to dynamic range is basicphotography-lighting and exposure. There are plenty of good books on the subject, so we'll say no more except to stress, once more, that when you're
profiling acamera, lighting the target evenly and capturingboth the highlights and shadows, is critical.
Control over colortemperature, on the other hand, depends largely on
the capabilities of the camera. and on the file format you're shooting.

In-camera gray-balancing. Scanning backs and some three-shot cameras

let you gray-balance the camera itself-you expose a gray card, then use
the camera software to adjust the analog gains of the individual channels
to obtain a neutral gray
Whitobalancing color filter array camera. "One-shot" color cameras use an area array sensor with color filters applied over the sensor, so
they're often known as color filter array CFA) cameras. Each element in
the sensor captures a single color, usually but not always red, green, or

158

R u l World Color Mamgement, M

d Edition

blue. The color image is made by processing the raw capture--which is

grayscaleto interpolate the missing colors for each pixel.
Almost all CFAcamerasofferawhite balance feature, hut it worksquite
differently from the gray balance in scanning backs.Wlen you ser white
balance in the camera, it doesn't adjust the camera'sresponse to the light.
Instead. it affects the processing of the raw image. Most professional and
prosumer CFA cameras offer the choice ofshootingJPEGorshooling Raw
format. The impact of white balance differs between the two.
When you shoot JPEG, the raw data is converted in the camera, using
the on-camera settings for white balance (and contrast, sharpness, color
matrix, etc.),so when you profile a camera for P E G capture, it's critical
that you set a custom white balance when you shoot the profiling target.
When you shoot raw, however, the white balance setting has no effect
on the captured raw data-it's simply recorded in the image's metadata
as a tag that raw converters can use for a default interpretation. However.
since the raw datais a grayscale image that requiresprocessing to become
a color one, you can't profile the raw capture directly-rather, you're profiling the camera and a specific path through a specific raw converter. It's
still a good idea to set acustom white balance when you shoot the target.
and use thal white balance in the raw conversion, rather than relying on
either the camera's auto white balance or the raw converter's best guess
on white balance.
If your camera is unable to set either a custom white balance or a
custom gray balance, it probably isn't a candidate for profiling. Instead,
simply open the images in a well-choseneditingspace, and use your calibrated monitor as your guide for any necessary edits.

Capturing the Target

Make sure that the target is lit as evenly as possible, and try to capture
it as squarely as possible. Most profiling tools allow you to correct small
perspective distortions, but if they're more extreme you may get a bad
profile, or the software may simply reject the capture.
Ifyou're profiling raw, most raw converters offer a "linear" (gamma1.0)
capture mode, which is generallythe best choice for profiling sinceit uses
only the minimum processing required to produce a color image. In the
early days we had a notable lack of success with this approach, but with
current raw converters and camera profiling tools, we find that it works

Chapter Z Bulldins Input Plonkr

159

reasonably well. It's also the simplest path to profiling high-bit captures.
Figure 7-6 shows a linear-gamma capture of the ColorChecker SG target
opened in GretagMacbeth's F'rofileMaker Pro 5.0, with perspective corrections applied.
ngun 7-6
Cropping the target

!M

Building the Camera Profile

The process for building camera profiles is basically identical to that for
building scanner profiles. You provide the profilingtool with a target description file and a capturedTIFFimage of the target, crop the target, and
press a button. The profiling tool then calculates the profile, and you're
done.
Well, not quite. Most camera profiling tools offer a wealth of options
that, more than in any other kind of profile, blur the line between color
management and color correction. The huge number of variables that
come with the territory in camera profiling just make it an inherently
complex undertaking. For example, we've found that most cameras respond quite differently to daylight and to tungsten, so we recommend
making separate proliles for daylight and tungsten shooting.
Most camera proflers allow you to set contrast, gray balance, saturation, andevenexposure compensationin the profile itself. Gettingthehues
tight is probably the easiest part of camera profiling. The trickier aspects
concern the exposure and contrast. In our experience,it makes more sense
to build a family of profiles with different tonal handling than to try to get
everything right in one single profile. Figure 7-7 shows the options offered
by GretagMacbeths ProfileMakerPro 5 for camera profiles.

160

Real World Color Management, Second Edltlon

ngure 7-7

ProfileMaker Pro 5 offers a

wealth of options for camera
profiling.

GretagMacbeth
ProfilemakerPro 5
camera options

C."..#

111

C""

m,-.

'..,.".-

**

srrr-

G...;".".io",:"T~G;<-,i~

~~

>r a
-inc.i

The Espos~~re
Comper~satior~
tab lets
you bias the exposure for cameras whose
built-in metering consistently under- or
overexposes the image.

The Gray Balance tab, left, lets

you choose between using the
camera's gray balance, or
making the profile force
near-neutrals to gray.

, - -- .m

The Satnrntion aad Conlrosr lab lers j m r

build highlight, shadow, contrast, and
saturation tweaks into the profile.
The Spot Colors tab lets you
bias the profile to reproduce
specific measured colors
accurately, which is useful
for product shots.

Chaptu 1: Buildlng Input Proflln

161

If the sheer number of variables makes your head hurt, one interesting alternative is the approach taken by Adobe's raw converter. Adobe
Camera Raw, which contains not one but two built-in profiles for each
supported camera. We discuss Camera Raw in some detail in Chapter 12,
The Adobe Common ColorArrl~itecnrre,
and suggest a strategy for using
Camera Raw's built-in Calibrate controls to tweak the built-in profiles to
match the behavior of your specific camera in Chapter 9, Evalrcntingand
Editing Profiles.

Starting Out Right

A good input profile can be avaluable time-saver. Good scanner profiles
produce scans that reproduce exactly what's on the film, which is usually

the best starting point for an image. Camera profiles are a bit more slippery We've found that even with the best camera profiles, we'll still have
to make significant edits to tone, and when camera metamerism rears its
head, to color, too. But even in the worst-case scenarios, a good camera
profile will at least nudge theimagein the right direction, saving you time
and effort. Which is really, after all, what color management is about

This page lefi blank. ..

Building
Output
Profiles
Anal Destinations

Output profiles do more than just deliver your known, calibrated color
to your output device. They're also the map you use when you're deciding where you want to drive that color. Output profiles not only help the
color management system (CMS) produce the right numbers to represent
your color on the output device, but they also help the CMS show you, on
your display or on another printer, how the output device will render the
color before you print it. Most of our captured images contain colors that
our output devices can't reproduce, so we have to decide how we want
to handle these out-of-gamut colors: the output profile is the map that
shows you the possible destinations. If that map doesn't describe the territory accurately, you're liable to get lost, and if an output profile doesn't
describe the behavior of your output device accurately, your color won't
end up where you expected it to.
This capability to preview the results before they happen is one of the
most valuable things color management brings to the table-it lets you
use relatively inexpensivedevices like monitors and inkjet printers to predict what will happen when yousend your color to much more expensive
devices like printing presses or digital film recorders, so you can take any
necessary corrective action beforehand.

Rral World Color Management, Second Edition

But to do so, the color management system needs accurate profiles.

The process of building a profile is quite simple--you feed the profiling
software some measurements, and it munches them and spits out a
profile--but just as profile accuracy makes or breaks color management,
the accuracy of the data you collect makes or breaks the profile.
If you skipped straight to this chapter without reading Chapter 5, Measrrremenl, Calibration, and Process Control, because you figured it didn't
apply to you, Stop, Do Not Pass Go, Do Not Collect $200. If you just print
the profiling target, measure it, and let the profiling software do its thing.
you may get a decent profile. But the chances will be a lot higher if you
spend some time making sure that the device you're profiling is working properlv before you print the profiling target-that's why we spent a
whole chapter harping about calibration and linearization.
In this chapter, we talk in detail about the things you have to do before
profiling to make sure that the data you collect represents the behavior
of the device you're profiling as accurately as possible, and after profiling
to make sure that your device keeps behaving the way the profile expects
it to. Even if you don't plan to build your own profiles, relying instead
on any of the many reputable profile-buildingservices available online
or through consultants, you need to make sure that the profiling target
gets printed properly, and you need to be aware of the things that can
affect yonr device's behavior, and hence the accuracy of its profile. after
the profileis built

The Moving Target and The Rubber Ruler

As we pointed out in Chnpter 5 , h4eosftrerr1errt,Calibmtron, and Process
Control. profiles are snapshots in time. In the case of output profiles. all
that they can do is to describe how the output device was behaving when
it printed the pro6ling target.
Butmost output devices changeovertime, though different kindsdoso
in differentways. Sounlessyou like trying to measure moving targetswith
rubber rulers, you need to develop strategiesfor making sure that you collectgooddata to build the profile and for keeping the device's behavior in
sync with the profile. Inthii chapter, we'll point out the kiodsofvariabili~
that plague different types of output devices, and suggest strategies both
for collectinggood dataand for tracking thedevice's behavior so that you
know the data remains good. The primary tool for doing all these things
is the measuring instrument.

Chapler 0: Building Output Protila

165

Measuring Instruments
If you're serious about building output profiles, a dedicated measuring
instrument is essential.
In theory, you can use either a reflective colorimeter or a reflective
spectrophotometer to build output profiles. In practice, we recommend
reflective spectrophotometers because they offer the most bang for the
buck.They don't cost a wholelotmore than colorimeters, hill they're more
accurate and versatile.
Do yourself a favor and choose an instrument that's directly supported
by your profiling software. If you're a tinkerer by nature, you can probably
figure out the data format vour profilingsoftware expects-the patch order,
the measurement type, and the actual formatting of the me-and then expon the measurements to a text file that vour profiling software can read,
but it's a lot of work! (Bruce remembers using Red Ryder 10.1, a venerable
Macintosh terminal emulator, to take readings with a Gretag SPM50, and
blames it for a considerable amount of hair loss.) It's just so much simpler
to use a device that your profiling software can talk to directly.
There are profiling packages that use a flatbed scanner to measure the
printed targets, but in the final analysis, scanner-based profllers are like
the talking horse: what's amazing isn't how well they do it, bur that they
do it at all. We don't recommend them.

Measurement Geometry
Spectrophotometershave two different flavors of measuremenrgeo~netrjthe direction from which the light strikes the sample, and the direction
kom which that light is collected. Instruments with d10' or O0ld geometries are often known as "integrating sphere" or simply "sphere" instruments. They measure the color of the sample without regard to surface
texture and are most oftenused in the forn~ulationof paints, inks, plastics,
and other colored materials. In the graphic arts world, we favor instruments with a O"145" or 45"/0 geometry (the two are functionally equivalent), because they make measurements that correspond more closely to
the way the sampleappears to the human eye, taking into account surface
texture and its effect on the apparent color (see Figure 8-1).
Most sphereinstruments offer an option to "exclude the specular component," producing a measurement more like the 45"10 geometry: so if
you do use a sphere instrument, use this option if it's available.

166

Real World Color Management, Second Edition

64

ngure 8-1

light source

Measurement geometries

In a 0/45"instrument,
thesample k illuminated
by a light beam thatk
perpendicular to the
surface. The detector
reads the light thati
reflected at a 45-degree
angle.

I sample

OV45"geomeny
detector

2
f

.$

I sample

d10" geomary

In a d/O"instrument, the sample is

illuminated bya diffuse lightsource that lights thesample
euenlyfrom all angles. 7'hedetectorreads the light that's
reflected perpendicular to the su&ce.

Aperture Size
All spectrophotometersmeasure light through an opening, or aperture.

Most of the spectrophotometersused in color management have a measurement aperture of around 4 4 mm in diameter.Smaller aperture sizes

Chapter 8: Building Ontput Mia

167

are pretty rare, but some instruments have larger apertures in the 12-15
nun range. The disadvantage of larger apertures is that you have to print
targets so that the patches arelarge enough to be read individually by the
instrument, which uses more paper-a concern if you're profilingexpensive art papers or specialty photo-inkjet papers.
However, there are two situations where a larger aperture can help.
One is when you're profiling output with very low line screens, such as
billboardprintets, where a small aperture may not capturea largeenough
area to correctly determine the ratio of coverage between the halftone
dots and the paper on which thev're printed. The orher is when you're
profiling a "noisy" print process, such as uncoated watercolor paper or
canvas on an inkjet printer, where each patch contains substantial variation The larger aperture effectively averages out the "noise."

Filters
Some papers contain fluorescent brighteners that convert ult~a\ioletlight
into visible blueish light, fooling oureyes intothiinking the paper is whiter
and brighter. (Many laundry detergents use a similar technique, which is
why they're often blueish in color.)
Unlike our eyes, spearophotometers don't have white point adaptation and mn't fooled by paper brighteners, so they seethe paper as being
blueish. Profiling tools compensate by adding the complementary color,
yellowv, so the resulting output may have a yellow cast from highlights to
midtones, or sometimes even green quartertones, as the yellow ink and
bluish paper combine.
One way to deal with this problem is to use a spectrophotometer with
a W filter. Bruce sees this as the colorimetric equivalent of sticking your
fingers in your ears and yelling lalalalala-it deals with the problem by
pretending that it isn't there--but he admits that it can produce bener
results than no W filtering. The only spectrophotometerwe knowofthat
has user-switchable W filtersis the Spectrolino. Other instruments,such
as the X-Rite DTP-41, come in two versions, one with a (permanent) UV
filter and one without.
We'd prefer to see profiling tools handle UV brighteners intelligently. rather than pretending that they weren't there-GretagMacbeth's
PmfileMaker 5.0, for example, hasan option to defect and compensate for
optical brightenersbut thus far it's the exception rather than the rule.

168

Real World Color Management, Second Ediflon

Polarizing filters are sometimes useful for measuring very glossy papers, particularly when the gloss has a directional component. We've
found them more useful with dye-sublimation printers than with glossy
inkjet papers. Theyte also used in instruments designed to measure wet
ink on press. Forprocess control when you're profiling, you must let the
ink dry before you measure it!)

Backing
The IS0 standard recommendation is to measure targets on a black backing. While we hesitate to disagree with such an august body, we feel bound
to point out that measuring over black can create problems when the
stock you're measuring is thin or translucent, because the black hacking
results in artificially low luminance readings. (Unless, that is, your final
output wiU be viewed on a black backing, which is typically not the case
with thin papers.)
Our recommendation for translucent papers is to measure overa stack
of blank stock where possible, or, if you're using a strip reader (which
makes stacking paper impractical), to measure over white instead-strip
readerssuch as the DTP-41 generally give you thechoice. A white backing
will still skew the measurements slightly, but much less so than a black
onewill.

Handheld Instruments
Handheld instruments are great for making spot color measurements
you position the measurement aperture on the sample and press the
measure button. They're handy for process control, where you need to
take relatively few measurements, but most profiling tools requiwyou to
measure hundreds or even thousands of color patches.
With a handheld instrument, you position the measurement aperture
on the first patch, press themeasure button, move theaperture to thesecond patch, press the measure button, and so on. several hundred times.
Handheld instruments are usually less expensive than the automated
varieties and can produce very good results. They're great if you're on
a tight budget, have plenty of time and strong wrists, and only need to
make profiles occasionally-we've used them more often than we care to
remember-but if you're like us, you'll find about halfway through measuring your first target that you'd really likeaninstrument that offersmore
automatic data collection.

Handheld reflective spectrophotometers include the Color Sawy

ColorMouseToo! CM2S and the X-Rite Digital Swatchbook (DTP-22).
Handheld reflectivelemissive spectrophotometers, which can measure
monitors as well as hard copy, include the various incarnations of the
venerable Colortron series, which still receives some degree of support
from X-Rite, Spectrostar's Spectrocam, and GretagMacbeth's Spectrolino
and EyeOne Pro (see Figure 8-2).

b u n8 4
Handheld
spectrophotometers

Part of Bruce's collection of handheld spectrophotometers, old and new

ClockwiseFefromlefr: ColorMouseToo!CIUZS, Colomon 11,X-RireDTP-22
Digital Swatchbook, and GretagMacbethEyeOne Pro.

The Spectrocam and EyeOne Pro offer a unique measuring capability

that other handheld spectrophotometerslack. In addition to taking spot
readingslikeother handhelds,you can also use them in a"scanningVmode
by dragging them over a row of patches on an appropriately designed
target (the instmments need targets where each patch is significantly
different from its neighbors to figure out where one patch ends and
another starts). These instruments offer a very quick way of collecting
data at relatively low cost. The only real downside is that it's relatively
difficult to design your own targets, which is a fairly esoteric concern.
(GretagMacbeth's MeasureTool5.0, part of the ProlileMaker Pro 5.0 package, can produce custom targets formatted for the EyeOne Pro as well as
for various common strip-readinginstruments,but it only lets you control
the number of patches, not the content of the patches.)

170

Real Workl Color Hanagcmcnt, kcond Editlon

XT Plotters
XYplotters are the most automated reflective spectrophotometers. They
use a mechanism that moves the reading head in both dimensions across
the face of the target, so you can program them to read an entire target
automatically. XYplotters demand the least human interaction for reading patches, but many of them don't make it easy to measure a single color
or a custom set of colors. The two most commonly used XY instruments
are GretagMacbeth's Spectroscan (seeFigure 8-3) andX-Rite's Spectrofiler.
But the two are very different and are suited to different tasks.
The Spectroscan is actually made up of two parts-the handheld
Spectrolino previously mentioned and a mechanized table with a robot
arm that moves the Spectrolino from patch to patch. The Spectroscan
measures targets up to 31 by 24 cm (12.2 by 9.4 inches) on material up to
1.5 mm in thickness. It lets you specify measurement positions to within
0.25 mm and to average multiple measurements per patch, which makes
it very useful for measuring "noisf' print processes such as inkjet on rag
paper, where there may be substantial variation within each color patch.
Measurement is automatic, but quite slow. On the plus side, once you've
started the measurement process, you can simply walk away and come
back when it's done.
X-Rite's Spectrofiler is more specialized-it's designed to make
fast measurements of targets printed in the trim area of a press sheet,
either control bars for process control or specially formatted targets for

GretagMacbeth's
Spectmscan

-I

Bruce$ Spectmscan at work

Chaptn 8: Building bu(put Proflla

17l

profiling. The patches can be as small as 3.EX1.6 mm, with target widths
up to 80 inches (less-expensiveversions of the instrument handle 20-inch
and 40-inch sheels, respectively).It measures an appropriately formatted
extended IT8 target in under four minutes.
Printing the target in the trim area lets you piggybacktarget printing onto
amoney-makingpress run, so the Specmfiler is great for profilingpresses,
but it's less suited to dealing with smaller-format output devices.

Strip Readers
Strip readers are generally the fastest measurement instruments available,
like the EyeOne Pro or
although a handheldscanningspec~~ophotometer
Spectrocam can come close. They need more human involvement than
XY plotters-you need to load the strips into the i o m m e n t and keep
them properly aligned-and they're ill suited to making spot measurements for comparison or process control. But they're quick and efficient
for reading profiling targets.
The ship reader that's most widely supported by profiling tools is the
X-RiteAutoscanSpectrophotometerDTP-4 1. It's afast, efficientworkhorse
for making KC profiles and is also directly supported by some Electronics for Imaging Fiery RIPS and the Bestcolor proofing RIP as a calibration instrument. Of late, the DTP-41 has been given a serious run for its
money by GretagMacbeth's ICColor strip reader, which is fully supported
in GretagMacbeth's software and is fast garnering third-party support.

Profiling Packages
A1 profiling packages work by comparing known reference data with
measurement data. In the case of output profiles, they compare known
RGB or CMYK values (the ones in the profiling target) with the IAB values we obtain by measuring the printed target. Profiling packages run
scannerthe gamut (pun intended) from relatively inexpensive (~$100)
based packages to industrial-strength solutions with price tags to match
($2500-$5000).Generally speaking, we find that you get what you pay for,
thoughdepending on your needs, you may find that you don't always use
everything you pay for! For example, if your profilingactivities are focused
on RGB inkjet printers, you don't need sophisticated control over black

179

Real World Color Management, W n d Edition

generation and other CMYK separation parameters. Likewise, if you're
working with a handheld measuring instrument, the ability to define custom targets with thousands of patches probably isn't of great interest.
All output profilers work in basically the same way-they provide a
profiling target that you print on the device you're profiling. you measure
the target, feed the measurements to the profiler, and then it chews on
the measurement data and spits out a profile. (For details on the chewing, see the sidebar, "Profile Creation.") But within those broad confines,
you'll find plenty of differences. We can't possibly review every profiling
package in this book-it would get out of date very quickly if we did-so
instead. we'll give you a quick mndown of the kinds of features on which
different packages compete.

Instrument Support
We don't think there's been any collusion between the vendors, but they
all seem to have made a conscious decision to support as many of the
instruments incommon useas they possibly can-this is true even ofvendors like GretagMacbeth, who make instruments in addition to profiling
software.
Most proliling packages can support any instrument by importing
measurements as text files, but it's a lot more convenient if the package
actually drives the instrument directly.With handheld instruments, direct
support usually means that the target is displayed on the monitor with a
cursor showing youwhichpatch to measure--very helpful when you stop
to answer the phone halfway through a target. With automated instruments, it's perhaps less important but still a real convenience.
With instruments that require specially formatted targets, such as XRite's DTP-41 and GretagMacbeth's ICColotstrip readers, orthe handheld
scanning EyeOne From GretagMacbethand Spectrocam from Spectrostar,
the profiler may limit you to a subset of all the available targets. We can't
say weGe ever found this to be a real problem, but in some situations-for
example. if you want to measure the industrp-standard IT8.713 target to
compare your press's behavior with published standards--you may have
to do somesearching to find an IT8.713formatted for the EyeOneor Spectrocam. (You can always measureit one patch at a t m e , but that's quite
tedious. See the following two tips.)

Chaptcf 8: Bulldlng Output P r d a

'I73

Profile Creation
Jusl what do output profilers
do when they, as we rather simplistically put it, "chew on the
measurements'? In practice,
they have to do quite a lot. As
we pointed out in Chapter 4.
All Ahout Profiies, every output
profile contains six tables, one in
each direction between device
RGB or CMYK and PCS values For
each rendering intent. Building
all these tables takes quite a bit
ofwork.
Profiles generally don't use the
raw measurement data ro build
the profile. Fnr example, the
profiles we used to print this book
were made from aprofiling target
containing 875 patches. producing 875 measurements. Rut each
AtoB table contains only 17 grid
point% each one of which has

fowvalues for C. M. Y,and K and

three values for L*. a* and h*. (A
profile that contained grid points
for each measurement would get
very large, very qulckly, because
it would contain 875 points, each
of which had four values for C,
M.Y,and K, and three values for
La,a*, and b'!) So the first thing
the profiling package does is to
interpolate the measurement
data down to a manageable size.
and build the AtoBl table, which
contains the wiative colorimetric
rendering intent. Then it calculates
the AtoBO (perceptual)and AtoR2
(saturation) rendering intents.
To go in the opposite direction.
the profiling packageuses double
the number of grid points-33 in
this case-to improve accuracy.
(The profiling tool knows what

LAD colors tesult from the CMYK

values in the target, but it doesn't
have m y actual data going in
the other direction.) So it has to
interpolate the AtoB tables to
twice the number of grid points,
then reverse the tables so that
they go from LAB to device value^
to produce the BtoA tables.
That's the simplified view. In
addition. the profiling tool may
apply some smoothing ro the
data, it may build in some "secret
sauce" to work around some of
the limitations of LAB, or it may
do other things that we've yet lo
figure out. But whatever the
story, when yon see a progress
bar during the profile creation
process, rest assured that the
profilingtool is doing some pretty
hefty calculations.

Tip: Download Targets from Different VendorsYour profilingtool may

not offer a target formatted for your instrument, but chances are that
some other vendor does. Most profiling tools are available as downloads
in demo mode--you need to buy a dongle or license key to actually save
profiles, but you candownload the package anduse the targets and target
reference files contained in the package without buying anything. The
only tricky part is formatting the target reference file for your particulat
profiling tool (see the next tip).

Tip: Use a Spreadsheet to Format Target and Measurement Files. The

o n e speed bump in using both measurement files a n d target reference
files that weren't created by, or designed for. your profling tool is getting
them into the format your profiling tool expects. We find that the best
approach is to use a spreadsheet program like Microsoft Excel to refor-

174

Real World Color Management, Second Edition

mat measurement and target reference files. They're almost always tabdelimited text, so you can easily switch things around by simply moving
the columns in the spreadsheet.
Profilers almost always include sample measurement files. Use one of
these as a template to figure out the data format the profiling tool wants
(see Figure 8-4). Likewise, profiling tools almost always include target
reference files that record the RGB or CMYK values for each patch in the
target.Youneed to take the target reference file for the "foreign" target and
reformat it for your profiling tool.
ngure 8-4

Target data reference files

GretagMacbeth'starget reference
file for the IT8 713 target

Integrated Color Solutions'

target referencefile for the BasICColor Print 3c RGB target

Chapter k Bulldlng Oulput Profila

175

Targets
One of the biggest areas of difference between profiling packages is in the
profiling targets they use. Most packages support the IT8.713 target for
CMYK pmfiling, but the IT8.713 has some shortcomingsits main usefulness is as a standard for publishing reference data for print standardsso
most packages also offerone or more pmprietaq CMYK targets. In addition, the ECI 2002 target, developed by the folksat the Eumpean Color Initiative (ECD is startingto gain traction-it includesallthe patches fmm the
lT8.713 but adds additional patches to provide a more uniform sampling
of device behavior. The almost-ratified 17'817.4 is a superset of the aforementioned targets, but it's simply too new to have gained support from
proiilingtools as we write this.Thereareno standard targets for RGB output
profiling--each profiling package uses its own proprietary RGB targetts).
Besides any special requirements imposed by your instrument, the big
difference between different packages' targets is the number of patches
they require you to measure. Most packages need at minimum two to three
hundred measurements,some require closer to a thousand, and some give
you the option of measuring several thousand patches. Don't assume that
more measurementsalways means a better profile in the end-in our experience, it depends very much on the profilingsoftwareand the device being
profiled.Avery general guideline is: the closer your device is to being linear
and gray balanced, the fewer patches you need; and the more non-linear
andlor color casted the device. the more patches you need. In some cases
you can end up introducingnoise when profiling very linear devices if you
measure too many patches. But no matterwhat, y o ~ umileage will vary.
If you use a handheld inshument, you probably want to measure the
smallest number of patches possible-the difference between three hundred and eight hundred patchestranslates to alot of time and alot of stress
on yourufists! When you have to measurea large number of patches with a
handheld instrument, you also run agreater risk of mismeasurement,which
will produce a bad pmfile. With an automated instrument, the number of
patches is obviouslyless of an issue, and it's common for packages to provide feedback in the form of an error message if mismeasurementoccurs.
Some profiling tools also let you generate your own custom targets. In
the past, we felt that this wasstrictly for the hardcore color geek, but GretagMacbeth's ProfileMaker Pro 5.0 makes custom target generation relatively painless (though measuring a 10,482-patch target may be another
story!] Figure 8-5 shows targets used by a variety of profilingpackages.

176

Real World Color ManagI, k o n d EdlUon

Figure 8.5
Profiling targets

Profiling targets come in

all shapes and sizes, but
they all attempt to do
the same thing-provide the profiling
tool with a sufficiently
representative sample
of thedevice's behavior
to build an accurate
profile.

The Il8.713 CMYK target

Most profiling tools

include proprietary
CMYKtargetswith fewer
patches than the
industry-standard
IT8.713 and EC12002.
However, unless it's
impractical to measure
the larger number of
patches, we recommend
using the EU2002 targetfor most CMYK work,
because the target is
supported by almost all
profiling packages, so
the measurement data is
reusable. Measuringan
industry-standard target
also makes it easier to
compare your device's
behavior with industry
standards.

The EU2002 CMYKtarget,designed by the European Color Initiative

Chapter 8: Buikllng Output Pmflkr

Profiling targets,

177

'

continued

Both pages of
MonacoPROFILERS
1728-patch RGB target

I
1
"I

I'

?
e s m t s C l _ r * ~ a s i r s o =i
.-..-O----

Colorlrision'sProfiler Pro 729-patch

RGB test chart

GretagMacbeth's TC9.18 RGB test chart

(designed by B111Atkinson)

If you're profiling a stable but non-linear process such as an Epson

inkjet profiled on top of the driver's No Color Adjustment setting, using a
target with a very large patch count can indeed produce a better profile,
but it's a great deal of work for a relatively modest improvementdon't
even think about it unless you have an automated measurement device!

178

Real World Color Hanagcmrnt, Iacond Edition

Data Averaghg
Ifyou're planningon profilinganinherentlyvahbleprintingprocesssuch
asa pressorcolor copier, you'll almost certainly need to measure multiple
targets and average the results. Some profilingtools will do this for you.
You can always open the measurement files in a spreadsheet program
such as Microsoft Excel and average them there, but having the profiling
tool do it for you is much more convenient.

'lip: Download GntagMacbdh's kleasufetool 4.1.5. Profilerpro 4.1.5 is

still available as a demo download from GretagMacbetlisWebsite.You need
to buy a license to unlock the ProfileMaker and ProfileEditor applications.
but the MeasureTool application offers two very useful capabilities even
when the package is running as a demo--itletsyou c o r n p a measurement
files, and it lets you average measurementfiles. So if your profiling package
doesn't offer data averaging, you can always use MeasureTool to average
your measurements before feeding them to your profiler. [MeasureTool
5.0, part of the ProMeMaker 5.0 package, requires a dongle to allow comparisons and averagineit was good of GretagMacbeth to keep the older
version available.)

Linearization
Some profiling packages make profiling a two-step process,whereby you
first print and measurralinearizationtarget,like the oneshown inFigure
8-6. to determine the linearity of the printer-the degree to which changes
in the control signals produce proportional changes in the printed color.
The profiler then uses the measurements to generate a profiling target
that's optimized for the specific device.
Other packagesuse linearizationas aprofile updatingfeature--you reprint and measure the Linearization target (whichhas many fewer patches than the full-blown profiling target), and use these measurements to
update an existing profile. In some cases, you need to build the original
profile using the linearization step to be able to re-linearize and update
the profile.
Do these packages work any better than those that don't offer linearization? In extreme c a s e s f o r example, when a device starts to plug up.
producing the same color at any value higher than 70%-they can. They
work by changingthe valuesin the profiling target (and itsaccompanying

target reference file) to eliminate redundancy by lightening the patches

so that they lie inside the usable range of the printer.
Ifyou're worldngwitha CMYK RIP that supports linearization, it's usually easier to linearize the device through the RIP prior to profiling-that
way, you eliminate the need for linearizing the profiling target. However,
some RIPSdo a better job, or are easier to use, than others, so when in
doubt, experiment. Bear in mind that if you linearize partway in the RIP
and the rest of the way in the profiling tool, you're mutiplying the number
of variables you need to keep under control. You may get the best possible results that way, but at the cost of introducing more opportunities
to screw up.

Parameter Controls
Most profiling tools offer explicit control over CMYK separation parameters such as total inklimit, blackinklit, and blackgeneration, although
a few entry-level packages simply offer presets for different CMYK processes. More advanced packages offer control over black start (the point,
typicallysome considerable distance from the highlights,where blackink
is first introduced),black shape (the rate at which black ink is introduced
as the color gets darker), and black width (the strength of the Gray Component Replacement, or GCR, which dictates how far from the neutral
axis black ink is introduced into color combinations). Figure 8-7 shows
the separation parameters offered by some of the leading profiling packages.
Afewpackages let you control the trade-offs that are made in rendering out-of-gamut colors using the perceptual rendering intent, letting
you choose the compromises between maintaining hue, saturation, and
lightness. Figure 8-8 shows the rendering controls offered by Heidelberg
Printopen 4.5 and MonacoPROFILER 4.5.

180

Real World Cdor Management, krond EdHlon

nguw 8-7
Separation parameters

The controls provided

by the three packages
shown here are roughly
equivalent, even though
the terminology differs
from package to
package.

rwrwwt.lcr-me.-

om-

r ern

~ a a

mr*m-crna~*~)
lqanpl-m

rcm.r.ollmarm
I*lma

mmcura-wa

m*wA**-111

M o n a c o f >FILER offers control over black and total ink limits, blacksta. .,
black shape, and GCR strength, in addition to automatic settings.

connolo& blackand iotalink limiu: blackstart.

I'

Flgun 8-8

Rendering controls

MondROFILER lets you choose how much lightness or saturation

to sacrifice for perceptual rendering.

YeidelbergF'rintOpenlee you build custom percepnr

renderings by connollinglightness and contrast.

189

Real World Color Management, Second Edlllon

Some profilers let you decide whether to aim foraneutml gay axis that's
dead-on neutral or one that's neutral relative to the paper stock. In our
experience, very blue-white papers can cause profiling packages to overcompensate and make true neutrals appear yellow, while the paper-relative
neutrals appear much more neutral even if they don't measure that way.

Profile Editing
Some packages not only generate profiles, but also let you edit them.
Profile editing can be extremely valuable, with the important caveat that
before you start editing a profile to address a problem, you need to be
very sure that the problem does in fact lie in the profile, rather than in the
measurement data or the device behavior. See Chapter 9, Evnlrmringanrl
EditingProfiles, for an in-depth discussion of profile editing.

Getting Good Data

If you just print a profiling target, measureit. and feed the measurements
toyourprofilingsoftware, you mavbe luckyenough to get a good profile-but unless you take special care when printing the target and taking the
measurements, it3 a lot more likely that you'll get a bad one.
You want to make sure that the output device is working properly when
you print the profiling target.Then you want to capture measurements that
represent the good working state accurately Lastly, after you've built the
profile,you want to make sure that the device keeps working the way it did
when you printed the target, and hence that the profile staysvalid. Different
device types present different challenges,each of which requires a different
remedy, so we'll breakthe discussion down by types ofdevice, point out the
issues you're likely to run into with each type, and suggest remedies. So feel
free to skip from here to the device types that interest you

Opening and Printing the Target

Profiling targets are possibly Iheonly case where wedon't careatallabout
color appearance-all that mattersis thenumbersin the targetfiles.Therefore, it's absolutely vital that you do nothing to change thenumbers inthe

Chapter 8: BuildingOutput Pmflkr

183

target files. We make a practice of never, ever, assigning or embedding a

profile in aprofilingtarget, both because they're simply unapplicable,and
because it's all too easy to trigger an automatic conversion when a source
profile is present. (It's possible to trigger one even when there's no source
profile, but no worknow is absolutely bulletproof!)
We almost always print our profiliig targets from Photoshop, because
Photoshop has very explicit color management controls throughout the
process. But whatever application you use, you must make sure that the
target doesn't get converted when you open it.
You can make life simpler by also making sure that nothing happens
to the values in the target when you print it, and that the numbers in
the target file simply get sent directly to the printing device. See Chapter
11, ColorManagement in the OperatingSystem, for more information on
printer driver color management.
Sometimes you may want to deviate from this strict rule-for
example, when profiling an RGB inkjet printer, you may want to use one
of the auto-correction features in the printer driver, or when profiling a
proofer, you may want to profile on top of linearization curves, or curves
set to make the proofer simulate another printing process.
If you do anything l i e this, be aware that the profile is only valid for
that printing condition. When we make profiles in situations l i e this, we
always document exactly what settings we used, and we put some clear
indication in both the internal and external name of the profile that it's a
non-standard prolile that applies to a particular set of printer settings.

Opening the Target

When we open the profiling target, we always make sure that Photoshop's
missing profile warning is turned on, and we always choose the Leave as
is (don't color manage) option, as shown in Figure 8-9.
Fi#ure 8-9

Opening the target

leaue the numbers in the

target unchanged.

lR4

Real World Color Management, Second Edition

That way, we don't run the risk of inadvertently embedding a p r o f i l e

it's all too easy to close the file, get the prompt asking if you want to save,
then click the save button. If you assigned a profile, it gets embedded
when you save.

Printing the Target

When we print, we always set the Source Space to Document (which,if we
opened the target correctly, is always Untagged RGB or Untagged CMYK)
and set the Print Space to Same As Source, as shown in Figure 8-10. That
way, the numbers in the target are always passed along unchanged to the
printer driver.

Printing the target

Document and the Print

Space to Same As Source
to pass the unchanged
numbers to the printer
driver:

ka*:[h*.h...

.,,,I

r c ~
uu sack sDin w s a

- ~ - ~
~

mn

Profiling- RCB lnkjet Printers

As we've previously pointed out, there's really no such thing as an RGB inkjet
printer-they all use cyan, magenta, and yellow inks, most also use black,
and some also add "light" versions of the primaries to help improve highLight detail. We call them "RGB inkjets" because the vast majority require
RGB signals as their input, forcing us to profile them as RGB devices.

Ckrp(cr 8: Building Output P d l n

185

The reason all non-PostScript inkjets function as RGB devices doesn't

have anything to do with the devices themselves;it's simply that the data
pipeline from the application to the printer driver offered by QuickDraw
on the Macintosh or GDI (Graphics Device Interface) on Windows can
onlypass RGB (or grayscale)data, not CMYK. So even though some applicationswill let you send aCMYKlile to the printer, they perform a hidden
conversion back to RGB before handing off the data to the printer driver.
(The Mac OS X display architecture doesn't have this limitation-it can
pass CMYKaswell asRGB orgrayscalebut we've yet to see an OSXdriver
that exploits this capability.)
In theory, inkjets driven by Postscript R P s should function as CMYK
devices, but in practice we've found that agood many PostScriptRIPsalso
perform a hidden CMYK-to-RGBconversion. Sometimes theonly way to
find out which will give you the better results is to make both RGB and
CMYK profiles for such devices, and use the one that performs better.

Before Profiling RCB Inkjets

lnkjet printers are h~icallyvery repeatable and stable. RGB inkjets don't
offer any calibration features, though they do usually have a plethora of
softwaresettings.To get a good RGB inkjetprofile,there are threevariables
you need to worry about.

, Software settings in the driver have a huge impact on the printer's

behavior--see Figure 5-2on page 118 for some dramatic examples.
lnkjet prints take time to reach their final,stable color. Even "fast-drying"
inks, which appear to he dry when the page comes out of the printer,
continue to change color until they reach stabiliw.
Unannounced manufachuing changes in inks or paper can change
the color that the printer produces.
The first two are things you need to takeinto consideration before you
profile.The last one is simply something you need to watch out for when
you change consumables.

Media settings. Most inkjets offer settings lor different media types, generally those sold by the printer vendor. These settings control the amount
of ink laid down, the black generation, and in some cases, the screening
algorithm, so they have a huge effect on the print. If you're using vendor-

186

Real World Color Management, Second Edition

Comparing Colors
The ability to compare measuredcolors is invaluable in color
management, whether you're
tracking press variation from
sheet-to-sheet, comparing different batches of consumables.
or determining drying time. Visual comparisons can identi&
gmss problems, but subtle differences can be exaggerated by the
profiling process. So some means
of comparing measurements and
determining how different they
are is extremely useful.
Q e h e . By far the most common

way ofexpressingcolor differences

is in CIE LAB delta-e units. Other

difference,while in midtone neuh-alsadifferenceof 0.5delta-e may

be visible.
Inproductionwork,adelta-ein
the rangeof3-6 is usually considered a good commercial match.
and a delta-e ofless than 2 isconsidered unattainable due to the
many variables in print praduction. But when you're evaluating
devicedrift orcornparingprofiling
targets, you need to be a little
strmer, because small differences
in the profilingtarget can often be
amplified in the profile.
I3mce's usual method of track.
ing doice drift on uncalibratable
devices such as RGB inkjets is to
print and measure the profiling
target andthencompare the measurements with the ones from
which the profile was built. It he
finds an average delta-e of more
than 2, or a peak delta-e of more
than 6, he builds a new profile.

color-differcncing schemes exist,

and some are a bit more accurate
than LAB, but LAB is still the most
widely used. If someone reports
delta-e values without any further
qualification, it's safe to assume
that she means LUldelta-e.
In rheury, LAR is perceptually
uniform, and in theory one delta- Tools. Most spectrophotometers
e unit is the smallest difference come with software that lets you
perceivable by humans with nor- compare two LAB values and
mal color vision, but in practice determine the delta-e difference
it falls a little short of that goal. In between them. It's less common
the saturated yelluw region, for to Hnd tools that let you compare
example, wen viewers with very an entire set of measurements
discriminating color vision may and calculate the average and
be hard-pressed to see a 3 delta-e peak delta-e diffetences.

One well-kept secret is GretagMacbeth's MeasureTool 4.1.5,

which is one of the modules in
their ProfileMaker Pro 4.1.5 software. You can still download
ProfileMaker 4.1.5 from GretagMacbethb Web site (www.gretagmacbeth.com)-it doesn't let )vu
save profiles unless you purchase
a license, but the MeasureTool application is fully functional without a License, unlike the newer
MeasureTool 5.0 in ProfileMaker
5.0. For this situation, it allows
vou to compare two sets of measurements of the same target and
shows you the average and peak
delta-e values (It even shows you
which patches have the highest
delta-e valuessee Figure 6-13.)
Perhaps the biggest omission
bom the iirst edition was our failure to mention ColorThink, lrom
www.chromix.com. ColorThink
offers more ways to visualizecolor
than any other tool we've seen, m
both2Dand30, weusedit to make
the 3D gamut plots in this book,
andwe find it indispensible.
Occasionally, you'll nm into
measurement filesthat your tool of
choice can't read. Open them in a
spreadsheet application,and then,
mpy the measurements into a file
that you h o w the tool mn m d .

supplied paper, simply choose that paper type. If you're using third-party
papers. you'll need to experiment to find out which settings produce the
best gamut and the most linear tone scale (often, you'll need to w m p m mise one for the sake of the other]. See Figure 8-11.

ngllre 8-n
Media settings

The paper settings

in most inkjet
printer drivers
control the ink
limits and black
generation.Finding
the correct settings
for your paper is
key togem'ng good
profiles.

A target like the one shown in Figure 8-12 is useful for all sorts of tasks,
including this one: the ramps let you judge the tonal behavior, while the
solid colors let you evaluate the gamut.

This target contains 21 steps of red, p e n , blue, cyan, magenta,

yellow, and RGBgray (R=G=B).

Don't expect perfection. Our experience has been that if you achieve
the maximum gamut, you'll have a fairly non-linear tone scale with some
blocking of the shadows. It's often preferable to sacrifice some linearity
in favor of a larger gamut-a prolile can compensate for non-linearity to
some extent, but it can't do anything about the gamut size. Watch out for
puddling, bronzing, or ink bleeding, all of which are signs that the printer
is laying down too much ink for the paper to handle. If you encounter any
of these problems, switch to a media setting that lays down less ink.

188

Real World Color Management, Second Edition

Color settings. Most inkjet drivers offer a variety of color-handling options. Epson inkjets, for example, offer options with names lievivid and
Photorealistic, along with some slider controls and a gamma setting. Fortunately, they also offer an option labeled, "No Color Adjustment."
Wherever possible. look for anraw"setting thatjust sends the numbers
unchanged to the printer driver, like the aforementioned No Color Adjustment. It's the simplestway to profile. In some cases, though, you may find
that you get better linearity using some other setting.
Once you determine thesettings that yield the best compromise, you
can profile this condition. with the important caveat that when you use
the profile, you'll always have lo have the software settings set the same
way you did when you made the profile. This also means that you can't
use the profileinside the printer driver, because the place you choose the
driver-level color settings is the same place you'd choose the profile--you
can only have one or the other, but not both. This isn't a problem when
you're dealingwith color-managedapplications, butit may he if you need
to print from a non-color-managed application such as Powerpoint.
Resolution settings. Most of the inkjets we'vetried seem to be fairly consistent between different resolution settings, but there's no guarantee that
yourinkjet will work that way
We find that a single profile works for all supported resolutions on a
given paper stock, but if that doesn't seem to be wor!dngforvou, you can
do a quick reality check by printing the RGB process control target shown
in Figure 8-9at the different resolutions, and then measuring each one
and comparing the results. See thesidebar, Tompaling Colors,"earlier in
this chapter. If the average delta-e in LAB between different resolutions
is greater than 1, or if the peak delta-e is greater than 6, you'll almost certainly want to build separate profiles for the different resolutions.
Curing time. One of the biggest gotchas that people overlook when
profiling is that inkjet prints need time to cure. The ink may appear dry
when the paper leaves the printer-itwon't smear or rub off-but it almost
certainly hasn't reached its final color appearance because the solvents
take time to evaporate and the ink takes time to react with the paper coating. Every inkjetwe've everused takessome time forthe color to stabilize.
(The actual curing time depends on the specific inkset and paper.) If you

Chapter 8: Bulldlng Oulpul P r d l a

189

start measuring the profiling target as soon it leaves the printer, you'll go
crazy trying to figure out what's wrong with the resulting profile because
you really are measuring a moving target.
The easiest way to determine the length of time your prints need to
attain final color is to keep measuring the profiling target with an automated measuringinstrument until you see little or no difference between
measurement passes. It's time-consuming,but you only need to do it once
for each paper and ink combination. It's the combination of paper and
ink that is the issue. The same paper will have different curing times with
different brands (and sometimes models) of printer that use different
inksets, and the same inkset may have radically different curing times on
different papers (see Figure 8-13).
ngum 8 4

Comparing
measurements

GretagMacbeth's
MeasureTool is invaluable
for mmparingsets of
measurements. Note that
in this w e , measuring the
entire target reveals drift
in the greens that would
remain undetected ifyou
simply measured the
primaries.

If you're working with a handheld instrument, though, you probably

don'twant to torture yourselfmeasuringhundreds or thousands ofpatches over and over again. Instead, you can use the RGB Process Control
target shown in Figure 8-12. Print the target, note the time it was printed-preferably on the target itself-and set it aside for a couple of hours.
Then print the target again, notingthe time the new one was printed, and
make a visual comparison of the two prints.

190

Real World Color Managemmi, a n d Edlllon

If you can see visual differences, you know the print takes at least two
hours to cure. Re-examine them periodically until they appear identical,
and note howlongit takesforthesecondprint tolookidentical to the first
me.This gives you a rough estimate of the drying time.
Youneed to refine thisrough estimatewith a fewwell-chosen measurements. Subtledifferences between targets that aren't obvious to the naked
eye can be amplified during the profile calculation process, so measure
nr leastthe solid and midtone patches for R, G. B, C. M, l', and RGB. When
the prints have cured, you shouldn't see any differences greater than 0.5
delta-e LRB between targets. See the sidebar, "Comparing Colors." earlier
in this chapter.

Creating RCB Inkjet Profiles

Profiling tools usuallv offer few or, sometimes, no parameters to set for
RGB output profiles. A good rule of thumb is to start by building a profile
at the default settings offered by the profiling tool and depart from them
only when you've found a reason to do so.
Some profiling packages offer a two-step process in which you f i r s
print and measure a linearization chart. The profiler then uses the linearization measurements to generate a custom profiling target tailored
to the specific printer at hand Read the accompanying documentation
caretidly: some packages, such as Integrated Color Solutions' BaslCColor
Print3c. recommend using thelinearizationstep for inkjet printers, while
others, such as Monaco Systems' MonacoPROOFand MonacoPROFILER,
recommend against doing so.
Some packages,including MonacoPROOE MonacoPROFILERand GretagMacbeth'sProfileMaker, offer a choice ofgray rendering aseither absolute or relative to the paper. We've noticed that with some veryblue-white
paper stocks, neutrals that measuredead-on neutral appearyellow to the
eve because our eyes alwaysadapt to the paper white. In those situations,
rendering gray relative to the paper seems to work better. In the case of
ProfileMaker, Preserve Gray Axis will produce neutrals that measure neutral yet look yellow, whereas the Paper Gray Axis will produce neutrals
that visually appear neutral. These options apply only to the perceptual
rendering intent. Other intents are always built using Paper Gray Axis.
Afew packages offer somecontrol over gamutmapping for theperceptual rendering intent. There's no right way to do perceptual rendering: if

Chapter 8: Bulldlng Output Ploffla

191

you're of a mind toexperiment, by all meansdo so, but we suggest starting

out with the default.
When you build the profile, it's always a good idea to give it a name
that's as informative as possible--one that contains the printer model
and the paper stock.

Profiling
- True RGB Printers
\Ye use the term rrlre RGB printers to refer to devices such as the Fuji

Pictrography,Cymbolics Lightlet, Durst Lambdaseries, and anything else

that images directlyonto photosensitive paper or film.These types of devices usually have robust built-in calibration features that make them
highly repeatable. The photographic media they use tends to be very
stable with no curing time.
The only real source of variability we find in these devices is that the
profiling target may print slightly differently depending on its orientation
Colorgeeks refer to t h i s p r o b l e m a s n n & o m pobfuscatoryway
~
ofsayingthat theprinterproducesslightlydifferent colorwhen printing the same
Image in portrait and landscapeorientations. In our experience, the effect
of anisotmpy is usually quite subtle, and unless you're picky, you may not
even notice the effect it has on your images printed through the pmlile.

Before Profiling True RGB Printen

If you do notice anisotropy in the printing of the profiling target. we offer
two ways to work around the problem-scrambling rhe patches on the
target, or printing the target four times, rotated 90 degrees each time, and
averaging the results. If your profiling package either offers to scramble
the patches for you [MonacoPROFILER will do this). or uses a target with
a randomized set of patches (such as the ones offered by Integrated Color
Solutions' BasICColor Print3c), use the randomized target.

Tip: Use GretagMacbeth EyeOne Targch. Many profiling packages that

don't offer random patch arrangements have targets formatted for the
EyeOne. Since the Eveone's scanning mode requires big differences between each patch, the EyeOne targets are, in effect, randomized, which
makes them ideal for dealing with anisotropy. even if you're using some
other instrument to measure them.

Real World Color Ma~gemcnt,Second Edltlon

Randomizing the patches yoursell might seem like less work than
printing and measuring four targets. Trust us--it's not!You not only have
to move the patches in the target file itself; you also have to rearrange
the numbers in the target description file that the profder uses to relate
the measured values to the control signals that produced them, and the
chances are extremely high that you'll screw something up in the process.
Just bite the bullet and print the target four times at different rotations!

Creating True RGB Printer Profiles

Theoptionsoffered by- profilingpackagesforhue RGB prinlersare the same
as for RGB inkjets, but they ma); not always produce the same results.
Of those packages that offer linearization, the ones that don't recommend it for RGB inkjets do recommend it for true RGB printers, and vice
versa. Rendering gray relative to the paper tends to have a more pronounced effect on these types of printers than on inkjeteifit's available,
we recommend t ~ n itgas your first choice.

Profiling Three-Color Dye-Sub Printers

Non-Postscript dye-sublimation printers function as RGB devices even
though they use CMY d);es. The main variables that plague dye-sublimation
printers are anisotropy (though for quite different reasons than hue RGB
printers) and manufacturing variation in consumables.
Dye-sublimation printers vary in their color performance due to thermal latency: theymakecolorhy heating thedyesin the donor ribbon until
they sublimate--turn into a gas--and get absorbed by the coating on the
paper. The laws of physics limit the speed with which the print head can
heat up and cool down, so the color can v q f depending on what color
was printed immediately before the current one.
Aswith true RGB printers. the remedies areeitherscramblingthe patches or printing the target at multipleorientations and then averagingthe results. We've generally had better results from printingmultiple targets than
we've hadfrom patch-scrambling, but it means more measurement-so if
you're workingwithahandheld instrument, you may want touse scrambled
patches on a single target to avoid excessive wear and tear on your wrists.
Some dye-sub printers also exhibit directional nwtamerism, where the
color appearance (and the measurement) changes when you rotate the

Chapta 8: Building Output Proflla

193

sample 90 degrees. Ifyou average four targets at different orientations, the

differencegetsaveraged out. If, however,you're measuringa singletarget,
a polarization filter can help (if your spectrophotometer has one).
Some dye-sub printers offer software settings that perform some kind
of auto-enhancement on incoming images. Make sure that any such options are turned off before you print the profiling target.
If your profiling package offers a linearization step, it's likely that your
dye-sub printer will benefit from it. Other than that. the default settings
for your profiler should work well.

Profiling Composite CMYK Printers

-

CMYK printers differ enormously in dynamic range, gamut, available paper stocks. and inksets. In this section, we'll talkabout color laser printers.
color copiers, solidink/waxprinters, inkjets. and digltd presses. (Printing
presses have their own section later in this chapter.) In addition. we'll assume the device is driven by a Postscript RIP
As withRGBoutput devices, you first need to make sure that the printer
is stable, linearized. and gray-balanced. The different printing technologies are prone to different sources of variability, so we'll deal with each
class's typical problems and the strategies for dealing with them, separately.
Once you've stabilized the printing process (we provide guidelines for
doing so below under each printer type), printed the target or targets,
and collected the measurements, you're ready to build the profile. Most
profding packages offer default settings for the different types of devices.
and these are usually the best place to start. The most common options
are total ink limit andUCR1GCRsettings.More sophisticated packagesallow finer controlover black generation, includingblackstart. the shapeof
the blackcurve, and the amount of GCR. See the sidebars. "Ink Limiting"
and "Black Generation." for a discussion of these terms.
In case your package doesn't provide presets for categories of CMYK
printers, we've provided starting points for common printer types. These
are guidelines, not rules. If you have ink bleeding all over the page and
you're using 400% ink, lower the total ink limit! Likewise, if we suggest a
medium blackgeneration, but a lighter or heavier amount gives you better results, go with the settings that give you betterresults.

194

Real World Color Management, Second Edition

Before Profiling CMYK Color Laser Printers

Color laser printers, color copiers, and some digital presses form a class
of devices generically called eltermphotostaticdevices. In these devices.
a laser writes the image onto a light-sensitive drum that converts light
into a static charge. The drum thenrotates, transferring this invisible image made of patterns of positive and negative charge onto the paper. The
paper passes by a toner cartridge containing the first colorant in the form
of very small and charged particles, which attach themselves to the oppositely charged areas of the paper. Finally, the paper passes through a
heated fusion roller that meltsand fuses the toneronto the paper. In older
color lasers the process is done separately for each colorant. and in the
latest generation all four colorants are applied in a single pass.
We explain this technology because it helps you understand its inherently variable nature. The variability comes from two sources:
b

Paper needs some humidity to hold a static charge--it can't be completely dry. Variations in humidity translate directly to color variation.

When a tonercartridge isnew. therearelotsofeager toner particles that

are itching to jump on charged paper. Over time, the eager toner particles are gone, leaving behind the less eager toner particles, so there's a
steady decrease in the densiiyyou can achieve with each colorant.

If you used up the exact same amount of each colorant (extremely

unlikely), you'd notice a steady decrease in saturation. More likely you
consume the colorants at different rates, so the relationship in density
among the four colorants is constantly changing.
Electrophotostatic printers are a prime example of the moving target.
The onlysolution is frequent calibration-twiceaday is typical, and once
an hour is not unheard of.The newer generations ofmachines tend to be
more stable than the older ones. So first determine how often you need
to calibrate to keep the device reasonably stable--periodically printing
and measuringa CMMC process control target like the one in Figure 8-14
is the easiest way to do so.
We offer two strategies for printing the profiling target:
b

Cdibrate the printer and immediately print the pmliling target.

Chapter 8: Bulldlng Output Proliln

195

CMYK process control

target

This taiget contains21 -step mmps of C+M+KM+K C+K C+M,

GM,KandK.
b

Calibrate the printer, print a series of targets (decidehow many you're

prepared to measure) throughout the previously determined calibration interval, and average the results.

The first approach will produce aprofile that's accurate for the device's
optimal behavior. The second will give you a profile that's accurate for
the device's average behavior. If you can enforce regular calibration, use
the first approach If you can't guarantee that the device will always be
calibrated when necessary, use the second.

Creating CMYK Color Laser Profiles

Color laserprinters and copiers almost always have fixed built-ininkliiking, so you don't want to limit it any further. These devices are usually
designed to print business graphics, so the colorants arenk particularly
well gray balanced.You'U, usually get the best gray balance by forcing as
much gray component as possible to print with blackink by selecting a
maximum GCR separation.Here are our suggestedseparation parameters
for color laser printers.
b

Total Ink L i t : 400%

Black Ink Limit: 100%

Black Generation or GCR: Maximum setting

196

Real World Color Managcmenl, keond Edition

Ink Limiting
We don't know of any CMYK
printing processes that can use
100% coverage of each of the four
colorants. If you apply 400% ink
to paper, the paper is likely to
disintegrate, whether the ink is
being applied by an inkjet printer
or a printing press.
But even when the colorants
aren't wet, as in the case of color
laser, dye-sublimation, or solidink printers, maximum density
is achieved long before you reach
400% coverage, and in some
cases density may even decrease
as you add more colorants. So
some kind of ink limiting is
always needed in CMYKprinting.

With composite CMYI( printers, you often have the choice

of limiting the inks in the profile
or in the RIP. If the printer is a
type that can take different paper stocks with different ink requirements, such as an inkjet, it's
usually easier to set the limits for
each paper type in the RIP: and
then set the profile ink limits to
400%.

If the printer is either limited

to asingie paper [dye-sub).orthe

paper has almost no efiect on the
point at which the colorants reach
maximum density lcolor laser.
solid ink), it doesn't matter where
you set the ink limits.

In either case, though, it's

critical that you set them in one
place or the other, hut not both.
For presses, it makes a great deal
more sense to set the ink limits in
the profile.
There's never a reason to lay
down more colorant than isneeded to achieve maximum density.
There is, however, often a good
reason to lay down less.
You may wish to sacrifice some
density in favor of stability, but
an even more compelling reason
is simply to save money by using
less ink.

If prints made through the resulting profile seem oversaturated or

plugged in the shadows, it's possible that the printer doesn't have builtin ink limiting (the Xerox DocuColor series is one example). In that case,
we suggest the following as a starting point:
b

Total Ink Limit: 260%

Black Ink Limit: 100%

Black Generation or GCR: Maximum semhg

Black Startionset: 5%-10%

Before Profiling CMYK Inkjets

Most of the things that apply to profiling RGB inkjet printers applv to
profiling CMYK inkjet printers as well. The main variable, apart from the
software settings in the RTP, is the time the ink needs to stabilize o n the
paper (see "Curing time." earlier in this chapter).

Black Generation
Composite CMYK printers usually need a fairly specific black
generation scheme, but on presses, the way you introduce black
ink to the cyan, magenta, and
yellow primaries involves a series of trade-offs that represent
one of the most critical aspens of
printing. The main reason we use
black ink is to produce a denser
black than we could obtain with
CMY primaries on their own, but
important secondary reasons
include saving money-black
ink is generally cheaper than the
colored inks-and avoidingpmblems on press caused bymisregistration or fluctuations in density
The two basic black generation strategies are Under Color
Removal IUCR), and Gray Component Replacement [GCRI. UCR
separations use black only in the
neutral and near-neutral areas.
while CCR Is a more aggressive
strategythat replaces the amount
of CMY that would produce a
neutml with K, even in colors that
arequite a long way fmmneutral.

Each has its own strengrhs and

weaknesses.
UCR separations are often used on newsprint presses
because they're generally less
vulnerable to changes in black
density-newsprint press operators tend to run up the black
until the type looks nice and
dense, andalittlemore blackthan
you'd bargained for can turn a
CCRseparation to mud. But UCR
separations are also used on highend sheet-fed presses to print
content that has important dark,
saturatedcolon such as shoeand
leather catalogs.
GCRseparationssave money by
uangmore blackinkrelative to the
colors, and they cau mask slight
variations in registration between
the colors that would wreck UCR
separations. GCRseparationsalso
make it easierto maintain gray balance, since the black plate carries
most of the neutrals and are often
used for images whose important details are neutral or close to
neutral.

So one key point to make is

that Ule choice of UCR or GCR
is dependent not only on the
printing process, but also on the
image content. When we profile
presses, we generally create a
family of profiles that differ only
in blackgeneration so that we can
lailor the separatin~nto the content. Computer screen grabs. for
example, generally benefit from a
much heavier GCR than we would
consider using on scanned photographic images. There's no right
answerwhen it mmes to blackgcneration on press so don't be afraid
to experiment
Most of the work in profiling
a press is in collecting the data.
Once you've done that, it's trivial
to generate profiles with different black generation schemes.
You can tell a lot about a profile's
behavior without ever committing ink to paper, so don't think
that experimenting necessarily
involvashugeprint bills. SeeMOutputPmfiles" in ChapterS.Etnrltm~ing and Editing Profiles.

If your RIP allows it, it's usually easier and more convenient to set ink
limits and linearization in the RIP before profiling than it is to apply ink
limits and linearization in the profile. You won't get the best possible results by applying ink h i t s and linearization in both; all you're doing is
multiplying the number of variables you have to contend with. Pick one
or the othel:

198

Real World Color Management, Second Edition

Some RIPSapply ink limiting but don't let you conrrol it other than by
selecting different paper types. If you want to use a paper that isn't supported directly bythe RIP, you'll need to experiment, but you should begin
by building a profile that assumes the ink limits have been set in the RIP,
We rarely find a reason to print multiple targets, though if the inkjet
printer is very low resolution (such as a billboard printer) or is very noisy
(such as when using uncoated watercolor paper), you may benefit from
taking multiple measurements and averaging them-which, with a strip
reader, may mean printing and reading two targets.

Creating CMYK Inkjet Profiles

Due to the wide array of inkjet printers, paper stocks, and RIP capabililies, the recommendations we give here are guidelines for a reasonable
starting point, not hard and fast rules. Coated papers can generally handle
more ink than uncoated ones. If your RIP applies ink limiting (whether
you can control it or not),we suggest the following as a starting point:
b

Total ink Limit: 400%

Black Ink Limit: 100%

r Black Generation: Medium to Heavy (not quite Maximum)

F

Black StartIOnset: 30%

Ifyour RIP doesn't apply inklimits, we suggest as a starting point:

Total Ink Limit: 260%

Black InkLimit: 100%

r Black Generation: Heavy (not quite Maximum)

b

Black StartlOnset: 30%

Before Profiling Solid Ink Printers

Solid ink printers (aka crflvoln-jetor crrqlon-spiners) aren't popular (perhaps because people don't realize how inconsistent color laser printers
can he and aren't aware of solid ink as a viable option).Their main drawback is that the prints are quite delicate--an eraser removes the ink from
the paper. heat melts the ink, and sunlight can drastically alter its color.

Chapter 9: Building Output Profilm

199

Solid ink prinrers work by maintaining a heated reservoir of melted

inkThe liquid inkis squirted onto a rotating drum and isthen cold-fused
with pressure onto the paper. They're very versatile when it comes to paper t y p e w e ' v e wen seen one print on a tortillaand the color is veq
consistent between different substrates as long as they're white.
Older models ofthese printers had to be shut down each day, because
theliquified inkcooked over timeand changed color. Newerprinters don't
have this problem.
Even though these printers have four inks, we usually profile them
either as CMY, o t as CMYK with no blackgeneration. We've never found a
reason to average multiple targets.

Creatlng CMYK Solid Ink Printer Profiles

You may need to experiment for the best results. But in the case of the
XeroxPhaser8200,here are the recommended settings.This effectivelybuilds
a CMYKpmfile that generates separations with no black channel--apparently the black channel is best computed by the printer itself:
b

Total Ink Limit: 300%

Black Ink Limit: 0%

Black Generation: None

Black StartlOnsel: Not applicable

If you end up with disgustingresults, fall back to the suggested settings

for the color laser or copier.

Before Profiling CMYK Dye-Sublimation Printers

Cfvll'K dye-sub printers operate identically to CblY dye-sub printers

except that they have Postscript RIPS.Whether they have three or four
colorants, they're still considered CMYK devices. But since they operate
identically to non-Postscript dye-sub printers, the same considerations
before profiling apply, including printing and averaging of multiple targets in multiple orientations.
Most dye-sub RIPS have fixed, built-in ink limiting. If you should encounter the very rare one that doesn't, don't make the mistake of thinking
that dye-sub can handle 400% coverage--you'll typically get less density

200

Real World Color Managemenl, Second Willon

at 400% than at 320% or so, because the extra heat required to lay down
400% coverage sublimates some of the dye out of thepaper and back into
the donor ribbon!

Creating CMYK DyedublimationPrinter Profiles

If your dye-sub has built-in ink limiting, use the follor5ng settings:
b

Tolal Ink Limit: 400%

Black Ink Limit: 100%

Black Generation: UCR

Black StartlOnset: 60%
In the rare case that the dye-sub doesn't have built-in limiting, try

Total Ink Limit: 320%

Black Ink Limit: 100%
Black Generation: UCR

Black StartIOnset:60%

Printing Presses
Rr i t lirhography, flexn-ograpliy,or gravure, the printing press isThe Beast

when it comes to printers. On a desktop CMYK printer, you worry about

one sheet at a time. On a press, you worry about five thousand sheets at a
time. Because mistakes are extremely costly in terms of time and money,
you need to take into account some special considerations prior to going
down the road of profiling a press.
With other devices, you typically calibrate them to optimal behaviot
and then profile them. With presses, that's just one option.You may decide
on standardized press behavior rather than optimized press behavior.
Or you may decide not to profile the press at all (see the sidebar, "Don't
Profile the Press?!").
The key factor in deciding how you want your press to behave, and
whether or not to profile it, should be your proofing methodology. 771e

Chapter 8: Building Output Pmflkc

401

press must beable to match theproof: If it doesn't, you're in the situation

we colorgeeks technically term hosed, and you must either do whatever's
necessary to force the press to match your proofs or change your proofing
system so that it does a better job of predicting actual press behavior (see
"Printing and Proofing" in Chapter 10, Color-ManagementWorkfloru).
There are essentially two approaches to running presses.
b

Do whatever's necessary in prepress to massage the data so that i t

produces the desired result on press.

Treat the press as a $40001hr, color correction station by adjusting the

press to produce a sellable result from questionable separations.

In thelatter case, you may as well forget about profiling the press unless
you'rewilling to spend agood deal of time findinga stable and reasonably
optimal set of press conditions, and a somewhat larger amount of time
retraining the press operators to hit that condition consistently instead
of making artistic decisions based on the current job. You may. however,
find it worthwhileto profile youf proofing system (see the sidebar, "Don't
Profile thePress?!." on the next page).
In the former situation, profiling the press is eminently practical, but
even then you may decide that building a profile, or a family of profiles.
forevery paperstockis simply too big a headache. Ifyour current proofing
system takes the same CM?X data as your p r e s s t h a t is, it doesn't rely on
a color conversion between press space and proofer space--and youVre
confident that you can matchvour proofs on press, itkeasierandcheaper
to profile the proofer instead.
If you aren't happy with your proofing system, or it's one that uses
profiles to convert from press to proofer space, you'll need to profile the
press, but vou may only need lo make one press profile (see the sidebar,
"Profile the Press Once." later in this chapter).

Before Profiling Presses

The tirst consideration is to get the press behaving in as stable a manner
as possible-which means v e y different things for a brand-new sheetfed press and a 50-year-old web. This isn't a book about press process
control-there are many good ones-so we won't tell you how to do that:
we simply point out the necessity of doing so.

202

Real World Color Management, kcond Editlon

Don't Profile the Press?!

Given the variation inherent in
printing on a press. as well as
the cost, you may want to consider nor profiling the press.
That's right, you read correctly.
Consider profiling your contract proofer instead. If you can
successfully match your proof on
press, and with reasonable ease, it
makes sense to make the proofer.
rather than the press, your aimpoint. Itk a lot easier toget a perfectproofthanitis to get aperfect

press run that hits your in-house

process control target aimpoints
dead center.
Traditional contract proofers, such as those that produce
film-based laminate proofs, use
colorants that are similar to press
inks on substrates that make
the colorants produce similar
dot gains to the press. The great
workflow benefit is that you can
send exactly the same separatinns to the proofer and to the

press. A few digital proofers.

such the Kodak Approval and
the Iris inkjet series. offer similar
capabilities.
You can create a profile for
a contract proofing device hy
measuring a single target, so it's
relatively quick and easy. When
you create the profile, simply use
the ink limits and black generation you want on press. That way
the separations resulting from
this profile are press-read!!

A second consideration is whether you want to optimize your press

to take full advantage of its capabilities, or whether you want to make
your press conform to some kind of standard or reference behavior such
as SWOPlTROOl for magazine publication, the nascent sheetfed TR004
standard from GRACol, IS0 12647-3 for newsprint, or a contract proofing
device. See the sidebars "Optimized or Standardized Press Behavior," and
"Intentionally Non-Linear," later in this chapter.
Printinga profiling target on press demands asignificant commitment
oftime and money. Moreover,themeasurement data you collect from the
target run will likely be reused, because building the best possible press
profile is almost always an iterative process. So no matter whether you
opt for an optimized or standardized press condition. you should treat
the target run as a critical job.

Choosing profiling targets. Most profiling packages offer proprietary targets as well as IT8.713, largely due to the latter's shortcomings. However,
a couple of new standard targets are worth considering:
The 1,485-patch ECI 2002 target from the European Color Initiative
performs better than the lT8.713.
The brand-new IT817.4, which contains 1617 patches. is effectlvely a
superset of the IT8.713 and the ECI 2002. At the time of writing, the

Chaptat 8: Building Outpul M n

PO3

Profile the Press Once

What if you aren't happy with presslinklpaper combination as
your current proofing situation? an aimpoint for proofing.
Once you've profiled the press
If you print to more than a hand~LII
of paper stocks, it's probably and established your aimpoinl.
impractical to createand manage you can profile your proofing
press profiles for each paper. One devlce, use a color sewer to
viable strategy i s to take a paper make the necessary conversions
stock with middle-of-the-road from your press profile to your
behavior on press and profilethat proofer profile, and account for

varying dot gains on different

paper stocks by applying transfer curves in the platesetter or
imagesetter. The single press
profile then heconles the
aimpoint for RGB-to-CMYK
conversionsand also serves as the
source profile for proofing clientsupplied CMYK.

IT817.4was simply too newto be supported by any profiling tools, but

we expect that to change during this book's shelf-life.
Both targets come in randomized versions. GretagMacbeth already
offers theECI 2002 as their preferred target for profiling presses. We recommend using a randomized target if at all possible (such as the version
GretagMacbeth offers for the EyeOne Pro). Otherwise, gang copies of the
target in four different orientations and expect to do extra measuring.
When you print a run of targets, you'll have hundreds if no1 thousands
of press sheets from which to choose. We recommend selecting between
10 and 20 sheets to measure and average. There are two equally valid
schools of thought on which sheets you want to measure:
Measure targets representing the full range of the press output Don't
measure 10 press sheets in the middle of the stack-get a sample of
sheets representing the natural oscillation of press behavior, with
some really good press sheets, some avenge press sheets, and some
below-average press sheets that nevertheless wouldn't be rejected.
b

Measure targets representing the sweet spot of press output. Sample

only those press sheets that represent the ideal press behavior you
wish i t always had from the beginning to the end of a press run.

In the first case, the intent is to ensure the proNe represents the overall
average sellable product from the press. Use this technique when press
process control and consistency are acceptable, but not outstandingvou don't have a wholelot of near-perfect sheets or below-average sheets,
and most of them are average.

204

Real World Color Managemenl, ktond Edltlon

Optimized or Standardized
Press Behavior
It's possible to have a press condition that's both optimized and
standardized, but itas easier to
understand if you consider them
as separate goals. A fullyoptimized
press is all about maximizing its
capability-the lowest possible
dot gain, highest ink densities,
and best contrast the individual
press can achieve, without regard
to any enernal specifications or
standards. If your press can do
better than a specification, you
let it do so.
The problem is that this
creates a unique press condition
for which there's no standard
means of creating separations
or proofs. But if you profile the
press, your applications can do
the necessary conversions to

create separations tailored to

your unique and optimized press
behavior.
Another approach is to make
tbe press conform to some khd
of reference. This may involve
something relatively easy such
as changing ink densities on
press, or creating custom transfer curves in your RIP, imagesetter or platesetter (see the sidebar
"Intentionally Non-Linear"). Or
it may he more involved, such
as changing waterlink ratio,
temperature, blanket pressure,
ink viscosity, ink tack, and dozens
of other variables, to essentially
force the press into the desired
behavior. The desired bebavior could be a house-standard
proofing system that you want all

your presses to match. It could be

SWOPiTR 001 to take advantage
of file portability,as it is one of the
most common targets for color
separations in the US.,or FOGRA
in Europe.
With a fully optimized press.
you'll need to make a profile for
your press because the press behavior is unique, and there's no
other way to make separations
or proofs. In the second case, you
can use good process control lo
make the press conform to the
desired behavior, and you can
use either standardized characterization data, such as that provided
by SWOP TROO1.FOGRA, or IS0
12647-3, or profile your in-house
contract proofingdeviceinstead of
profiling the press.

In the second case. the intent is to make sure that average o r belowaverage press sheets don't add noise to the profile. Use this technique
when press process control and consistency are excellent. Most press
sheets will hit the intended aimpoints; therefore, you want the profile to
expect those aimpoints.

Creating Profiles for Presses

While most packages supply default ink limit and black generation settings for different press types, consider them starting points--each combination of press, ink, and paper has its own requirements.
The ink limits for a particular press are usually known [and if they
aren't, they should be determined as part of the exercise of stabilizing
the press), but black generation is more of an open question. Enhylevel
packages may simply offerpreset UCR and three or four strengths of GCR

Intentionally Non-Linear
L~nearizingthe Imagesetter or
platesetter should be a regular
event in a print shop, because
these devices drift so that a 50%
dot in an image file ends up creating something other than a 50%
dot on filmor on plate. Commonly, we linearize imagesetters and
platesettersand thencompensate
for dot gain produced on press in
our reparations. That's one purpose for profiling a press or contract proofing system.
Rot there's a legitimate case to
he made tor intentionally nonlinearizing your imagesetter or
platesetter in order to force one
or more presses to conform to
a specific behavior. In fact, it's
common in the flexography and
print-packaging industries tn
compensate for large amounts
of dot gain in the image- and
platesetters, though it seems
to he a relatively recent idea in
lithography.Using transfercunres,
you can compensate for the more
subtle variations in the behavior

among presses Without having

to make custom profiles for each
one of them.
In an organization with 10
presses, it's not realistic to make
LO profiles: quite apart from anything else, you can't delay the
prepress department unt~lthey
know which press will run a job.
Instead, use transfer curves to
compensate for differencesin dot
gain, giving all presses the same
tone response curve. Then use a
single profile for all presses.
This is especially important
in C 7 P (Computer-To-Plate)
workflows, where a linear platesetter produces a sharper dot
on press than a plate made fmm
film. For these environments
to take advantage of file portability and standard separation
methods (see the sidebar "Optimized or Standardized Press
Behavior"), they need lo use
the transfer curve function to
effectively "add-in" dot gain.That
way, these workflows have the

benefits of l;TP. while still being

able to produce plates with the
same dot size as plates made with
film for those jobs that need it.
You can even use this method
to compensate for the difference
in dot gain from different line
screen settings. If all other things
are equal, the higher the line
screen. the higher the dot gain. If
you regularly print at 150 and 175
Ipi with a particular inkset and
substrate, you can use transfer
curves to compensate, and use
just one profile for both.
Rut he warned: transfer functions are an expert feature that
can cause a great deal of trouble,
especially if they're saved into
image files. It's a great deal safer
to download a transfer curve to
the imageserter or piatesetter
RIP thai it is to huild one into an
image. Unless you know exactly
what you're doing, and have
communicated it clearly to your
prepress and printing providers,
don't mess with transfer curves.

black curves. More sophisticated packages offer complete contro! over

the black curve shape, start, and end points, a n d also let you decide how
far to extend blackaway fromthe neutrals a n d into saturated c o l o r s s e e
Figure 8-15.
When making press profiles, we rarely make a single profile. Instead.
we build a family of profiles that differ only in their black generation to
accommodate different image types (see the sidebar, "Black Generation,"
earlier in this chapter). Often, after we've used the profiles, we'U go back
to the measurement data and generate new profiles with slightly refined

906

R u l World Color Management, kcond Wltion

~ a e b e t h 'P@leMakerPro
s
o m presetr fora cnar&tyofdi@mtpress types, but also allows you to
control the tomiand black ink limits, the blackstart,and theblack width, the last ofwhich fine-tunes the
extent to which bhcke*tL)ndFt?vmthe n e u W into colors.

black curves. Until you've used the profiles on live jobs, it's pretty much
impossible to tell if your profile would be improved by setting the black
start value to 38%rather than 40%, for example. See Chapter9, Et~fllmting
and EditingProfUes,for profile evaluation techniques.

Profiling Multichannel Printers

Sou might think any printer that uses more than four inks qualifies as a
multichannel (sometimes referred ro as deviceN Color) printer. Nothing
could be further from the truth. In an overwhelming majority of cases.
printers that have six, seven, eight, 10, sometimes 18 inks are treated and
profiled as either RGB or CMYK.The challenge is in knowing which one.
In most cases, you'll need to make both an RGB and a CMYK profile to
see which is better. The RIP or printer driver separates the RGB or CMYK
data you send it into the control signals for the specific inks in the printer.
A few inkjet RIPSactually let you control each inkindividually and offer
support for six-,seven- andlor eight-color ICC profiles. In this case, making such a multichannel profile may be worthwhile. ln the case of printing presses. it's really the only way to go because you need multichannel
separations to generate the film and plates for use on press. Be w e d ,
however, that few profiling packages provide support for multichannel
profiling, and those that do often do so only at extra cost.
Building multichannel profiles is the trivial part of the task. Printing
the target optimally and collecting the measurement data are the tricky
parts. No one has massive amounts of experience building multichannel
profiles--you're on the bleeding edge--so the best advice we can offer is
to remember the guiding principles of profiling:
b

Do what's necessaryto make the device behave the way you want it to.

Print the profiling target while the device is behaving that way.

Make sure that the measurements you collect truly represent the
device's behavior.

308

Real World Color Management, Second Edition

Checking the Map

If you're careful about following the principles we'velaid out in this chapter, you may get perfect profiles on the first try-maps that show exactly
where your color will go. Our experience tells us that although a newly
built profile may be very good indeed, it can almost always be improved.
In the next chapter, we look at techniques for evaluating and editing
profiles.

Color Orienteering

In BNC~'S
native land, Scotland, many otherwise-normal people happily
spend rainy Saturday afternoons tramping across heath and bog, peering
through wet glasses at the map in hand, pondering how to relate its contents to the ankle-deep water in which they're standing. They're indulging
in the grand British pastime of orienteering-negotiating terrain using a
map as their guide.
Evaluatingand, optionally, editingprofilesis alot likeorienteering. The
profile is the map, and the device is the territory, hut now you're looking
at the map from the middle of the territory, possibly in ankle-deep water,
and figuring out just how closely the two correspond.
One view of evaluating profiles is that it's an exercise in determining
how lost you are. Another is that it's fundamentally futile to try and put
a metric on people's subjective preferences. We think it makes sense to
find outhow inaccurate your maps are. Once you know,you can decide to
make themmore accurate by editing them to better match the territory,or
you can decide to allow for the inaccuracies and learn to head in the right
general direction while keeping an eye on the nuances of the terrain.
We have one more reason for putting profiles through some kind of
systematic evaluation. Almost all color matches are the product of at least
two profiles. If you don't take the necessary steps to evaluate each profile's
accuracy in isolation, it's hard to pinpoint the culprit when things go

910

Real World Color Management, kcond Edition

Judging the Map

Compdshes, sextants, and other such devices are indispensable tools for
navigation, but at some point, you have to simply look at where you are.
By the same token, measurementsplay an essential role in color management, butwhen it comes to evaluatingprofiles,your eye has to be the final
arbiter-if it doesn't look right, it's wrong.
But beware of mirages: back in Chapter 1, we pointed out some of the
many tricks your eyes can play on you. So when you evaluate a profile,
you need to set up your viewing environment so that you can say with
certainty that any problems you see do, in fact, lie with the profile.
Your eye must be the final judge, but it doesn't have to be the only judge.
We'll show yousome objective tests that can help determine aprofile's colorimetric accuracy. But these only apply to colorimetric renderings-perceptual renderings always involve subjective judgments, because there's
no single correct way to reconcile two different gamuts. So objective tests.
while they're useful, don't tell the whole story At some point, you have to
make subjective judgments, and to do that, a stable viewing environment
is critical.
You may be rempted to edit profiles to fix problems that really lie in
the device calibration or the data collection. We can tell you from bitter
experience that doing so is akin to buying a one-way express ticket to the
funny farm. It's often quickerand easier to recalibrate andlor reprofile to
fix the problem at its source. We tend to use profile editing as a last resort
(though it's also an integral part of our workflowsl, so throughout this
chapter we'll point out which problems can be fixed more easily through
other means, and which ones are amenable to profile edits.

Viewing Environment
Back In Chapter 3, we pointedout that virtually all the color matches we
create are metameric in nature-that is, they're dependent on the light
sourceunderwhich we view them. Sowhen you come toevaluate profiles,
it's vital that you view hard-copy samples under a controlled light source
such as a D50 light box But simply plonking an expensive light box into
an otherwise-imperfect viewing situation is like sticking a Ferrari engine
into an AMCPacer-it'U cost you plenry, but it may not get you where you
wanted tn PO.

Chapter k Evaluating and ldlling Pmfla

211

Some color management purists insist that you must work in a windowless cave with neutral gray walls, floor, and ceiling, and a low ambient
level of DSO light, while wearing black or neutral gray clothing. We agree
that this represents an ideal situation for evaluating coIor matches, but
it's a distinctly less-than-ideal situation for most other human activities.
[We know shops where rooms just like this improved color matching but
dramatically increased employee sick days.)
So rather than insistingon ideal conditions, we offer a series of recommendations for reasonable ones.We'll let you decide just how far you want
to go towards the ideal.

Surfaces
Surfaces within your field of view-walls, floor, ceiling, desktop--should
be as neutral as humanly possible. You most certainly don't want strong
colors intruding into the field of view because they'll throw off your color
judgment. But pastels can be just as insidious: Bruce moved into a workspace with vent pale pink walls, and he found that until he painted them
white, he introduced cyan casts into all his images!
If you decide that neutral gray is the way to go. Chris has compiled
some paint recommendations from a variety of sources:
b

Shenvin-Williams paint code: 2129 ZIRCOM. You'll want flat paint

rather than glossy. (Man Louis, Louis Companies, Arlington, Texas)

California photographer Jack Kelly Clark recommends mixing one

gallon of Pittsburgh Paint's pastel-tint white base #80-110 with Lamp
Black (B-12/48PPG). Raw Umber (L-36/46PPG),and Permanent Red
(0-3148 PPG).Write those numbers down and take them with you to
the paint store if you want to try the mix yourself.

A similar Kelly-Moore Painr Co. formula from photographer John

Palmer uses apastel-tint white as a base with three colors to create an
interior, flat latex similar to Mumell8 gray: Lamp Black (4148 PPG),
Raw Umber (27148 PPG). and VioIet (2148 PPG).

If all this seems a bit extreme, the main consideration is to ensure that
the field of view you use to evaluate hard copy is neutral, and that color
in the room doesn't affect the light you use to view the hard copy Bear in
mind that white walls tend to reflect the color of the ambient light-it's
manageable as longasyou'reaware of it and takesteps tocontrol it. Glossy
h1""L

',."I.

:A""l ,
.
:
.
h
"
-

"" 2.

A"-

-"..",.A:

-.-,,.:----

n..,..:---

919

Real World Color Mmagemcnt, a n d Wltlon

Lighting
The IS0 (International Standards Organization) has set standards for illumination in the graphic arts. For example, IS0 3664 specifies D50 as
the standard illuminant for the graphic am. It also specifiesluminosity of
500 lux for "Practical Appraisal" and 2,000 lux for "Critical Comparison."
See the sidebar. "Counting Photons,"for a definition of lux, lumens, and
candelas. The standard takes into account the fact that both apparent
saturation and apparent conhast increase under stronger light (these effects are named for the scientists who first demonstrated them-the Hunt
effect and the Stevens effect, respectively).
But the standard wasn't created with monitor-to-print matching in
mind-it mandates that the ambient illumination for color monitors
should be less than or equal to 32 luxand mustbe less than or equal to 64
lux. For monitor-to-print matching, all thesevalues are way too high-the
IS0 has acknowledged this. and is still working on standards for this kind
of match. So until these standards are published and ratified (which may
take several years),we offer some practical advice that we've gleaned empirically (that is, by trial and error) over the years.
While it might be mostly a concern for theupcomingboard game Color
Geek Trivia-Millennirrm Edition,we should be clear that D50 isan illuminantwithavery specific spectral power distribution that no artificial light
source on the planet can replicate. The term used for most light sources
is correlated color temperature for which there are invariably multiple
spectra. 5000 Kis an example of correlated color temperature, not an illuminant. Different 5000K light sources have different spectra and produce
slightly diierent appearances.

Ambient light. Ifyou're workingwitha CRT monitor, youneed lowambient light levels because CRTs just aren't very bright.The color temperature
of the ambient light isn't terribly important: what isimportant is that you
shield both the face of the monitor and the viewing area for hard copy
from the ambientlight. Filtered daylight isOK. FulIsunfrom a west-facing
window is not. both because it's too bright and because it imparts a lot of
color to your surroundings.
With some of the latest LCD displays, such those from Apple and EIZO,
you can use much higher ambient light levels simply because these displays are so much brighter than CRTs, but the same rules about excessive
coloration are as applicable to LCDs as to CRTs.

Chapter 9: Errluatlng and Edtllng Pmfllu

413

Counting Photons
Photometry-the
science of
countlng photons-is a basic
building block for measuring color. But there's more than oneway
to count photons, and different
methods use differentunits.
The Iumiizo~rsp l ~ x[or lrcmitroos yotcwr) is the amount of
visible light energy a light source
emits over time in all directions.
It's computed by measuring the
power in watts thc light source
emit*, weighted by the spectral
sensitivity of the eye.The unit for
luminous flux is the llcmen ilm).
which is the luminous fluxa reference light sourceemits.We uselumens to describe the overall light
output of lamps without regard
to the lamp's reflectors or the distance from the lamp to what it's
illuminaling. A tvpical 100-watt
lightbulb ha3 a lununous flux of
about 1700 lumens.
The lurninozts intensity is
the amount of visible light
energv over time from a point
light source in a given direcuon.
The unit for luminous intensity is
the ctrndela (cd),which is equivalent to l~rrneilsper srerndinn
(Im/srl. (A sferndimis a standard
unit ofsolid angle, lust as anangle
cuts out a subsection of a c k l e , a
solid angle forms a cone that cuts
out a subsection of a sphere and
represents theconeoflight travelling nut from the point source in
the given direction.) The candela

also has an equivalent definition

based on the luminous intensity
of a reference light source. The
candela is one of the fundamental
units of measumneot (alongwith
such things as grams and meters)
defined by the SI (SystemIntemationale &Unit&).
The Iuminnnce refers to the
light given off by a more typical tight source (which is alightemitting surfacr?.not a point) or
an illuminated
y surface.'lheluminance is the luminous intensity
per unit area of the light-emitting
surface. It's measured in candehs
per sq~inremeNr, which is sometimes also known as the nit (from
the Latin nirere, to shine). Luminance is thequantity represented
by Y in the CIE equations, and in
color management we most often
enCOUnM!' luminance as a measurement of the photons emitted
from a displa
Finally, the ill~rrninnncerefers
to the amount of light arriving
at an illuminated scene. It's just
the luminous flux per unit area of
the illuminated surface, and is
measured in h~niensper sqtrnre
mmer, or iux. (A nonmeuic unit
is the footcnndle, which is one
lumen per square foot, or about
10.76 lux.) Typical supermarket
lighting has an illuminance of
about 1000 lux. Public areas like
hotellohhies have an illuminance
of around 300 lux, while the full

moon produces an illuminance

of about 0.4 lux.
In the definitions forthelumen
and candelaunits above, wementioned a refprence light sottrce.
This reference has an interesting history. Much of photometry
stems from an early unit called
the candle.This was the perceived
intensity of light in one direction
from a specified standard candle
(116-pound s p e r m w h ~ l e wax
burning at the rate of 120 grain?
an hour).
This reference candle was later
replaced by specific types of oil
lamps, then by carbon filament
lamps, and in 1948 by a blackbody
radiator at a specific temperature
(to he precise, the freezing point
of platinum-about 2042 Kl.
Because blackbody radiators are
n pain to constmet, the reference
light source was redefined in 1978
as a light source with a specific
frequency (equivalent to a wavelength of about 555 nm, which
is where our brightness sensitivity hits its peak) and a specific
radiant energy (11623 watts per
steradianl.
In all cases, this seemingly arbinary definition is just the standards people's way of coming up
wirh a definition that produces
about the same unit quantity that
people are already using-in this
caye tracing all the way back to
that whalewax candle.

914

Real World Color Management, Second Edition

Tip: Use a Monitor H o d Y o u can increase the apparent contrast of CRT

monitors significantly by using a hood to shield the face of the monitor
From the ambient light High-end CRTs often include a hood, but ifyours
didn't, you can easilpfashion one for a few dollars from blackmatte board
andadhesive tape--it's one of those investments whose bang for the buck
is simply massive. It's also a good idea to wear dark clothing when you're
evaluating color on a monitor, because light clothing causes reflections
on the screen that reduce the aovarent contrast.
Hard-copy W n g light. Unless yourprofle buildingapplicationhas addilionaloptions,thecolorscience onwhich ICCprofiles arebased isdesigned
to create color matchesunderD50lighting,so for your criticalviewinglight
D50 illumination would be ideal, but as we've already pointed out, there's
no such thing as an artificial light source with a D50 spectrum. So-called
D50 light boxes are really "D50 simulators."The minimum requirement is
a light boxwith a 5000 K lamp and a Color Rendering Index (CRI)of 90 or
higher. The best available solution is a 5000 K light box that lets you vary
the brightness of the light, so [hat you can turn it up when you're comparing two hard-copysamples and turn it down to match the brightness of the
monitorwhen you're comparing hard copywith the monitor image.

Tip: Tailor the Ptotlle to the Light Soum. A few packages (such as GretagMacbeth's ProfileMakerPro 4.1 and later) let you build profiles tailored
to the spectral power distribution of specific physical light sourcesthey
include spectral measurements for some common light boxes and also let
you define your own. So instead of D50 LAB, you can end up with "GTI
LightboxL4B."Thislets you make verycritical color judgments under your
viewing light. the trade-off being that those judgments may not translate
perfectly to otherviavingconditions. Ifyouuse this feature, use some other
light source as a reality check-we really like filtered daylight!
A variable-intensity 5000 K light box is a fairly expensive piece of equip-

ment. but if you're serious about color management. it's a worthwhile

investment. Rut it's not the be-all and end-all, either. As previously mentioned. many artificial light sources have spectra with more pronounced
spikes than the relatively smooth spectrum our sun produces, and with
some inksets, such as the pigmented inks used in some inkjet printers,
which crlso have fairly spiky spectra, you may see significant differences

Chapter 9: Muadng and Edldng Plonla

P15

between hvo5WOKlight sources, though you'reunlikely to see differences

with press inks.
Bruce always uses daylight,preferably filtered,indirect daylightas areality
check.Wheneverhe's measured daylight anywhereclose to sealevel,it'sbeen
significantly cooler than D5Ctypically between 6100 Kand 6400 K-hut the
smoother spectrum eliminates some ofthestrange behaviorthat results from
the combination of a light source with a spiky spectrum and an inkset with
a spiky spectral response curve. We also like the 4700K Solitx lamps--while
they're a tad warm, they produce a vei-ysmoolh spectrum with an extremely
high CRI of 98. See www,solux.net for more information.

Tip: Urn a Viewing Light Chccker.Viewing light checkers are ingenious

littlecolor Largelsprinted with twocolors [hat match underD50butshow
different colors under other light sources. One such is the GATFIRHEM
Light Checker-we've included one on theinside-back cover of the book.
Place the checker on top of the sample you're evaluating. If you can't see
stripes, your lighting maybe good enough for all but the most critical
color matches. If you can see obvious stripes, you definitely need to improve your tightingsituation-these test strips don't validate good lighting
conditions, they simply identify bad ones.
Monitor-to-print comparisons. CIE colorimetry wasn't designed to handle monitor-to-print comparisons--colormonitors didn't exist when the
models on which color management is based were beingdeveloped. Nevertheless, it's capable of doing a surprisingly good job.
Some punditswill tell yon that it'simpossihle to matchamonitorimage
and a printed one, because the experience of viewing light reflected from
paper is simply roo differentfrom that of viewing light emirtedfrom glowing phosphors. In averynarrow sense, they may be right--one important
difference between hard-copyand monitorimagesisthatwith the former,
our eyes can invoke color constancy (also known as "discounting the il1uminant"-see "Color Constancy" in Chapter l. Wharls Color?),whereas
with the latter, we can't since the illuminant is the image.
This may help to explain why so many people have difficulty matching
animageonamonitorcalibrated toa5000Kwhite pointwith ahard-copy
image in a5000 Klight box.There may be other factors, too--a lot of work
remains to be done on cross-media color matching-hut whatever the
reason, the phenomenon is too well reported to be imaginary,

46

Real World Color Management, Second Edition

We offer two pieces of advice in achieving monitor-to-print matching:
Match the brightness, not the color temperature.
Don't put the monitor and the light box in the same field of view.
\%en givena choice between 5000 Kand6500K wecalibrate our monitors to 6500 K becauseit'sclosertotheirnative white point; we can obtain
a better conuastratio than we can at 5000 K, which requirestumingdown
the blue channel. We dim the light box to match the contrast ratio of the
monitor-a sheet of blank white paper in the light box should have the
same apparent brightness as a solid white displayed on the monitor. If
you're happy working with a 5000 Kmonitor, by all means continue to do
so, hut no matter what the white point, it's critical that you throttle back
the light box to match the brightnessof the monitor.
The second trickistoplace thelight boxat right angles to themonitorso
that you have to look from one to the other. This accomplishes two goals:
it lets youuse your foveal vision--the cone-rich area in the center of your
retina where color vision is at its most acute--and it allows your eye to
adapt to the differentwhite points. (Short-term memory has been shown
to be very accuratewhen it comes to making color comparisons.)
We've been using these techniques to do monitor-to-print matching
for several years, and we find them very reliable. Does the monitor image
perfectly match the print? Probably not But frankly,short of apressproof,
we're not sure that we've everseen aproof that matched the finalprint perfectly. Laminated proofs, for example, tend to be a little more contrasty,
and perhaps a shade pinker, than the press sheet-not to mention their
inability to predict wet trap or print sequence--but we've learned to filter
out these differences. You have to learn to interpret any proofing system,
and the monitor is no exception.

Evaluating Profiles
Once you'\ e 11onedall the kinks out of pourviewingenvironment, you can
safely stan evaluating yout profiles. One of the trickier aspects of profile
evaluation is being surewhich profileis responsiblefor any problemsyou
might see. because most of the color-matching exercises we go through
need at least two profiles, sometimes morr.

Chapter 9: Evaluating and Editing Proffltl

917

Anything that you view on-screen goes through your monitor profile,
so the monitor is the first dwice to nail down-once you know you can
rely on your monitor profile, you can use it as a basis for comparison. If
you just trust it blindly, and it's flawed. you'U create a huge amount of
unnecessaryworkfor yourself. Once you've qualified your display profile,
it's much easier to use it as an aid in evaluating your input and output
profiles.

Checking the Display

-

b u r display is not only the first device you need to nail down, it's also the
one type of device where calibration and profiling are usuaUy performed
as a single task. Since it's much more common to find problems with the
calibration than with the profile, you need to check the calibration first-if
it's bad, the resulting profile will be, too.

Monitor Calibration
The two most common problems with monitor calibration are incorrect
black-point setting and posterization caused by trying to apply a gamma
that's too far away from the monitor's native gamma for the 8-bit tables
in the video card to handle.
To check the calibration, you need to take the profile out of the display
loop and send RGB values directly to the display To do so, set your monitor profile as the default RGB space in the application of your choice so
that RGB is interpreted as monitor RGB, and hence sent directly to the
screen with no conversions.

Tip:

hoof Setup in Photoshop. In Photoshop 6 or later, choose

Viewproof Setup>MonitorRGB to send the RGB values in the file directly

to the monitor. When you choose Monitor RGB. Photoshop automatically
loads the Monitor RGB profile in Proof Setup and checks "Preserve Color
Numbers," effectively taking your monitor profile out of the loop.
We generally use Photoshop for this kind of testing. though almost any
pixgl-based editing application should serve.

918

Real World Color Hanagclmnt, Second Wltlon

Black-point check. Setting the correct black point is the biggest challenge for monitor calibrators for two reasons:
F

The monitor output is relatively unstable at black and new-black.

Most of the instruments used for monitor calibration are relatively

inaccurate at black or near-black-they attempt to compensate by
averaging a lot of measurements.

Be warned that the following test is brutal at showing the flaws in most
monitor calibration and can lead to significant disappointment! (See
Figure 9- 1.)

In Photoshop (or the pixel editor ofyour choice),create a

solid black image, and make a selection in the center of the
document with the marquee tool.

Chapter O: Evaiuating and Editing Proflia

PI9

Black-point test,
continued

Hideall palettes, switch tofull-screen mode with no menu bar; then hide the
"marchingants."Open Curves or Levels,and target the 0,O point in Curves.

Drag the dialog box off the screen, leaving only its title bar showing (theflare
from the &log's large expanse of white can preuentyoufrom seeingsmall
differences, which is what this test isabout).

220

Real World Color Management, Second Edition

Black:-point test,
continued

Use Ilre up nrrou, to rnfre tlre level of tlze selected nrefl-eoch preys rnrses rt by
one lel~el.

With excellent calibration systems, you may see a difference between

level 0 and level 1. More typically, you won't see a change until somewhere around levels 5 to 7, or sometimes even higher. If you don't see
any change when cycling through the first 12 levels, your black point is
definitely set too low and you should recalibrate, requesting a slightly
higher black point.
If the first few levels that are visible have a color cast, you may have
set the bias controls incorrectly (if your monitor has them). Try making a
small adjustment to the bias-for example, if the first few levels are red,
uy lowering the red bias slightly. Then recalibrate.

Gamma check.To refine the gamma setting, display a black-to-white gradient. We usually flip the top half of the gradient horizontally to produce
a test image like the one shown in Figure 9-2.
If your application allows it, display the gradient in full-screen mode.
hide all other user interface elements, and then lookat the gradient closely. in an ideal situation, you'll have a perfectly smooth, dead-on-neutral
gradient, and black will fade smoothly into the nonpicture area of the
monitor.

Chaptor 9: Evaluating and Editing Profilcr

221

n g u 9-P
~

Gradient test image

Display the gradient rest i n q e in fir//-screen mode, with all user interface
elements Ridden, for best results.

In practice, this rarely happens.You'll almost certainly see some slight

banding or posterization in the shadow areas, and you may see some
colorwhere you should see neutral gray. Make sure that you're looking at
raw RGB sent straight to the monitor-at this point, you're checking the
calibration, not the profile.
Color crossovers in araw RGB gradient almost always indicate a fatally
flawed calibration tool. If you've adjusted the R, G, and B bias controls on
yourmonitor, it'spossible to introduce color crossovers,so ifyou see them,
the first thing to try is to reset the bias controls to the factory defaults. If
you still see color where you should see grays, and you're absolutely sure
that you aren't looking at the gradient through a profile, you need to toss
your calibration tool and get one that works-this isn't something that you
can fix with any amount of recalibration or profile editing.
More commonly, you'll see some slight amount of posterization in
the shadows and three-quarter tones. Until the video card manufacturers decide that higher-precision videoLUTs are more important than the
number of shaded polygons the card can draw per nanosecond, we'll all
have to live with some slight irregularities in tonal response, but you can
often improve the smoothness of the calibration by recalibrating, and by

222

Real World Cdor Management, Second Edition

changing the requested gamma to something closer to the native gamma
of the display system. Color management needs to know your monitor
gamma, but it doesn't really matter what the absolute value is, only that
it's known--so feel free to experiment with different gammas between 1.8
and 2.4 or so, and choose the one that gives you the smoothest gradient.
There are so many combinations of monitor and video card that we
can't give you any magic numbers. Try raising the requested gamma by
0.1.Iftheposterizationgets worse.tryloweringit by 0.1. Eventually, you'll
Find the best compromise. When you've done so, don't forget to generate
a new profile!

Monitor Profile
Monitor profile problems are relatively rare-flaws are usually either in
the monitor calibration, or, in cases of gross mismatches between screen
display and printed output, in the printer profile that senres as the source
in the conversion from print RGB or CMYK to monitor RGB. Nevertheless,
some monitor profiles may work better than others. If your profiling tool
offers the ability to build different types or sizes of profiles, you may find
that one type works much better than another.
Repeat the gradient test shown in Figure 9-2, but this time, set your
working space to a gamma 2.2 profile such as sRGB or Adobe RGB, or
simply assign one of these profiles to the gradient image, so that you're
displaying the gradient through the monitor profile. Ifyousee color crossovers that weren't visible in the raw display test, it's possible that your
profile contains too much data-we often see LUT-based monitor profiles
producingpoorer results thansimplermatrixor gamma-value profiles, so
don't assume that bigger is always better.

Reference Images. Of course, you'd probably rather look at images than

at gray ramps. But the classic mistake many peoplemake when evaluating
a monitor prolile is to make a print, then display the file from which the
print was made, and compare them. The problem is that the display of
the print file is controlled as much by, or even more by, the printer profile
used to make the print as it is by the monitor profile, so you can be fooled
into trying to fix problems in the printer profile by editing the monitor
profil+which is like trying to fix the "empty" reading on your car's gas
gauge by tweaking the gauge instead of putting gas in the tank

Chap-

9: Evaluating and Edltlng Protiles

3P3

When it comes to comparinga monitor image with a physical reference,

the only reliable way to make sure that the monitor profile is the only one
affecting the image is to compare an image in LAB space with the actual
physical sample From which the LAB measurements were made. There are
very few sourcesfor aphysical sample and amatching calorimetrically accurate digital file--Koda!& now-discontinuedColorFlow ICC Profile Tools
supplies a stringently manufactured photographic print with an accompanying digital file, but we don't h o w of any other vendor who does this.
So unless you own ColorFlow, you'll need to use a little ingenuity. Here
are some of the techniques we use.

The Hacbeth CdorChttkw. The Macbeth ColorChecker, made by Gretaghlacbeth, has long been used by photographers to check colors, and is
available From any decent camera store. GretagMacbeth has never published LAB values for the 24 patches, but the LAB values shown in Figure
9-3 are ones that Bruce has collected and averaged over the years From
various Macbeth targets in various stages of wear (with considerable help
from Bruce Lindbloom and Robin Myers). Or you can simply measure the
24 patches yourself.

ColorChecker

The Macbeth ColorChecker with its associated LAB values

924

Real World Color Management, Second Edition

Then. create a LAB image in Photoshop (or any other application that
lets you define LAB colors), and compare the physical color checker, appropriatelyilluminated, with theimage on screen. With CRT monitors, the
saturated yellowpatch (row 3. column 4) is outside the monitor's gamut,
so you'll likely see a slightly more orangey-yellow. LCD monitors can hit
the yellow, but may miss the sky-blue patch in row 3, column 6. But in
either case, both the color relationships and the tonal values of the gray
patches should be preserved reasonably weU.

Scanner targets. If you have a scanner target, then you have a physical
target and the LAB values of the colors it contains. All you need to do is
create an image of the target containing those LAB values. Creating hundreds of LAB patches in Photoshop, though, is too much pain even for us,
but fortunately,there's a solution, and it's free--see the sidebar,"Measurements to Pixels and Back," on the facing page.
Use the procedure outlined above for the Macbeth Color Checker to
compare the physical target to the LAB imageon screen.You'll likely have
more out-of-gamut colors than you would with the Macbeth target, but
the relationships between all but the most-saturated colors should be
preserved.
Printer targets. Leverage your printer target!Yourveprinted a target and
measured it. Turn the measurements into pixels-see the sidebar, "Measurements to Pixels and Back" on the facing page--and compare the onscreen image with the printed target under appropriate illumination.
Your printer target will likely contain more patches than a scanner target, and fewer of them will lie outside the monitor's gamut, so it can give
you a good idea of your monitor profile's color performance.
We don't advocateeditingmonitor profiles. It's not a philosophical objection-we simply haven't found it effective. If you find a display problem
that you can more effectively solve by editing the monitor profile rather
than by recalibrating and reprofiling,please let us know-we'll gladly eat
our words, if not our hats!

Chapter 9: Evaluatlng and Editlng M k r

995

Measurements to Pixels and Back

Logo ColorLah 1s an undocu- ushlg Nearest Neighbor interpolamented and unsupported free tion in Photoshop. I1 the patches
tool that you can download from have borders. you need to do some
GrelagMacheth's Web site. We workgetling rid of them fint.
ColorIab's requirements for
don't pretend to understand all
the things it does, hut it ha? one measurement or target files are
very useful capability we've come pretty straightforward.They need
to find indispensable. It lets us to be in tab-delimited text format
rake measurement files in LAB. with a header that looks like the
or target definition files in RGB or following:
LGOROWLENGTH
24
CMYK, and turn them into TIFF
BEGIN-DATA-FORMAT
images in whicheach color patch
U\BL L A K A LAB-B
is represented by a single pixel.
END-DATA-FORMAT
Since thecolor patchesare solid
BEGTN-DATA
colors. you only need one pixel per
The
first line tells ColorLah how
color-you just print and displav
many patches per column [even
them at very low resolutions.
ColorLab can also go the other though it says row),and the numway. turning pixels into target files her is separated from the rest of
(RGBor CMW or measurement the tine by a single space. MU,
files (LAB), the only catch being LAB-4, and L A R B on the third line
that the image files need to he in areseparated by tahs.AfterBEGIN1-pixel-per-color tormat. If the tar- DATA, record the three LAB values
get consists of solid patches with for each color, separated by tahs,
no borders, you can usually lust and separate each colorhya Return
downsample to one p i d per color (soeach color is on a new line).

Input
- Profiles

At the end of the data, youneed

a final Une that says
END-DATA
We usually use a spreadsheet
like Microsoft Excel to massage
measurement files into theshape
ColorLab needs-when you save
an Excel file as Text, each column
isseparated hy a tab and each row
by a return.
We keep the header in an otherwise hlank Excel doc, and simply
paste the measurement values
from measurement files into it.
ColorLah can tolerate other information in the file, so try simply
opening your measurement files
in ColorLah fint. But if you get a
"format error" message, the foregoing header always works.
Be warned that ColorIab is unsupported, occasionally bugg):
and sometimes hard to 'ld. You
can find out itscurrent location hy
going to www.colorernedirs.com/
realworldcolorldavnloads.hunl.

Scanner profiles are, as we noted in Chapter7, BrciliingIr~prrrPmfi[es,relatively easyto build. Scanners have aflxed, reasonably stablelightsource.
a n d as long a s you make sure the software settings remain consistent,
they usually behave quite predictably. Digital camera profiles are much.
much harder, both becausecamerasfimction undera hugevarietyof light
sources, and because they have tocapture photonsreflectedfromamuch
wider variety of objects than d o scanners. That said, there are a few tests
that you can apply to either one.

226

Real World Cokr hnagemenl, W n d Wlllon

Basic Test for lnput Proflles

The first simple test is to open the scan or digital-camera capture of the
profiling target in Photoshop,and assign theinput profile to it.Then comp~are
the image on your calibrated monitor with the appropriately illuminated
physical target Ifassigningthepmfle doesn'timpmvethematch,you should
pmbably just start over, double-checldngyour methodology. If the match is
impmved (as we'd hope). yo11 may want to use a more objective test.

Objective Test for lnput Profiles

This lest uses a simple principle: we check the profile by comparing the
known LAB values in the profiling target with the LAB values predicted by
the profile. In an idealworld, they'd beidentical, but inourexperience, this
never happens. Ifyou care why, see the sidebar, "Objective Objectives," on
the facing page. The reasons we do objective tests on input profiles are to
help us understand and optimize our capture device's behavior, to identify problem areas in the profile that may respond to editing, and to help
us understand what we see when we do subjective tests with images.
To perform tests such as this one, you have a choice. You can use tools
that are available at no cost but that require considerable work on your
part, or you can spend money on Chromix ColorThink Pro, which does a
great deal of the work for you and has many otherinvaluable capabilities.
We'll cover both approaches.
First, the h e method:You can employ the ever-useful Logo ColorLab
(see the sidebar,"Measurementsto Pixels and Back," earlier in this chapter)
to make the comparison.You need two TIFF images of the target with one
pixel per color patch: a LAB image of the target values, and an image containing the LAB values predicted by the profile, which is a little trickier.
Some profiling tools save the RGB values they capture from the target
scan as a text file, which you can then open in ColorLab. Others. including
those from GretagMacbeth and Heidelberg,store the captured RGB values
right inside the profile as a "private" tag-private in name only because
you can extract the data by opening the profile with a text editor, then
open the text file in ColorLab.
Forthose that do neither, youcan downsample thetarget scan ordigital
camera capture to one pixel per patch using Nearest Neighbor interpolation in Photoshop, or build a I -pixel-per patch image manually in Photoshop. Building the target by hand is time-consuming but more accurate
than resampling.

Chapter 9: M w # n gand Cditlng Prolila

997

Objective Objectives
You can lead yourself down all
surts of fascinating rabbit holes
looking at numerical evaluations
of profiles-we know because
we've been there. So let's make
somethingveryclear. Wedon't do
objective tests to come up with
some unambiguous benchmark
of profile quality, for Iwo reasons:
b

It's impossible to determine

a profile's absolute accuracy

accurate pmfile can very easily

produce posterizatlonand wlorbanding where a less-accurate
one will produce smoother
transitions. We often sacrifice
absolute accuracy for smoothness, because our eyes are a
lot more forgiving on small
discrepancies in hue, saturation, or lightness than they
are when they encounter
wdden obvious discontinuities.

unless you're wining to sample

and measure every possible colSo what's the point of objective
or combination in the profile's tests? Simply#they help us underspace, which is, to put it mildly, stand how the profile behaves,
impractical-you'd need 16 and why we see the results we do
million-some measurements in our subjective tests They help
in RGB, and 4 billion-some us understand the limitations of
measurements in CMYK. The our devices. and they point us
objective testswe do realtyonlv to areas in the profile that may
measure the profile's ability to respond to profile editing.
predict or reproduce the colors
You may be tempted to use
in the test target we use. Differ- objective testing to compare
ent targets v d l return different profiling packages. The single
average and maximum delta-e best word of advice we can give
values.
you is, don%
What the tests we've described
b Accuracy isn't the only factor
in determining a profile's qual- here study is the profile's ability. All CMMs operate at finite ity to predict and reproduce the
precision, making calculations colors in the target from which
from limited data, and a highly the pmfile was built. These tests

fail as an objective comparative

benchmark for two reasons:
b

They give an unfair advantage

to profiles that were actually
built from this target.
They tell you nothing about
the reproduction of colors that
aren'r in the target.

You can use these types of

comparison to gauge thc relative accuracy of profiles built
from the same target, and as a
leiy rough comparison between
profiles built from different
ones-hut you'll get different
numbers, though generallyheadlng in the same direction, if you
use different targets.
So you can't use rhese tests to
say that profile roo has an ovetall delta-e of 1.28 while profile
bar has one of 1.69, except in
the contea of the specific set of
colon in the target. (Again,if you
werewilling and able to measure
16.7 million RGB patches. or 4.3
billion CMYK ones, you might be
able to make such a statement.
If you get around to it, let
us know!l

A third alternative is to create a text file bv hand, sampling the RGB values and enteringthem into the text f i l e w e generally use Excel to do so. If
you're building either a t e a file or an image by hand. it's easier if you s t a n
outwith atemplate. For text, open the target description file in Excel, and
replace the LAB values with your captured RGBs (don't forget to change
the DataFormat definition). Ifyou want to build an image. open the target
description file in ColorLab, then save it a s an image. Figure 9-4 shows the
ctnn.

rnm,i-rl

tn m l k n th.3 rnmnlrirnn

PP8

R u l World Cdor Manylancn(, kcond W k n

Figure 9-4

Step 1: Capture the RGB ualuesfrorn the scanned or photographed target.

Comparing acNal and

predicted LAB values

Ifyour profiling tool

includes the RGB data in
the profile, you can open
the profile with a text
editor, then cow and
paste the RGB
values into a new text
document. Start at the
'BEGINDATA FORMAT"
line, andfinish at the
'EivDrnTArnRMAT'line.
Ifyour proJiling tool doesn't capture the RGB numbers in a way you can
access m i l 3 you'll need to use one of the following methods. You need to
like the one above, or an imane
wind up with either a textfileformatred
. .
with one pixel per color patch To createan image, it's easiest to use ColorLab
to create a templatefmm the target descriptionfile.
In ColorLab,choose
Openfmm the File
menu, and open the
target description file.
You'll get something that
looks like the
image at the right.You'll
need thisfile anyway to
make the comparison.

Since t h e m target is nonrectangular,you need to

chooselTB toRectFormatfrom ColorLab's Tools menu
tofill the restof the target with neutral gray.

m=w
Comparing actual and
predicted lAB d u e s .
continued

Ib wnvert thejile to a
TlFFWyoumopen
in Phomhop, choose
&portfrom ColorLzb!!
Pile menu, and ckoose
~ffFomffrom
the

Format menu. m u
mczy haw m change the
.atatemion tn
.tifmanucrllyl

Open the newlysavedfile in Photoshop, zoom in to 1600%so thatyou can see whatyou're doing,
convert the image to RGB (any oldRGB-it doesn't matter because you71 be replacing all the RGB
values with onesfrom your target capture), and saw it with a name like 'ScannemmeRGB."

To transfer the RGB valuesfim the target capture to the Scannername RGBfile, open the target scan and
arrange thefiles side by side. Set the marquee tool to Fixed Size, 1 pixel by 1 pixel, and select thefirst patch
in Scannername RGB.
Then select the eyedropper tool, set it to 5 x 5 average, and Option-click on thefirst swatch in the target
capture image to set that color as theBackground color Press Delete tofill the selectedpixel in ScannernameRGB with that cola5 press the right arrow key to advance the selection to the next pixel, Optionclick the next swatch in the capture target, and repeat until you'wfilled all the patches in Scannername
RGB with the valuesfim the capture mrget.

930

Real World Color Ilbnalpnrm(. Second Mllon

...

Step 2: Convertthe capturedRGB values to LAB (Photoshop),or

Comparing actual and
predicted WiB values,
continued
Ifyou built the RGBfile
by hand in Photoshop,
do the conversion to LAB
there as well.
Assign the capture
profile by choosing
Assign Profilefrom the
Mode submenu of the
Image menu.

Convert the image

to LAB by choosing
Convert to Projilefrom
the Mode submenu of
the Image menu. Use
Absolute Colorimetric
rendering.

After conversion, fill

the background
with neutral gmy,
LAB 50,O.0.

Chapter 9: Evalualing and W3na P M l r

R ~ r 9a4

Step2: Convert the captured RGB values to LAB (ColorLub).

Comparingactual and
predicted LAB values,
continued
Ifyou have an RGB text
file, do theconversion to
LAB in ColorLab.

Open theRGB textfile

in ColorLab.

Choose LogoSyncfrom
the ICCsubmenu of ColorLab's Filter menu.

In the ensuingdialog

Finally, chooselT8 to
Rect Formatfrom ColorLabS Tools menu.

Real World Color M a n a m , kcond Edltkn

Rgun 9-4

Step 3: Make the comparison.

Comparhg actual and

predicted LAB values,
continued

Open both the target

descriptionfile and the
file containing theLAB
ualues you obtained by
converting the capture
RGB numbers through
the profile. Makesure
that bothfiles are
converted to rectangular
format using Colorlab's
IT8 to RectFormat
command. Then choose
Compare Withfrom
ColorLab'sTools menu,
and choose thefile you
want to compare with
the current onefrom
the ensuingselect Document dialog box.
ColorLabgenerates a
comparison that shows
you the difference
between the actual LAB
values in the target and
the LAB values that the
profile predicts. It shows
you the average overall
delta-e, the average
delta-e for the best
90% and wont 10%
of patches,and the
maximum delta-e for
the best 90% and worst
10% of patches. The
worst 10% patches are
outlined In yellow,
and the wont patch is
outlined in red. You
can check any patch's
delta-e by clicking it.

&4 - 0 S S a m k m

14.6 I U l - W .

N.9

1.9 CMta m 428

S M 14.4 l4.e Ollt.. tdD

Chaptar 9: Evaluating and LdHing Plollla

433

This test provides a decent metric for comparing input profiles built
tom the same target.For example,your profiling tool may provide several
options for profile size. This test lets you determine which option gives
ie most m a t e profile (see Figure 9-5).

Comparing large an
s m d pm6k

A comparison of small (above)and large (below) profiles built

from the same scan shows that the large profile offers a clear improvement in accu~acyouer the small one.

934

Real World Cdor Mananmnent, kcond Wltim

With this pair of small (above)and large &low) digital camera

profiles,w $ n d that the reverse is true-the smaller profile is
slightly more accurate than the Iarge one.

Colorimetric accuracy is important, but it's not the only concern.

Sometimes we have to sacrifice accuracy for smoothness, for example.
A profile that renders many colors very accurately at the cost of heavily posterized images is usually less useful than one that distorts colors
slightly but produces smooth transitions between tones and colors.

Chapfn 9: Evaluating and Editing PmRk

!235

Figure 9-6 shows a typical scanner profile evaluation. Note that we

decided to improve shadow detail by raising the scanner gamma and
reprofiling, rather than editing the profile.

Scanner pmEle
evaluation

Here we seea
comparison ofprofiles
built by scanning the
target at different
scanner output
gamma settings.

-,..w-,-."

v v r m thisscanner;u

gamma of 3.0 produces

the most accurate
profile, but there are
always trade-offs. Some
colors, such as theS6
patch in this case,
m a y reproduce more
accuratelyat different
settings, but thegamma
3.0 setting produced t
best overall results wl
the smallest errors.

*
111

y
1
1

L.I

lY

w,

zm
,a

0.4,

I+.--YIUS~U

.*I->-

--

Gamma 3.0

Alternatively, you can perform the preceding tests using ColorThink

Pro, which does most of the work for you. If your profiling tool saves the
capture data in the profile, setting up the comparison is very easy indeed,
as showninFigure 9-7. ColorThiiPro automaticallyextracts the requisite
data from the chosen files, then goes ahead and makes the comparison.
Rgun 9 4

ColorThink input
test setup

For profiles that don't contain the capture data, you can create a onepixel-per-color image of the target capture using any of the methods previously outlined, then feed it to ColorThink Pro to create a Color List. In
either case, ColorThink Pro guides you through the comparison, and produces areport like the one shown inFigure 9-8.TheTest Feedbackwindow
shows the summary results, including overall, minimum, and maximum
delta-e, along with the same information for the best 90% and for the
worst 10%. The Color Worksheet shows the delta-e for each individual
color sample, alongwith a user-configurablecolor label that indicates the
delta-e range in which it falls.
ColorThinkPro also lets you make the comparison shown in Figure 9-6.
Simply load the target reference data, then the profiles. You can toggle
between the different profiles to see the delta-e values between reference
and sample, as shown in Figure 9-9.

Chap* 9: Evaluating and Editing ProRia

Fbum94

ColorThinlr inplt pmgleevaluation

B7

98

Real World Color Management, Second Edltlon

Subjective Tests for Input Profiles

Any test thatlooks at the rendering of images is necessarily subjective.The
key in evaluating an input profile is to select a representative sample of
images, with low-key and high-key elements, pastels as wen as saturated
colors, and neutrals. Profiling involves a series of trade-offs, and if you
judge the trade-offs on a single image, you're likely to regret it later. Of
course, you need to use some common sense as well. If you are scanning
or photographing only people, then you'll be more concerned with skin
tones and you'll want a suitable set of test images to reflect a variety of
skin tones.
It's also helpful to include afewsynthetic targets as a reality check: they
can often show you problems that natural images may mask or miss. Tho
such images that weuse regularlvare the Granger Rainbow,developed by
Dr. Ed Granger, and the RGBFXPLORER8, developed by Don Hutcheson.
both of which are shown in Figure 9-10. Note that we're constrained by
the limits of ourprinting processin howfaithfutlywe can reproduce these
targets-the RGB versions will look quite different on your monitor!
The point of looking at synthetic targets like these isn't to try to get
them to reproduce perfectly-that ain't gonna happen-but rather to provide clues as to why your images behave (ormisbehave) the way they do.
Sudden jumps in tone or color on the synthetic targets may happen in
color regions that arewell outside anythingyou'relikelyto capture. or thev
may lie in a critical area. It's up to you to decide what to do about them.
but we believe that y o u k betfer off knowing they're there.
In thecase of RGB input profiles, weuse the synthetic targets by simply
assigning the profile in question to the target and seeing what happens.
Often, the synthetic targets make problem areas very obvious. We always
make our final decisions by evaluating real images, but synthetic targets
can be real time-savers in showing, at a glance, why the profile is reproducing images the way it does. Figure9-11 shows a good example of this.
One digital camera profile, applied to images, seemed a little weak in the
redsand muddy in some yellows, but simply looking at a collection of images made it hard to pin down the specific flaw that one profile had and
another did n0t.A single glanceatthe Granger Rainbow, with eachprofile
applied, makes it very clear where the deficiencies lie!

C
h
a
w 9: Evaluatingand Editing Rollla

The Granger Rainbow.You can easily make this target in Photoshop.

M a h a horizontal rainbowgradienr,adda layer, makea black-to-white
vertical gradient, and set the layer blending to luminosity.

Don Hutcheson's RGBXPLQRER8. You can download this targetfrom

www.hutchcolo~com-click
theFree link.

239

ngum 9-11
Synthetic target
evaluatir-

effects areshown on
this page seem to
behave quite similarly to
the one whose
effects are shown on
the facing page.

A close emmination
of the images points to
the profile on the facing
page havinga problem
with saturated orangereds, but it's quite hard
to spot in these images,
even though they
contain a lot of reds.

Chap* 9: baluatlng and Wltlng P r d k ,

When we assign the

respective profiles to
the Granger Rainbow,
we can see at a glance
that the profile on the
previous page has a
reasonably smooth
response throughout the
reds, while the profile
on this page has a huge
"hole" in the reds.

This area of the image

hints at the problem, but
theRainbow makes it
clear at a singleglance.

Ri
u l World Color Management, Second Editlon
%!

Editing Input Profiles

With good scanners. we find we rarely need to edit the profile-the key parameter is finding the "ideal" gamma or tone curve for the scanner-but
with low-end scanners that don't provide sufficient control, or sufficiently
consistent control, we may resort to profile editing.
The simplest way to edit input profiles globally-tbat is, all rendering intents simultaneously-is to edit the capture of the profiling target.The only
catch is that you have to make edits in the opposite direction from thebehavioryou want in the profile. If the profile'sresultsare too dark, you need to
darken the target; if there's a red cast, you need to add red, and so on.
This is very much aseat-of-the-pantsprocedurethat absolutelyrequires
a well-calibrated and profiled monitor, but it has the advantage that it
doesn't require profile editing software, just an image editor. Small changes
can make big differences to the profile-with practice you can be surprisingly precise in edits to tone, saturation, and even to hue in selective color
rangesbut it's a fairly blunt instrument, and it affects all the rendering
intents. Soif you go this route, proceed with caution (see Figure$-12).
For more precise edits, or for edits to a specific rendering intent, you
need some kind of profile-editingsoftware. Most of the high-end profiling
packages offer editing in addition to profile creation. They all have different user intelfaces, but they work in generally the same way: they let you
open an image that'sintheprofle's space, anduse it asa referenceto make
edits. It should go without saying (but we'll say it anyway) that for this to
work, your monitor calibration and profile need to be solid.

Edit order.There's a classic order in which to make profile edits-it's the

same order that traditionalscanner operators use for image corrections-and we generally advocate sticking with it, with one exception, which is:
fix the biggest problemfirst!Because if you don't. it often covers up other
problems that suddenly become visible after you've done a lot of work.
The classic order for profile edits is
b

global lightness and contrast adjustments

tone curve adjustments

t gray bdance adjustments

b

global saturation adjustments

selective color adjustments

F i g ~ 9-19
r~
Editing the target

1he capture of the target, wrtn the camera prople

assigned, is shown above. It's quire good, but a little
unsaturated.

..

aI

I
..

-*I

To increase thesaturation in the profile, we &saturate

the capture of the target, then build a new profileilefrom
the desaturated target.

. . I

...I

--

The capture of the target, with the new camera

profile assigned, is shown above. It's noticably more
saturated than the original.

244

R u l World Color Management, Second EdlWon

The order isn't arbitrary: it's designed so that the likelihood of an edit
undoing the edits that went before is reduced, if not completely eliminated, so unless there's a huge problem staring vou in the face that you
need to correct beforepu can make any sensible decisions about other
issues, stick to it.
Figure 9-13 shows lypical editingsessionsfor a good midrange scanner,
The scanner profile requires only minor tweaks, which is typical.
Digital camera profiles are more difficult to handle than scanner
profiles. for three reasons:
b

They don't use a fixed light source.

They're more prone to metamerism issues than scanners.

There's often no target reference against which to edit them.

If your work is a l l shot in the studio, you can and should control the
lighting. If you have controlledlighting,and you gray-balance the camera.
youshouldn't have many more problems than with ascanner profile.Ideally you should edit the pmfilein thestudio, with areferencescene set up.
so that you can compare the digital capture to a physical object.
One problem you're unlikely to find in a scanner is camera metamerism, where thecamera either sees two samplesthat appear identical to us
as different,or more problematically,where the camera sees two samples
that appear differently to us as identical. (Scanners aren't immune to this
either, but if you're scanning photographic prints or film, you're scanning
CMY dyes-digital cameras have to handle the spectral responses of real
objects, which are a lot more complex.)
Thisisn't somethingyou can %I with profile editing-you'll simply have
lo edit those images in which it occurs. So if a color is rendered incorrectly
in a singleimage, check other images that contain similar colors captured
from differentobjects to determine whetherit's a profile problem that you
can fixby profile editing or a metamerism issue that you'll have to fix by
image editing.
With field cameras, you have no way of knowing the light source for
a given exposure. If you shoot TIFF or 1PEG. gray-balancing (or whitebalancing) is the key to keeping the variations in lighting under reasonable control. Ifyou shoot raw. be aware that vou're profiling the entire raw
conversion,includingwhite balancesetting.You also have the problem of
what to use as a reference when editing the profile.

Scanner profile editing

Thefirststep in editinga
profile is always to
figure out which part of
the profileyou want to
edit. Input profiles are
almost always one-way,
so we edit the RGB to
LAB side.We usually
start with the relative
colorimetric intent,
then apply the edits
to the other intents
when saving.

'Ihese moyappmr to be big comtions,

but aghnceat the numbers in theValues
paletteshows that.we'reacruallymaking
pray small change-t-this tool is designed

246

Real World Color Management, Second Edition

We find that the best approach is to create a reference image by combining a wide variety of image types shot under varying conditions, and
aim for aprofilethat helpsthemall more or less equally. You can include a
few images that contain a known target such as theMacbeth ColorChecker, but resist the temptation to edit the profile so that it reproduces the
color checker perfectlywithout checking what it does to other images.
Figure 9-14 shows a typical field camera profile editing session. In this
example, we used Koda!& newly revamped Custom Color KC, which allows us to use Photoshop as a profiling editing tool, but the process is the
same for any profile editor, and it involves several steps.
r We set the raw converter to deliver linear gamma images.
We make a reference image by ganging several different captures into
a single document that we'll use to judge the edits.
We open the reference image and the profile in the profile editor.
We make the edits to the profile, using the reference image as a
guide.
We save the edited profile.

Digital camera
profile editing

Ourpnt task is to
configure the mu!
converter to deliver the
images correctly. Here, we
set Canonk Fileviewer
utility to deliuer lineargamma images to
Photoshop.

Chapngum 944

Next, we create a reference

image in Photoshop by
assemblinga selection of
different image types.Al1
the images we used here
were shot with the camera
whoseprofile we're
editing. The images are
linear gamma with no
profile assigned, so we
expect them to lookdark
and ugly at thb stage.

Next, we open the reference

image and the profile we
want to edit in the profile
editol:Here weseehdak's
Custom Color Import
Plug-in dialog bax. It lets
us choose the profile, the
part of the profile we want
to edit, and a reference
image.When we click
OK, the image opens in
Photoshop with the profile
applied, ready for editing.

O: baluatlng and Editing ProRln

Digital camera pro6le editing, continued

u7

948

Real World Color M a n a m , Second Edition

R g v n 944

Digital carnerapm6leedi~continued

The reference image opens in Photoshop with the camera profile already applied. (The fact that the
profile menu indicates that the image is in ProPhoto RGB is because Custom Color ICC converts the
image to Photoshop's RGBWorking Space and uses it to track all the edits.)
Ourfirst set of edits uses Photoshop'sCurves. Westart by opening up the three-quarter-tones to increase
shadow derail, then we turn to the individual channel curves to tweak thegray balance.
In doing so, we pay particular attention to the gray patches in the Color Checker SG, while lookingat the
effects of the edits on the other images. We adjust the individual red, green, and blue curves to make the
values in each channel the same for each of thegraypatchesfrom light to dark-the closer the three
values are to matching each other;the closer we are to a neutral gray
In this case, most of the work consists offine-tuning the upper portion of the blue curve, but we also make
small tweaks to the red andgreen c u m to improve midtone neutrals.

Chaplet 9: h l w t l n g and Wltlng Protila

Ugurc 944

s9

Digital camera profile editing, continued

Ourfinal set of edits make small tweaks to individual color ranges usingPhotoshop'sHue/Saturationtool.
As with thegmy balance moves.we use the Color Checker SG as aguide, this time concenmting on the
primary color patches while using the other images as a reality check.When we're done, we choose
ExporbCustom Color ProfileErport to save the edited profile.

Calibrating Camera Raw

IfyouuseAdobe Camera Raw as your preferred raw converter, you'velikely
already realizedthat youcan't use custom camera profiles, because the profiles are hardwired into the CarneraRaw plug-in.The goodnews here is that
most of the heavy lifting has already been done for you. But the built-in
profiles may not match your specific camera exactly, so Thomas Knoll was
kind enough to includeCalibrate conaols in the Camera Raw plug-in. There
are manyways to use these, but we'll offer a relativelysimple procedurethat
we've found effective.

950

Real World Color Managemenl, Second With

To carry out the calibration, you need a physical target-we use the
old 24-patch Macbeth ColorChecker, but the newer ColorChecker SG
includes the original 24 patches so you can use it instead if you wish-and
an accurate RGB version of the ColorChecker.
RGB Color Checker. If you're a true fanatic, you can measure the ColorChecker yourself, create a Lab file of the color patches in Photoshop, then
convert it to RGB in Photoshop using Absolute Colorimetric rendering. In
theory, you should be able to carry out this exercise using any of the four
RGB spaces-sRGB, ColorMatch RGB, Adobe RGB (19981, or ProPhoto
RGB-that Camera Raw supports. In practice, we find that it's highly desirable if not essential to build the calibration in ProPhoto RGB, perhaps
because Camera Raw uses a linear-gamma space with ProPhoto RGB primaries to do much of its processing.
If you're less fanatical or simply lazy, you can download a ProPhoto
RGB version of the ColorChecker, with the RGB values entered on each
patch to simpliimatching, fromwww.coloremedies.comlrealworldcolorl
downloads.html. It's shown in Figure 9-15.

Shwtlng the brgel.To obtain the best results,you need to ensure that the

target you shoot is evenly lit, and that the shot is correctly exposed. An
ideal exposure should produce values around level 238-243 for the white
patch with no Exposure adjustments in Camera Raw.

Chapter 9: Evaluatingand Editing P d k r

Setup lor calibrate. Arrange the windows on screen so that you can see

both the ProPhoto RGB ColorChecker and the target capture in the Camera Raw window. You'll use the ProPhoto RGB ColorChecker as the aim
point when you adjust the Calibrate controls-see Figure 9-16.
nguw 9-16

Setup for Calibrate

---

Tonal adlustmcnts. Before you can adjust the Calibrate controls, you'll

need to make tonal adjustments using the Exposure, Shadows, Brightness, and Contrast sliders to make the gray patches approximatelymatch
the values in the reference ProPhoto RGB image.

The four aforementioned controls work together to define a 5-point

curve, as follows:

-re

Exposure sets the white point, and Shadows sets the black point.

Brightness adjusts the midtone without affectingblack or white, like

the gray slider in Photoshop's Levels command.

Contrast applies an S-curve around the midpoint set by Brightness,

again without affecting black or white-see Figure Q-17.

9-17

Tonal adjustments

As you make the

adjustments, monitor
the multingRGB lxJues
~(singcrunem
w s
RGB readout, which
shows the RG% vvrkres
underneath the cursor.
RGB readout

Chapter 9: Cwlurtlng and Editing Profllcr

953

Don't try to aim for perfection. The white patch will always end up
darker than in the reference file. You should, however, be able to get the
remaining five gray patches to match the reference within three or four
levels. Once you've done so, white-balance the image by clicking on the
second-to-lightest gray patch. After white-balancing, you may need to go
hack and tweak the Brightness and C o n m t sliders slightly.
Tip: Use the Anow K q s to Adjust Valun.Thisexercise becomes a great deal
easier if you keep the cursor on the patch you're evaluating so that you can
see the RGB values in Camera Raw's readout. UseTab to move to the next
field and Shift-Tab to move to the previous one, and use the Up and Down
m o w keys to change thevalues in each field.

Once you've set the tonal behavior, you can continue on to setting the
Calibrate controls, but before you do so, check the blue. green. and red
patches on the thii row (these are the patches where you'll focus your attention). In most cases, the red value in the red patch will be higher than
the one in the reference, and the blue value in the blue patch will be lower
than in the reference. Use the Saturation slider to get the best compromise
in terms of matching the red value in the red patch, the green value in the
green patch, and the blue value in the blue patch to those in the reference
image before proceeding to the Calibrate tab.
Calibnlcadludrnmh.The Calibrateadjustmentsletyou tweak the hueand
saturation of the primaries for the camera's built-in Camera Raw profiles.
If you're concerned that the adjustments you've made so far lock you into
always usingthese tonal moves, or that the choice of ProPholo RGR means
that the calibration is only valid for ProPhoto RGB, put your mind at rest
The Calibrate controls let you adjust the relative hue and saturation of
the camera RGB primaries, and these adjustments remain valid for any
tonal settings and any output space. The adjustments you've made so far
simply massage the target capture to make it easier to use the ProPhoto
RGB reference image as a guide.
u
be able to make the edits in any order, but in pracIn theory. ) ~ should
tice we find that things go more smoothly when we adhem to the following
order: shadow tint; green hue and saturation: blue hue and saturation: red
hue and saturation. For each color's controls, the hue slider adjusts the
other two colors unequally to affect the hue. and the sahlration slider adjusts the other two colors equally to afftvt the saturation.

254

Real World Color Management, kcond Edition

Start by sampling the black patch on the target to check the neutrality
of the shadows. There should be no more than one level of difference
between the three channels on the black patch. If the difference is
w a t e r than that, adjust the shadow tint slider to get the black patch
as neutral as possible.
Continue with the green patch. Adjust the green saturation slider to
get approximately the right amountofred and blue relative to green in
thegreen patch, and adjust thegreen hue to fine-tune the proportions
of red and blue relative to each other.
b

Next, move to the blue patch-you may find that the blue value for the
blue patch has moved closer to the target value due to the green tweaks.
Use the blue saturation slider to balance red and green relative to blue,
and the blue hue slider to balance red and green relative to each other.

Finally, sample the red patch. Use the red saturation slider to adjust
blueand green relative to red, and the red hue slider to adjust blue and
green relative to each other.

Remember that you're adjusting color relationships here. It's much

more important to get the right relationship between red, green, and blue
than it is to produce a close numerical match, though with skill and patience it's possible to getveryclose tothenominalvalues. So, forexample.
if the red patch has a red Value of 132. as compared to the reference red
value of 122, take the target values for green and blue, and multiply them
by 1321122, or 1.1. So instead of aiming for a green value of 58, you'd aim
for 64. and instead of a blue value of 46, you'd aim for 51. You'll also notice
that each adjustment in the Calibrate tab affects all the others adjustments in the Calibrate tab. A second iteration of adjustments will often
help yougetmuch closer. Figure9-18showstypicalcalibrate adjustments
applied to the target capture, in this case from a Canon EOS I D Mk 11.
Spsclal c n a . The calibration you build using the Calibrate tab is usu-

ally good for any light source. However, some cameras have an extremely
weak blue response under tungsten lighting, so you may have to make a
separate calibration for tungsten. And if you have the misfortune of havingto shoot under fluorescent lighting, theveryspiky natureof the typical
fluorescent spectrum may require you to special-case it too.

Chapter 9: baiuatlng and Editing P r d k ,

955

Calibrate adjustments
The completed Calibrate
adjustments produce a
close visual match
between the captured
target and the reference.
Once we're done, we
choose Save Settings
Subset, then saue only
the Calibrate adjustments, so that we can
apply them to images
independentlyfrom
other ndjustments.

Onceyou'redone,you cansave the Calibrate settingsasaSettingsSubset,

and apply them to images, or incorporate them into the Camera Default.

Output Profiles
We never edit monitor profiles, and we try to edit input profiles as little as
possible, but we almost always wind up editing output profiles. Output
devices tend to be less linear than capture devices or displays, they often
use more colorants, and their profiles must be bidirectionalso that we can
use them for proofing as well as for final conversion-so they're by far the
most complex type of profile.

P56

Real World Color Management, Second Edition

That said, the techniques we use for evaluating output profiles are very
similar to those we use for input profiles. Unless our evaluations reveal gross
flaws, we tend to be cautious about editing the LAB-to-device (BtoA)side of
the profile, except possibly to make slight changes to the perceptual rendering intent or to make a very specialized profle. But we find we often have to
make some edits to the device-to-LAB side to improve our soft proofs.
Before doing anything else, we always recommend looking at a synthetic target such as the Granger Rainbow or the RGB Explorer (see Figure
9-7),with the profile assigned to it in the case of RGB profiles, or converted
to the profile in the case of CMYK profiles.
You'll probably see posterization somewherebear in mind that real
images rarely contain the kinds of transitions found in these targets--so
that shouldn't be a major concern.But if you see huge discontinuities,areas
of missing color, or the wrong color entirely, stop. The problem could lie in
the printer calibration or media settings, in your measurements, or in the
parametersyou set whiie buildingthe profile. So go backand recheck a l l the
steps that went into building the profle, becauseyou won't get good results
from this one. Figure 9-19 shows typical appearancesfor the Granger Rainbow converted to output profiles. If what you see looks much worse than
any of these, you need to back up and figure out where you went wrong.
R~UW
9-19

Granger Rainbow
through proIYes
It's common toseeslight
color-banding in some
regions of the rainbow
when you convert to
an output profile, and
equally common to see
some color regions grow
while others shrink.

Ifan output profile looks significantly worse than the exampi

something h a almost certainly gone wrong!

Objective Tests for Output Profiles

We do objective tests on output profiles not to provide an objective benchmark of profile accuracy, but to help us understand the profile's behavior,
and to identify problemareasthat may respond to profile editing-seethe
sidebar, "Objective Objectives," earlier in this chapter. (Bruce also does
them for fun, but even he thinks that's weird.)

Chapln 9: Evaluating and W i n g ProRla

257

The fundamental principle for objective evaluations is to compare the

device's known behavior with the behaviorthe profilepredicts.We do that
by comparing theactual measurements of the target-the color we know
the device produces in response to a set of numbers--with the LAB values the profile predicts. We obtain the predicted LAB values by taking the
target file, assigning the device profile to it, and then converting to LAB
with absolute colorimetric rendering. This lets us see what's happening
in the colorimetric AtoBl (device-to-PCS)table.
But since output profiles are bidirectional, we go Further. We take the
predicted LAB values, prinf them through the profile, measure the print,
and compare the measurements to the predicted LAB values. This lets
us see what's happening in the colorimetric BtoAl (PCS-to-device)table.
We usually see higher delta-e values on this side of the profile because it's
computed, whereas the AtoB table is based on actual measurements.
Lastly, we "close the loop" by comparing the original measurements
with those from the predicted LAB image printed through the profile.
Usually the errors in the AtoB and BtoA tables are in opposite directions.
producing a laver delta-e between the two measurement files than between either measurement file and the values predicted by the profile. If
you get higher delta-e values in this comparison than in the others, the
errors are moving in the same direction on both sides of the profile--this
may be a sign of bad measurements, so check them. If you're quite sure
they're OK, you may want to edit both sides of the profile identically,
though this is quite a rare situation.
As with input profiles, you have the choice of using Free toolsanddoing
alot ofwork, or spendingsomemoney andusing ChromixColorThinkPro
to make the comparisons much more easilv. And as with input profiles.
we'll cover both methods here.
You can use ColorLab to convert the RGB or CMYK device values to
LAB, but we prefer to do that in Photoshop, since it lets us choose different
CMMs easily. The CMM can make a big difference to the results--if your
profiling packageisdesigned to preferaparticular CMM, you canusethis
same test to determine whether the preferred CMM actually produces
better results, and if so, how much better.
Figure 9-20 shows the steps needed to evaluate the AtoB colorimetric
table, the BtoA colorimetric table, and the roundtrip result, using Logo
ColorLab to convert text to pixels and vice versa.

958

Rcal World Color Manag.n#nt, Second Edition

Figure 940

Step 1: Obtain predictedLAB values.

Output pro&
objective testing
In ColorLab,choose
Open from the File
menu, and open the
target descriptionfile.
You'll getsomething that
looks like the
image at the right.
Choose Spot Colors
from the Conversion
submenu of the Filter
menu, then click Enable
in thedialoghappaus.

Choose Expowfrom the

File menu and saw the
fileasa TIFE

Open the TIFFin Photoshop, assign the profile

beingeualuated, then
convert the TIFF
to LAB with absolute
colorimetric rendering
to obtain the predicted
LAB values. (The
illustrationshows an
RGB output profile
but the procedure is
identical for CMYK)

In ColorLab,open the
original measurements
from which thepmflle
was built, the roundzrip
measurements you just
made, and the TIFF
image mnlainingthe
predicted LAB curlues.

To ensure that all threejilesare available for comparison in ColorLab,choose Spot Colorsfrom the
Conversion submenu of ColorLab's Filter menu, then click Enable in the dialog box that appears.
This tells ColorLab that each patch is a spot color and lets itgenerate a color-by-color comparison.

260

Real World Color Management, Second Milion

Ugure 9-90
Output
testing,

Step 4:Make the comparisons.

Select the original measurement
file, choose Compare withfrom
ColorLab'sTools menu, then select
the oredicted Labs file h m the
Select Document dialog box.

. .

ColorLabgenerates a
comparison of the
measured LAB ualues
with the ones predicted
by the target. This
provides a gauge of the
accuracy of the AtoB
(device-to-PCS)side of
the profile.
Repeat the process to
compare the predicted
LAB ualues with
the roundtrip LAB
measurements. This
gives you an idea of the
accuracy of the BtoA
(PCS-to-device)side of
the profile.You'llalmost
always get higher errors
on this side of the
profile.

Close the loop by

comparing the original
and roundnip LAB
measurements.This
gives you an idea of the
ouemll accuracy of the
profile. The errors are
usually lower than
those in the previous
comparison.

numbers. ColorLabdoes
a gocd job of showing
you the uisual d i & m m
between the samples
(ifyou havea good
monitor profile, as
mllas the&-e values
A deltu-e Of3is much
less signijIcant in a color
that's almost black than
it is in one that's bright
blue, for a m p l e ,
Also, lookfor areas

where thepmfileseems
to have problems with a
spec% range of colors.
They w n help you spot
possible candidatesfor
p r o p editing.

Chaptar 9: Errluetlngand Ldnlng Pronla

961

Tip: If Colorhb b n ' l Work... ColorLab is unsupported software, and as

such, it usually has some bugs. In the current Mac OS X version as of this
writing, ColorLab 2.8, the Compare with feature seems to be broken. We
have no idea when or if this will get fixed, so we offer an alternative. The
MeasureTool component of GretagMabeth's ProfileMaker 4.1.5, which is
still availablefor download, can do the same comparisons in demo mode,
unlike MeasureTool5.0, which requires a dongle to do so.

Alternatively,you canmake the comparison agood deal more easily using ColoffbinkPro.You'11need to open the profiling target in Photoshop
and assign the printer profile, then convert it to Lab, to obtain a printable
target containingthepredictedLab values and to print that target through
the printer profile. But ColorThink Pro can show you the predicted Lab
values, so you don't need to wrestle with extracting the Lab data from
Photoshop.
If you're already a ColorThink user, you'll find that ColorThink Pro is a
very worthwhile upgrade with a wealth of options that simplify both numeric and visual color comparisons. Figure 9-21 shows the comparison
between the three sets of measurements.
Figure 9-9l

ColorThink Pro
profile evaluation

363

Real World Color Managcmcnt, Second Wition

ColorThink Pro
We've long relied on Chromix
ColorThink as an indispensable
tool for visualizing color-all the
3D gamut plots in this book were
produced using ColorThink, for
example. While it's no slouch at
providing numerical feedback,
what sets ColorThink apart is the
way it lets you model and analyze
color issues visually.
The new ColorThink Pro builds
on the solid foundation provided
by the original, and adds extreme-

Rsw 9-PL
ColorThink Pm
metamerism test

nglre 9-93
Coloffhhk Pro
Color Worksheet

ly usefulnewcapabilities,ofwhich
we can show only two here.
Figure 9-22 shows Colorthink
Pro'smetamerismtest, whichshows
youhowspectmllymeasuredcolors
will appear under Merent lighting.
It can help you decide whether or
not your printer needs illuminantspecificproliles.
Another new feature in ColorThink Pro is the Color Worksheet,
which allows you to simulate and
analyze all sons of workflows

from the simple to the very complex. Figure 9-23 shows a Color
Worksheet that tracks the conversions from an RGB to a CMYK
destination to an RGB proof, and
lets you track exactly what happens to the color at each stage of
the process.
If you often have the need to
analyzecolor problems,and especially if you like to think visually,
we can't recommend ColorThink
Pm too strongly.

Chapter 9: Evaluating and Editing P d i h

463

Lfoneprofileor CMM seems to produce more accurate resultsthananother, double-check by looking at the Granger Rainbow-accuracy often
comes at the expense of smoothness. Don't make any hard-and-fast decisions yet-to fully evaluate the profile, you still have some work to do.

Subjective Tests for Output Profiles

Objective rests give you a good idea of a profile's overall accuracy-how
much slop there is in the system-and the specific color areas where it
has eouble. They're useful in determining the profile's ability to reproduce spot colors accurately, and to simulate other devices' behavior as
a proofer. But they tell you very little about the profile's ability to print
images. To do that, you need-you guessed it-to print images!
When you evaluate a profile's image-handling capabilities, you need to
use avariety of image types with shadow and highlight detail, sanuated colors, neutrals, skin tones, and pastels.The images must be in either IAB or a
wen-defined device-independent RGB space--suchas Adobe RGB (19%) or
ColorMatch RGB-to make sure that your evaluation isn't compromised by
shortcomings in the input profile. (You can't use an original print or transparency to make the comparison, because the profiled print would then be
the result of the input profile plus the printer profile.) Print your test images
using both relative colorimetric and perceptual renderings.
Youneedto cany out two comparisons for each renderingintent. Compare the printed image, under suitable lighting, with the original image
you were trying to reproduce (if you're evaluating a CMYK output profile,
look at the original RGB image) on your calibrated and profiled monitor.
This comparison looks at the colorimetric and perceptual BtoA (PCS-todevice) tables, which control the way the output device renders color.
Then convert the images to the profile you're evaluating, and compare
its appearance on your calibrated and profiled monitor-this looks at the
colorimetric AtoB (device-to-PCStables), which control the soft proof of
the printed image on your monitor.
If you're happy with both, you're done, and you can skip the rest of
this chapter. I t 3 more likely, though, that you'll find flaws with one or the
other.
The two key questions you need to ask are, in this order:
Does the screen display of the image converted to the ptofile match
the print?

464

R u l World Color Managemrnt, Second Edition

If the answer is yes, we recommend that you don't try to edit the profile.
but instead soft-proof images and make any necessary optimizations
as edits to individual images (see "Soft-ProofingBasics" in Chapter 10.
Color Management Workflow). It's a great deal easier to achieve predictable results editing images than it is editing profiles.
If the answer isno, we suggest that you edit the AtoB tables to improve
the soft-proof before you even think about editing the BtoA tables.
Most profile editors work by displaying your edits on an image thatas
displayed through the profile, so fixing the display side of the profile is
pretty much essential before youcan turn your attention to theoutput
side.
t Does the print provide a reasonable rendition of the original image?

IF the answer is yes, fix the AtoB tables if necessary or just use the
profileas is. Lfit'sno, do you seeproblems with both rendering in tents.
or just with one? If you only see problems with perceptual rendering, and your profiling package allows you to change the trade-offs
between hue, saturation, and lightness in the perceptual intent, you
may want to generate a new profile with different perceptual rendering parameters rather than editing. If you see real problems with the
relative colorimetric rendering too, it's a candidate for editing.

Editing Output Profiles

Output profile editing isn't for the faint of heart You're standing ankle-

deep in a bog, you've decided that the map needs fixing, and all you have
to fix it with is your eyeballs and a laundrymarker.With care, practice,and
skill, it's possible to make profiles more useful by editing them, but it's a
great deal easier to screw them up beyond recognition.
With that cautionary note in mind, here are the types of edits that we
thinkare rational to attempt:
r Globally editing all the AtoB tables to improve soft-proofing. We m i cally make smallchanges to white point. overallcontrast,and saturation
(see Figure 9-24).
Making small changes to selective colors in all the AtoB tables to
improve soft-proofing (see Figure 9-25).

?lpBe944

Edith@the AtoB
white point

The papar-white
simulation of images
renderad mt h e m with
absolutemlorimaaic
rendering is o p n too
drastic tn be visunlly
useful.We addms this
by slightly reducing
the color cast and
i n c m i n g the tightness.
Notethatthkod~t
nlso*absolute
colorimetric rendering
to M - m ~ p m f e r r s
@st both.

Selective colorin
the AtoB tables

Here, we made small

changes to the softproofing of a small
range ofyellow-greens
to make the sofrproof
match the printed piece
more closely

266

Real World Cdor Management, a n d Edition

Making small changes to the BtoAO (perceptual)table only, to improve

the printed output ofout-of-gamut colors, but only after you've finished
editingthe AtoB tables: unlike the first three types of edit, this one achally changes the printed output.You maybe better offgeneratinga new
profile with different perceptual mapping-if your profiling package
offers these controls, make a new profile rather than editing this one.
You can also edit the colorimetricBtoA table to try to reduce themaximum errors, or to improve the reproduction of specific colors, but be
warned-it's like trying to adjust a cat's cradle. You pull on one part of the
color range, and another part moves. It's possible to mduce the maximum
error, or improve the reproduaion of specific colors that concern you, but
only at the cost of increasing the overall average error. Within limits, this
may be a perfectly sensible thing to do.
if your goal is to reduce maximum error, consider where in the color
range it's happening. A IAB delta-e of 10 is a great deal more obvious in
a light blue than it is in a light yellow. for example. So look at where the
errors are showing up, and decide if they're visually objectionable enough
to justify compromising the color in other areas (see Figure 9-26).

Edit,- Reprofile, or Update?

-

When you're new to profile editing, you'll be tempted to edit the profile
to lix problems that are really better addressed by either adjusting the
device's behavior (perhaps by linearizing) and reprofiling, or reprofiling
with different profile parameters. Going all the way back to the beginning
and starting over may seem like a lot more work than editing the profile,
but in the long run it's usually the easiest course.
Remember that a profile is the result of a lengthy process that'ssubject
to all the variables we discussed in Chapter 5, Memrrremenr. Cnlibmtion,
and Process Co~mol.
Small changes in the variables early in the process
tend to get amplified in the profile creation process.
If your profile has problems that are big enough to make you contemplate editing them, takea moment to reflect on all the steps that went before. Are both thedeviceand theinstrumentcorrectly calibrated?Did you
let the printer target stabilize?Are all your measurements correct?Did you
set the separation parameters for your CMYK profile optimally?All these
factors can have a surprisingly large impact on the resultingprofile.

Chapter 9: Lvaluallngand Wiling Profiles

267

Reducing maximum
delta-e errors

This mundbip
comparison led us to
edit the quns to fly to
improvethe visually
obvious delta-e on
the cyan patches.

Afrerediting we
compared the new
roundtrip ualues. Wedid
im&& improw the
cyans, butonly at the
apeme ofslighfly
higher awrage and
maximum delta-e
ualues. Newrtheless,
we consirier the @ r t
succesful because the
incrense in errors is very
small, and rhe worst
delta-e has s h i m toa
colorregrbnwhereitk
ulually much less
notieedble.
If the device has changed in a way that defies recalibration, or calibration isn't an option with your particular device, andthe result is a shift in
tone reproduction, you may be able to update the profile. Some packages
offer the ability to reprint a linearization target, and update an existing
profile. If the magenta in a color laser printer is printing heavier today,
for example, updating may be adequate. But if you've changed to a new
magenta inkset on press or in an inkjet, you'll most likely have to make
a new profile. Subtle changes in paper stock can often be addressed by

968

Real World Color Management, Second Edition

profile updating. The way to find out is to give it a tty-updating a profile is
quicker and easier than either editing the profile or making a new one.

Profiling Is Iterative
Getting good profiles requires rigorous attention to detail. a reasonably
stable device, a reasonably accurate instmment, and a reasonably good
profiling package. Getting great profiles takes a good deal more work. We
often go through multiple iterations of calibrating the device, measuring.
profiling, evaluating, editing, evaluating again, going back and adjusting
the device, and then repeating the whole process. Eventually,you'll reach
the point of diminishing returns no matter how picky and obsessive you
are, but just where that point occurs is a subjective call that only you can
make.

Where Do We Go from Here?

We've just about run the analon of profileas map into the ground, but it
still has alittle lifeleftin it. In thischapter, we've shown you howto determine the accuracy of the map, and how to improve at least some aspects
of its accuracy. In the next part, we'll talk about how to use the maps to
get to where you need to go.

Applications
rnd Workflow

This page lefi blank. ..

Color
Where the Rubber
Meets the Road
We've been straining the analogy of profile as map throughout this book,
so let's give it one more good hefty yank. Afew of us appreciate profiles,or
maps, as works in their own right, but for most of us, they're simply tools.
and maps aren't all that useful if we never go anywhere.
At this point, you probably know slightly more than normal people
need to know about profiles. In this chapter, we look at where the rubber
meets the road-how you use profiles with your various devices and applications to drive your color where you want it to go.
Wegive you the holistic overview of how your color management components interact, and describe the major principles that govern how you
configure your applications to let color management make your color
flow smoothly and predictably-from capture, through editing, to document assembly, to proofing, and to final output (or outputs).
You may be tempted to just skip ahead and delve into the chapters that
describe your applications. Instead, we beg you to take the time to digest the material we present here, because while the application-specific
material will tell you which buttons to push, vendors have a tendency to
move and rename those buttons each time they revise the applications. If
you understand the bigger picture we paint here, you'll know what kinds
of buttons to look for, and you'll understand how, why, and when to press
them to make your applications handle color the way you want them to.

479

Real World Color Management, Second Ldlilon

What Is Color Management Workflow?

-

"1Vorkj7ouP is one of those slippery words likc "quality," "art." or "postmodernism," where every expert claims to know what he's talking about.
but no two can agree on its definition. We won't try to provide a comprehensive definition of workflow. Instead, we'll just tell you what rue mean
when we talk about color management workflow.
In a nutshell. color management workflow is the art and science of
defining what colors the numbers in your documents represent, then
preserving or controlling those colors as the work flows from capture.
through editing, to output. In thischapterwe'll lookat how color management worldlow applies to three different areas:
F

The flow of documents or objects within a program. For each application, how do we configure it for color management, and what procedures should we use to open and save documents, import objects
into the program, and copvand paste colors between windows in the
program?

t The flow of documents between programs. As we move color docu-

ments from program to program, or even from person to person, how

do we keep colors looking right, and when and how do we convert to
different forms for output?
The flow of materials into or out of a color-managed environment.
How do we bring in documents and objects from non-color-managed
devices or applications, and integrate them with our color management documents? How can we get maximum benefit out of our color
management efforts before sending our jobs off to environments
where other non-color-managed steps may happen?
We color-manage two types of things: documents, and objectsin documents. One program's document may be another's object. For example.
a docrrmentin Photoshop consists of a raster image, but in Illustrator,
QuarWress, or InDesign, this very same raster image may be just one
object among many in an assembled document. And each one of these
objects may have adifferentprofile assigned to it [see Figure 10-1).

mure 10-1

Documents and objects

assumed
profile used in
Illustrator

digital camera profile

profilefrom
designer's monitor

Ultimately,all these elements will windup being converted to the same

output space.There are reallyonly two variables that differentiatethevery
large number of possible color management workflows:
b

When the conversions occur

How the color meaning is conveyed

Real World C o b Managcnrml, krond Mltkn

As we toldyou waybackin Chapter 3, these are the only two things that
color management systems do. Applications may dress color management up in fancier clothing, but all color management features ultimately
break down into some c o m b i i o n of these two basics.
But timing conversions and conveying color meaning raise different
types of issues.When to convert is a strategic decision that is in considerable part dictated bythe type of work you do, while the way color meaning
is conveyed is largely a tactical decision that's dictated by the capabiities
of your applications. So we'll look at the bigger, strategic question first.
But before doing so, we need to address one other issue.

Dlsplay Conwenions
It may not be immediately obvious, but almost every time you display an
image in a color-managed application,there's a conversion going on from
the document's space to your monitor's s p a c e t h e only exceptionis if the
document is already in monitor RGB. If this conversion didn't happen, all
your color would be displayed inaccurately.
From a workflow standpoint, however, the display conversionis out of
the loop. Color-managed applications apply the display conversion on
the fly to the data that gets sent to the video card-it never touches the
documents themselves.
The beauty of this approach is that it allows color management to account for the quirks of each individual's display, transparently and automatically, without affecting any of the actual data being pushed through
the workflow (see Figure 10-2).
Clgufe 10-P

Monitor compensation

display conversion

Color-managed applicationsperfonn a conuersion on the data sent to the

monito~fromthe document's source profile (or profiles) to the display profile.
Thi.~conversion happens outside the main color reproduction loop and never
touches the document data itself;so the same document can be displayed
correctly on many different monitors without convertingthe source data.

When to Convert
Hack in the 1990s, worMow was a hot topic that allowed consultants to
earn large sums of money, mostly by telling their clients what they already
knew and wanted to hear. But one useful distinction that dates from those
halcyon days is the distinction between a lore-bindingandan early-hiding
wolMow:
W

Early-binding workflow converts all the color into final output space
as soon as possible-traditional prepress workflows that took output
CMYK right from the scanner are a classic example.
Late-binding workflow delays the conversion to final output as long
as possible. In-RIP separations in the platesetter's or imagesetter's RIP
represent an extreme late-binding workflow.

However, late binding versos early binding isn't a binary, eitherlor

choice. Rather, late binding and early binding define two ends of a continuum. Most color management worMows fall somewhere between the
two extreme examples cited above.

Early Binding Advantages

The biggest advantage nf early-brnding workflows is simplicity. Early in
the workflow,everything gets converted to a single output space. As long
as they stay in this closed-loop environment, there's no ambiguity about
the color meaning of the devicevalues in each document and object.
A second advantage is that it's easy to Introduce color management
graduallvmto existingeady-bindingprepress worMows.The only changes that color management brings are that operators see accurate color
on their monitors instead of being forced to rely strictly on the numbers,
and it becomes possible to pull intermediate camps and preproofs from
inexpensive inkjet printers. The basic workflow doesn't change at all.
Athird early-bindingadvantage is that it prevents designers from using
non-reproducible colors in their designs. If everyone works in final output
space, it's impossible by definition to create out-of-gamut colors.

976

Real World Color Managrmcnf, Second Edlflon

Early Binding Disadvantages

'The huge disadvantage of early-binding workflows is their inflexibility.
Everything in the workflow is targeted to a single output process--all the
color is squeezed into the output gamut and optimized for the output's
tonal response.So early bindingis practical in situations where the output
process is always the same-a daily newspaper or monthly magazine, for
example--but it's pretty useless for a freelance designerwho may have to
work on jobs where the output process isn't even known yet.
A second disadvantage when the output is CMYK is that many creative effects, such as a large number ofAdobe Photoshop filters, are only
available in RGB. And even those available in CMYK often make terrible
assumptions about CMYK behavior, resulting in poorer quality effects.
A third disadvantage is that output files are usually CMYK, while most
capture devices produce smaller RGB files [they only have three channels,
as opposed to four-channel CMYK),so early binding means larger files.
and slower performance in opening, saving, or copyingfiles across a network. While workstation storage is reasonably cheap, an extra33 percent
on a network, on the internet, or in asset storage, is not insignificant.

Late Binding Advantages

The great advantage of late-binding workflows is flexibility. Maintaining
the colorgamut of the originals means that the work 1s easily repurposed
for conditions as different as sheetfed printing, newsprint, or the Web.
Late binding also allows you to do a great deal of useful work before
the final output conditions have been determined, and it's well suited to
situations where the color is assembled from many differentsources.

Late Binding Disadvantages

The major disadvantage of late binding is its inherent complexity, not
in the sense that it's difficult, but in the sense that there are many more
places for things to go wrong. A single incorrect application setting can
week one or more elements in the job.
A second disadvantage with extremely late-binding workflows such as
in-RIP separations is that you don't have a chance to evaluate the final
output datauntilvery late in thegame. inearly-binding workflows, particularly those based on an established tmditional prepress workflow, color
management can be reasonably called a luxury, albeit a very useful one.
But in a late-binding workflow, it's a mission-critical necessity.

Avoidiig Extremes
The late bindinglearly binding distinction is auseful one to keep in mind,
but it's important to remember that the terms represent two extremes.
Most real color-managed workflows fall somewhere in the continuum
that stretches between the two extremes.
It's also tempting to think of early binding as an all-CMYK workflow
and late binding as an all-RGB one, and in a great many cases, you'd be
right. But bear in mind that while all capture devices capture RGB (even
the big prepress drum scanners that produce CMYK output scan in RGB,
then conventhe data), not all outputs are CMYK, so the distinction really
does revolve around when you commit all your color to output space,
which may or may not be CMYK.
There's one more important issue to factor into this equation. Most
capture spaces and most output spaces aren't that well suited to editing
color, so many color management workflows use an intermediate space
between capture and output spaces to avoid the worst extremes of early
or late binding, and to simplify the workflow.

Intermediate Spaces
If color reproduction were simply about reproducing original imagery or
artwork as exactly as possible, it might make sense to keep all our color
imagery in the RGB space in which it was captured until it was time to
convert to final output space, as shown in Figure 10-3.
nsura roa
Scan to print

The CMS converts the

source color to monitor

color for display only.

\'?

The CMS uses thescanner

profile to define the source
colors.

The CMS converts the

colorfrom thescanner
profile to the print profile.

278

Rral World Color Manaecmcnt, Second Edition

But.inthe real world, things are rarely that simple. Our originals almost
always have a wider gamut than our output, and even the best profile
with the best perceptual rendering is unlikely to do equal justice to all
images-we typically need to make different compromises to a picture
of a black cat in a coal cellar and one of apolar hear in the snow, to cite a
couple of extreme examples. So we almost always need to edit our images.
even if they were captured perfectly.
h p u t device spaces describe the behavior of capture devices. and as a
result they usually have two properties that make them less than ideal as
image-editing spaces:
Input spaces are rarely gray-balanced.
Input spaces are hardly ever perceptually uniform.
In a gray-balanced space, equal values of R, G, and B always produce
a neutral gray, which simplifies one of the most powerful techniques for
correcting color: pull the nelrnals into place and the rest of the color follows. In a non-gray-balanced capture space, this is a lot more difficult to
accomplish.
In a perceptually uniform space, the same incremental change in the
numbers produces the samedwee of change in theappearance, no matter where in thecolorgamut and tonal range it takes place. Capture spaces
generally lack this propew, which again makes them more difficult to
use for editing.

Dcrlce-independent RGB. One solution, embodled in Adobe Systems'

applications but usable by others, is to use a device-independent, graybalanced. perceptually uniform spacesuch as Adobe RGB (1998)for editing.This approach has proven sufficientlypopular to spawn aplethora of
editingspaces--oftennamed fortheir developers--and debating the merits of each isdecidedly outside thescope of this book. Instead, we'll simply
say that the main criterion in choosingan editingspace is its gamut.
Bigger isn't necessarily better. The trade-off is between finding an ediLing space that won't clip colors in either your capture or your output, and
finding an editing space that doesn't waste huge numbersofbitsdescribing colors that you can't capture, display, print, or in some cases, even
see. In practice, it's pretty much impossible to find a space that contains

all your colors yet doesn't waste bits, so you simply have to pick the best
trade-off for your particular purposes.

CIEUB. Why not simply use LAB as the intermediate space? It is, after
all, gray-balanced and reasonably perceptually uniform. LAB-based
workflow6 are quite popular in Europe, but less so in the United States.
LAB-based workflows can be very predictable and productive, with two
caveats. Fitst, LAB is not a particularly intuitive space in which to edit:
most LAB-based worldlows use editing applications,such as Heidelberg's
NewEolor, that put an LCh interface between the user and LAB. Second,
LAB is averylargespace indeed, since by definition it contains every color
we can see. As a result, it wastes an awful lot of bits on non-reproducible
colors, so all major editing should be done at the capture stage on highbit data-an 8-bit-per-channelLAB lile is a fairly fragile thing that doesn't
respond well to big moves in tone or color.
So a more typical color management workflow might look like that
shown in Figure 10-4, where captures from multiple sources are converted
early on into an intermediate editing space.
camem profile

Capture to edit to print

am-8

rntemtediate
space profile

press profile
I

conversion
4
I

conversion

scanner profile
Scanner RGB and camera RGB both get converted to intermediatespace RGB
when the application opens them. Editing is done in intermediate space RGB,
and the color is converted to press CMYKat print time.

480

Real World Color Management, kcond Edition

The Intermediate Space Advantage

Intermedmte-space worMows combine most of the simplicity of earlybinding worMows with all of the flexibility of late-binding workflows.
Early in the process. all imported color gets converted to the intermediate
space, and all new colors get defined in the intermediate space, so there's
little room for ambiguity. And as long as the intermediate space encompassesthegamuts ofall likelyoutput processes,or atleast thecolorslikely
to be output. it's easy to repurpose the work for different outputs, or to
do most of the work before the output process has been decided. Adobe's
applicationsusethe term "WorldngSpace" to define an memediate editing
s p a c e s e e Chapter 12. The Adobe Common Color Architechrre.

Conveying Color Meaning

Even thoirgh we're discussing it second, conveying color meaning is re
ally the first part ofthe color management puzzle. You can't ask the CMS
to match a color if you can't tell the CMS what that color is by supplying a
source profile. And there are only two ways to supply the source profile:
We can tell the CMS explicitly by embedding a profile in the document
or object (or assigning a profilesee the sidebar "Terminology: Tagging. Assigning, Embedding, and Assuming," later in this chapter).
b

We can tell the CMS implicitly by configuring applications to assume

that, absent any indication to the contrary (such as an embedded
profile), all RGB content is in Proflex space, and all CMYK content is
in ProfileY space.

The two options aren't mutually exclusive--manycolor management

worktlows. including the one we used to produce this book, make use of
both embedded and assumed profiles.

Embedded Profiles
The least ambiguous and most robust way to tell the CMS what colorsthe
numbers in the document represent is to embed the profile that describes
Lhem. In an enrkdded-profile zuor@ow, the profile always travels with the
object to which it applies, and is always available to serve as the source
profile for any conversion (see Figure 10-5).

Chaptr 10: Color Marigemen1W o r b

981

Fisure 10.5

scanner
profile

intermediate

press profile

1) The scanner embeds the scanner profik in the image.

2) The receivingapplicationmnvertsfrom scannerRGB to intermediate-spaceeditingRGB.

3) The application embeds the intermediate-space editingRCBprofile.

4) The next application mnverts to press CMYK, and embeds the press CMYKprofik.

51 Final editing is done in press CiWK The edited image is saved with the press CMYKproJile embedded.

Premature Binding
We think it should be self-evident anyway andconvert their colo~to
that it's technically impossible, some'hiddle-of-the-road"
CMYK
and practicallyless than rational, longbefore theyneed to, andhope
to convert your color to output for the best. Bruce wryly terms
space before the output process this the "premature-binding
is known, but every day we see workflow," and points out that
people proving us wrong. People the middle of the mad isgenerally
workon jobs with no idea wheth- where you find roadkill.
er they're going to be printed on
Sometimes you'll be forced to
a direct-to-plate sheetfed press convert to CMYK for an unknown
or on newsprint with plates im- output process, and thus you'll
aged fromfilm,
but they go ahead be forced to use "middle-of-the-

road" CMYK. But even then, if

you want to avoid having your
job look like roadkill, you should
tryto find out as much as you can
about the likely output process.
At minimum, you should have
an idea of the ink limits and dot
gain, for example. See 'heparing
Maferialsfor aNon-Color-Managed
Print Environment," later in this
chapter.

282

Real Wolld Color Managcmenl, W n d Edition

Terminology: Tagging, Assigning,

Embedding, and Assuming
One of the less-delightful aspects
of color management is the great
variety of terminology that different vendors use to describe
the same thing. A good example
is the set of terms mentioned
above: all relate to associating a
source profile with a document
or object, and often their meanings overlap.

Tagging is simply a generic term

for the act of associating a source
profile with an object or document. A tagged document is one
that has a source profile permanentiy associated with it, as an
embedded profie or an assigned
profile. An untagged document is
one that lacks n s o m e profile.
Assigning a source profile is a
term that Adobe applications use
to describe the act of tagging a
document. Hair-splitters should
note that assigning a profile is
a distinct process from embed-

ding a profile-you can assign a

profile at any time, but you can
only embed a profile as part of the
Save process.

source profile, sometimes called

a default profile, for RGB, and
another for CMYK. (Some applications let you set different
default profiles for pixel-based
Embedding a source profile data and vector-based data.)
describes the act of saving a doc- When the application encounument with the source profile ters an untagged document
included as pan of the file data. or object, it uses the assumed
Here's an example that makes profile as the source for any
the distinction between assign- requested conversions,including
ing and embedding more than the one that converts the color to
mere hair-splitting. If you assign monitor RGB for display.
Assuming a profile results in a
a profile to a document, and your
machine crashes before you save, different behavior ftom assignwhen you reopen the file,it won't ing a profile-if you change the
have the assigned profileassociat- default proNe, all open untagged
ed with it. However, if you embed images change appearance, since
a protile, and the machine crashes. they're now being interpreted
the profile will still be associated with the new default profile as
with the document when you the source. But if you assign a
profile, the image is no longer
reopen it.
untagged as far as the applicaAssuming a source profile is tion is concerned, and keeps
a way to configure applica- its assigned profile no matter
tions' behavior toward ontagged what you do10 the default pmflle
objects. Typically, you set a settings.

When you embed a profile, you literally write a copy of the profile into
the document. AU color-managed applications know how to interpret embedded profiles inTlFE JPEG,and. o n the Macintosh platform. PICT documents. Some applications also offer some degree of support for profile
embedding in EPS and PDF donunents, but it's sufficiently variable that
we'll discuss it fully in the individual application-specific chapters that
follow this one.
Embedded-profile workflows are by far the safest choice when your
production chain takes inputs from multiple sources and converts them
to outputs for multiple destinations. such as a service bureau operation.

When you always embed profiles, there's never any ambiguity as to the
color meaning of the numbers in your documents. Many production
people will argue the safest choice is one without any profiles that deals
with color problems tbrough an iterative proof-then-color-correct process. It's only safe because it's familiar, and it's extremely inefscient to
use human skill to solve the problem of various devices having their own
peculiar behavior. Such use of skill and craft is often necessary even in a
color-managed workflow, but to nowhere near the same degree as in a
traditional one.
One major disadvantage of embedded-profile workflows is that they
increase file size, sometimes by a little, sometimes by a lot. Matrix-based
RGB editing space profiles are relatively tiny-less than a ldlobyte-but
if you're uploading 30 images to a Web page, and embedding the same
profile in each one, you're uploading 30 kilobytes of redundant data.
CMYK output profiles can be quite l a r g e t h e ones we used to make
this book weigh in at 2.4 megabytes apiece--and if we chose to embed a
profile in every single graphic in the book, we'd be shoving an extra one
or two gigabytes through the production chain.

Assumed Profiles
The key feature of the assumed-profile workfiw is that, instead of relying on embedding profiles in all our documents, we agree on a single
profile to use for untagged documents that we push through the production chain, and we configure all our applicationsto assume that untagged
documents have this profile as their source (see Figure 10-6).
Figure10-6

Assumed-profile

workflow

assumed RGB

assumed CMYK

In an assumed-projZe workflow, RGB and CMYKunlues are always

interpretedasassumed RGB andassumed ChWK respectively.Profiles
mn'tembedded, and conuersionsarealways between the assumed
prof iles'spaces.

484

R u l World Color Management, kcond Edition

Of course, an assumed-profile workflow puts all the onus on the user

to disprove the old saw that "when you assume, you make an ass out of u
and me '-you don't have the safety net offered by embedded profiles, so
you have to make sure that the untagged objects were brought correctly
into the assumed profile space to start with, and you also have to make
sure that theassumedprofiles areset correctly ineach application so that,
if conversions do occur, they're the right conversions.
As a result, assumed-profile workflows require a great deal more coordination between programs and people than do embedded-profile
workflows. so they're best suited either to situations where a single individual is in control of all the work, or where a group with good communications is workingtowards a single output, such as a bookor aperiodical. In
situations such as these. many people find an assumed-profile workflow
makes it simpler to keep things straight. All programs that deal with any
RGB color values use the same assumed RGB profile. All programs that
deal with any CMYK color values use the same assumed CMYK profile,
and all conversions from RGB to CMYKuse the same sourceand destination profiles.
8

Hybrid Workflows
In practice, most real color management workflows are hybrids, using
embedded profiles and assumed profiles either for convenience, or hecause applications or devices force one or the other.
For example. Adobe Illustrator can assign profiles to EPS files, but it
can't embed them. So when you transfer Illustrator EPS files to another
application, you have three options: use an assumed profilein the receiving application. manually assign a profile in the receiving application, or
considerthe EPS "print-ready."The last case means thatall colotmanagement must occur prior to saving the EPS file, and thevalues in the file will
be sent to the output device.
It's also important to note that embedded-profile workflows, while robust, aren't bulletproof. (The first person to make a bulletproof worMlow
willget veryrich,very quickly!) If one personopens an image in a non-color-managed application (such as Adobe Photoshop 41, does some minor
editing, and resavesthe file, the embedded profileisgone. In thisscenario,
if you spot what happened quickly you can save the situation by simply
re-embedding the profile with a suitable application.

Chapto 10: Color ManagrnunlWorkflow

485

But here's a worse scenario. If someone opens the image in Photoshop

5,and the application is configured to convert images to her working
space and to save without embedding profiles, not only will the embedded
profile get shipped out, but the numbers in the image will have changed
drastically and no profile will be embedded upon resaving the image. So
simply re-embedding the profile won't work-you'd have to find out what
workingspace Photoshop 5 was set to use, then embed that profile to retrieve the correct color meaning for the new set of numbers in the image.
which may be difficult for those of you who aren't mind-readers.
The lesson here is that profile embedding is a useful tool for communicating color meaning, but it isn't a substitute for clear communication
between the people in the production chain.

Workflow Within a Program

Workflow within a single program can be broken down into two aspects:
b

How the program handles untagged documents

How the program handles tagged documents

Remember, color management only does two things--convey color

meaning and convert numbers to preserve that color meaning. So there
are really only three possible things the application can do in either case,
though the implications are slightly different in each.

Opening Untagged Documents

Any color-managed application has lo make some assumption about
the source profile for untagged documents. It needs a source profile so
that it can display the document on your monitor and convert it to other
profiles'spaceson request. Sowhen the program encounters an untagged
document, whether by the Open command or by commands like Import,
Place, or Get Picture, it can only do one of three things:
b

Assume a profile

Assign a profile
Assign a profile, and do a conversion to some other profile

986

Real World Color Management, Second Ldltlon

Some applications allow you to do only one of these things, while othe n may allow you to choose between two or three alternatives, but these
are the only three possibilities-see the sidebar. "Terminology: Tagging,
Assigning, Embeddig, and Assuming," earlier in this chapter, for the subtle distinction between assuming and assigning a source profile.
The other factor to consider is that some applications only allow you
to set a default behavior that applies to all untagged documents automatically, whereas others offer the option of choosing the behavior on a
case-by-case basis by presenting a warning dialog like the one shown in
Figure 10-7.
Note that in all cases, the question is fundamentally one of supplying a
source profile forthe untagged document. Once the application has done
so, the only difference between tagged and untagged documents is that if
you change the default profiles in the application, untagged documents
take the new default profile as their source, whereas tagged documents
keep their source profile until youexplicitly changeit by doing an assignment or conversion.
Note that the "assume profile" possibility is the only one which leaves
the documentlobjectuntagged. Inallothercases thedocument effectively
becomes tagged-Macromedia FreeHand and CorelDRAW, for example.
always assign the default profile rather than assuming it. However, the
assumed-protile workflow, where documentslobjects remain untagged,
is the most common workflow.

Opening Tagged Documents

As with untagged documents, applicationscan only do oneof three things
when they encounter adocument-whether by theopen command or by
commands like Import, Place, or Get Picture--that contains an embedded profile:
ngure10.7

Photoshop's Missing
Pmfilewarning

Chapter 10: Color Management Workflow

b

487

Use the embedded profile as the source profile.

Use the embedded profile as the source profile. then convert From that
source profile to another profile.

Ignore the embedded profile and assume or assign a different

profile.

Some applications only offer one of these alternatives, while others

may offer two or three, but these are the only three possibilities. And
some applications only let you choose a single behavior to apply to all
documents automatically, while others may again let you choose on a
case-by-case basis by presenting a warning dialog like the one shown in
Figure 10-8.

Color Management Between Documents

When you move objects between documents in color-managed pagelayout or illustration applications, where a document may comprise multiple
objects, each with its own source profile, you have exactly the same three
options you have when you open, import, or place tagged documents:
b

Use the embedded profile as the source profile for the object.

Use the embedded profile as the source profile for the object, then
convert from that source profile to another profile.

Ignore the embedded profile and assume or assign a different profile

to the object.

Some applications, suchas QuarkXPress4 and 5, allow you to set these

options on a per-document basis as well assetting an application default,
which is potentially powerful, but also potentially confusing. So youneed

Photoshop'sProfile
Mismatch (Open)
warning

288

Real World Color Managcmenl, Second Edition

to pay attention to the color management settings for each document.

or decide ro standardize your settings for all documents (in which case.
make sure that you know what you're doing when you open a document
that was created by someone else, because she may have set the color
management preferences differently).
A different choice arises in applications such as Adobe Photoshop,
where each document has a single profile governing all elements. When
youmove pixels (whethervia copy and paste, or drag and drop) from one
document to another in Photoshop, and the two documents have different source profiles, you have only two choices:
b

Move the numerical values of the pixels (which means that their appearance will change because it will be governed by the source profile
of the receiving document).

Move the color appearance of the pixels (which means that the numerical values will change to re-create the original appearance in the
different profile space of the receiving document).

A less-wordy description of the choice is: convert, or don't convert (as

the dialog box shown in Figure 10-9 indicates).
Some applications will always do one or the other without giving you
a choice, in which case you need to figure out which one the application
in question is doing. But these are the only two possibilities.

Assigning, Converting, and Soft-Proofing

So far, we've looked at what applications let you do when you open or
import a document or object, transfer an object between documents, or
combine different objects into asingle document But most color-managed
applications also let you do profile assignments, profile conversions, and
Figure 10.9

Photoshop's Profle
Mismatch (Paste)warning

Chspm 110: Cdor Management Workflow

489

sofr-proof which are reallysets of conversions that are applied temporarily to let you use your monitor to simulate final output-see the sidebar,
"Soft-Proofing Basics," later in this chapter.
In the case of applications like Photoshop, where adocument can only
have one profile. these options are relatively straightforward.You can assign a source profile to a document, you can convert a document to another profile's space with a choice of rendering intents, and you can set
the monitor to provide a live preview of how the document would appear
after a conversion to some other space-so you can, for example, work
on an RGB document while previewing the effects your RGB edits would
have on the file if it were converted to output CMYK.
With applications whose documents may contain multiple objects in
different color spaces, such as page-layout applications, profile assignment becomes hickier.The main factor is that objectsnative to the layout
application are handled differently from linked objects such as imported
graphics. (By native objects, we mean text, lines, shapes, backgrounds,
and so on, that were created inside the application.)
Native objects inherit the assumed or assigned profile that applies to
the document in which they are placed-the key point is that you don't
assign profiles to native objects directly. If the documenr bas ProfileXassigned to it, all native objects assume that profile as their source. If you
change the document profile, you change the source profile for all native
objects.
For linked objects, however,a newly assigned profile only applies to the
specificinstance of the object in that page-layout document. It doesn't affect the original file-to do so. the page-layout application would need to
be able to go out and rewrite the linked file, which the major page-layout
applicationscan't do.
Essentially, the assigned profile acts like an assumed profile inside that
specific page-layout document, and only to the specific instance of the
object to which it's applied-if you import the image again, and place it
in a different part of the page-layout doc, it'll use either the embedded
profile or the default profile. We think that assigning profiles other than
the true source profile to linked objects in a page-layout application is
a dangerous practice, and avoid it in all but the direst emergency. We'd
rather go back to the application that originated the object and assign a
new profile there, then update the link in the page-layout application.

990

R u l World Color Management, Second Wltlon

Conversions in page-layout applicationsgenerally apply only to native

objects, not to linked or imported ones.We almost always try to create everything in our layout applications in final CMYK from scratch, but when
this is impossible (because, for example, the final CMYK isn't known when
we start creating the document), we use the page-layout applications'
conversion features to convert thenative objects.Then wemake sure that
all the linked objects are in the correct space, and if necessary, do any
conversions in the applications that originated them.
Simulations always apply to the entire document, whether it's a single
image in Photoshop or awhole hookin apage-layout application. Insome
applications, such as QuarkXPress or Adobe PageMaker, the simulation
controls are presented as pan of the color management preferences (see
Figure 10-10).
Figure 10.10

Monitor simulation

- .- ~
aWCdr-rum

;
I
]

+display simulation
-

Many applications, such as

QuarkXPress,lohose Color
Mana~ementPreferences
areshown at left, include
the display sirn~datiotz
control asparr of thecolor
management preferences.

,J

In other applications, such as Adobe Photoshop, Illustrator, and InDesign, simulations are controlled by commands accessed from the main
menu bar (see Figure 10-11).
However they're presented, the soft-proofingsimulationsdo the same
thing. They convert all the supported objects in the document-on the
fly, for display only-to the designated output profile space, then they
convert that simulation to monitor RGB to display the predicted output
correctly on the monitor. (Formats such as EPS or DCS EPS don't allow
page-layout applications to change their preview, so you can't do soft
proofs of objects in these formats.)

Chapter10: Color M a n a g w n t W o r M a

2P1

Soft-Proofing Basics
The nai've View of color management, aspresentedinall too many
vendors' marketingspiels, is that it
will make your prints match ynur
monitor. If you've read this far,
you know that the laws of physics
make it impossible to do so.
What color management can
do, however, is to show you on
your monitor how your color will
appear when it's reproduced in
print. This technique is known as
sop-pmofinzg,and we thinkit's one
of the most valuable things color
management brings to the table.
Soft-proofing first converts the
document's colors to the final
output, then renders those colors back to the display, on the
fly, without actually changing
the values in the document. It6
basically a preview of the final
output.
Soft-proofing has limitations.
Your display has a fixed color
gamut, and it can't show you colors that lie outside it, so some
saturated cyans and orangeyellows that are readily achieved
in print can't be displayed, though
the clipping is very slight. But
there are different ways to ren-

der a simulated print hack lo

the display and each one has its
strengths and weaknesses.

R&tin colorimetric rendcrlng.

This is thc way most applications
render color, including simulations, to the display. White in the
simulation-paper white-gets
translated to monitor wh~te,and
all the other colors are shifted to
match that white. This type of
soft-pmof doesn't show the effect
of the paper white on the overall
color, but since our eyes adapt to
the monitor white, It6 still a very
usehtl view.

b To

make this view work, you

musr hide any white user interface elements, otherwise
your eye adapts to that white
instead of the simulated paper
white. and the color seems to
have a color cast.

You may he templed to conclude

from the above that accurate
soft-proofing is impossible. In
the strictest sense, you'd be right,
but the truth is that we've never
seen any proof that matched
the final product exactly, except
perhaps for press proofs, which
are prohibitively expensive for
most real-world jobs.
We learn to interpret hard-copy
Akdufe colorhmtric rendering.
proofs,
whether film-based or
In theory, ahsolute colorimetric rendering, which shows the digltal, and we simply have to do
actual color of the paper white, the s a n e with soft proofs. Once
should offer a more accurate soft you've learned to interpret what
proofthan relativecolormetricren- your display is telling you, we bedering, but in pmctice, you imme- lieve you'll find that a correctly
configured soft proof is every hit
diately encounter two problems:
as reliable as any other proofing
b Toshowthecolorofpaperwhite,
system short of an actual press
the monitor has to displaywhite
proof, and it6 not only a great
as somethingless thanRGB 255.
deal cheaper, but also a great deal
255, 255, so you immediately
faster and more flexible.
lose some dynamic range, making the color appear flat

Some applications offer per-object control over the rendering intent

from the object's space to the simulated output space, some apply perceptual rendering to raster images and relative colorimetric rendering
to vector vaphics, a n d some simply use the profile default rendering in
all cases. Most applications do relative colorimetric mndering tom the
simulation space t o monitor RGB, though a few offer control over the
rendering from simulation to monitor space. We'll look a t these details in
the annliratinn-sn~rifirrhanters that follnw thisone.

292

Real World Color Management, Second Edition

Figure 10-11
C",,rn

Applicationssuch as AdobePhoro.
shop, uthose display
simularion controls are shou~rr
here, letyou change the
sofr-proof settings without
changing the colormanagement preferences.

Printing and Proofing

The last set of color management controls that work inside individual
applications are those that deal with printing both final color and proofs
of final color.
All the color-managed applications we know of allow you to perform a
conversion from the source profile or profiles in the document to a single
target output profile, as part of the print stream. The document itself remains unchanged-the conversion only applies to the data that gets sent
to the printer driver.
Tip: Disable Color Management In the Printer Driver When You Use
Application Color Management to Print. At the risk of belaboring the
obvious, if the application performs a conversion to the data being sent
to the printer driver, you almost certainly don't want the driver itself to
do a conversion on top of the application's conversion-so when you use
this feature, make sure that you turn off any color management in the
printer driver itself.

In applications whose documents contain multiple objects with individual profiles, the rendering intent controls are the same as the ones
offered for soft-proofing. Applications such as Adobe Photoshop, whose
documents contain only one source profile, may allow you to choose a
rendering intent (see Figure 10-12).
mure 10-1P

Printing controls

Photoshop's printing
controls letyou select
a printer profile and
rendering intent.When
you do so, Photoshop
performs a conversion
on the data thatgets
sent to the printer drive,:

In addition, most color-managed applications allow you to print a

simulation of the final output to a composite printer such as a desktop
inkjet or color laser.
In some applications,such as Adobe Photoshop, Illustrator, and InDesign, this feature is tied to the soft-proof setting, so that instead of choosing the document space as the source profile, you choose the soft-proof
space. This makes the application perform the conversion(s)from object
spaces to the soft-proof space as part of the print stream.You then choose
an output profile and rendering intent to cross-render this simulation of
the final output to your desktop printer, which makes the application do
one more conversion from the output simulation to your desktop printer
space using the selected rendering intent (see Figure 10-13).

994

Real World Color Manylclll.nt, Second Edition

U~UC
10-13

InDesign and Photoshop

cross-rendering

Source Space set to Proof

Source Space set to Proof

Adobe Photoshop and InDesign both tie cross-rendering to the soft-proof:When

you choose Proof Setup Photoshop) or Proof (InDesign)as the source space, the
applicationfirst converts the color using the destination profile specified in the
Proof Setup dialog box. Photoshopalso uses theProof Setup rendering intent
for this conuersion-rendering intents in InDesign are applied separately to
each object Then theapplimtion does a second mnumsion using the destination
profile and rendering intent specified in Print Space.

Chrptu 10: Color h n m m c n t Workflow

295

Other applications may force you to actually convert every element to

final output space as the only way to cross-render to a composite printer, in which case the conversion is simply one from document space to
printer space.

Workflow Between Programs

lou ran con\ev color mramng b e t ~ e e nprograms by either, or both, of

two methods:

w Embedding profiles in any objects that travel between programs

b

Ensuring that both the originating and receiving application use the
same assumed profiles

Our tendency, which is by no means a hard and fast rule, is to embed

profiles in RGB, and assume profiles in CMYK, for two reasons:

w RGB profiles are typically small and add little to the file size. while
CMYK profiles are often larger than the actual files themselves.
At least two important CMYK formats-XMYK EPS and C m K DCS
EPS-don't reliably support profile embedding, and are often treated
as "output-ready" formats, both by people and applications.
However,when we don't embed profiles in CMYK, we alwaysleavesome
kind of audit trail that tells us what flavor of CMYK the numbers in the file
represent. We may archive the files in a specific folder that also contains
the profile. or we may include the profile name in the document name,
in the file information, or in whatever form of metadata the file format
permits.
Workflow between applications is perhaps the one aspect of color management where common sense applies. All applications that deal with
color have to make some assumption about the colors represented by
RGB and the colors represented by CMYK. Color-managed applications
can only get those assumptions from one of two places, a default profile
or an embedded profile. If you make sure that your default profiles are
synchronized across all your applications, and you set your applications
to deal with embedded profiles the way you want them to, your color will
flow reliably and predictably Erom one to another.

996

Real World Color Management, k o n d Wltlon

Workflow lnto and Out of Color Management

Even though this is a book about color management, we can't deny the
fact that large chunks of the world aren't color managed. What we call the
"color management workflow" may be just a small piece of the overall
workflow. We constantly bring materials in from devices, software, and
people who don't use profile-based color management. At the other end,
we prepare documents for further non-color-managed stages. For example, with some workflows, as soonassomeoneconvertscolor to CMYK, he
employs more-traditional CMYK correction, targeting, and output techniques thar don't involve profile-based color management.
There's a widespread and totally incorrect assumption that color management is an all-or-nothing proposition-unlessyou have profiles fotall
your inputs and all your outputs, and you use them religiously, you can't
do color management. If that were true, color management would be a
p a t deal less useful than it is.
We routinely bring color from non-color-managed sources into a colormanaged workflow, and we equally routinely export color from a colormanaged workflow intoThe Great Non-Color-Managed Unknown.

Bringing Color lnto a Color-Managed Environment

The first question we always ask when we're confronted with bringing
non-color-managed documents into the color-managed workflow is,
"Can we deduce the source profile?"

Known sources. If the source profile is known, or, with a little deductive
reasoning, knowable, we can simply assign it, and from then on, the ofrending document is color-managed. A trivial example would be an image from a scanner whose software doesn't embed profiles, for which a
profile is available. We'd simply open the scan in Photoshop and assign
the scanner profile (and then we'dvery likely convert itimmediately to an
intermediate editing space for further work).
A less-trivial example would he an untagged image sent by a peerwho
is reachable by that most-overlooked color management tool, the telephone. Usually, a few questions could determine which space the image
was saved in, and again, we'd assign the correct profile.

Chap* 10: Color Management Workflow

9W

Unknown sources. In many cases, however, the source profile is simply

unknowable, either for technological reasons-the image was shot on an
unprofiled digital camera. or scanned from a color negative--or for human ones--the person who created the material is ekther unreachable or
is sufficiently clueless about color management that asking questions is
an exercise in futility. In that case, you have what Bruce fondly calls "mystery meat." We recommenddifferentstrategiesfor RGB mystery meat and
CMYK mystery meat:
b

In the case of RGB, we usually trya few different editing-space profiles

to see ifone produces a more reasonable appearance than anotherthere's an element of guesswork or mind-reading in this, so we don't
hope for miracles. It'soverwhelminglylikely that thecolorwascreated
in either a monitor-like space such as ColorMatch RGB or sRGB. or an
editing space such as Adobe RGB, so those are the ones we generally
try. If the image appears dark in a gamma 2.2 space, we try a gamma
1.8 space. If it appears washed our in a relatively large-gamut space
such as Adobe RGB, wetrya smaller one such as sRGB.We then assign
the profile that produces the most reasonable appearance (yes,this is
a very subjective criterion), and use it as the basis for any editing and
subsequent conversions (see Figure 10-141.
in the case of CMYK,we lean to the view that guessing games are
pointless because there are simply too many possibilities to address.
We take the not-entirely-unreasonable position that if someone sends
us a CMYKfile,he expects us to print it. so we assign theCMYK profile
for the project to which the document is related, and use that profile
as the basis for any necessary editing. If the on-screen result is truly
disgusting, we may try assigning some different output spaces to see
if we can find a suitable match, and then convert to our CMYK, or we
may simply edit the file, trusting our calibrated and profiled display
as an additional aid to the process.

When Color Management Starts

The details of the recipes we gave above for turning mysrery meat into
something more palatable are ones we find useful, but what's really important is why they m r k . And the simple reason why they work is that.

998

Real World Color Management, kcond w i t h

Mystery meat

This mystery meat

image appears both
dark and oversamrated when we open it
in our Photoshop workingspace,Adobe RGB
(1998).A lowergamma
value will lighten the
image, and a smaller
gamut will make the
same numbers produce
less-samratedcolors.

Colormatch RGB has

a lowergamma than
Adobe RGB (1998)gamma 1.8 as opposed
to gamma 2.2- and
it also has a smaller
gamut than Adobe RGB
(1998).When we assign
the ColormatchRGB
profile, the image
appears less dark, with
more natural-looking

Chapter 10: Color Management Workflow

999

as soon as you open a mystery meat document in a color-managed application, you're using a known source profile to look at the color. Ariy
application that can convert color from one color mode to another-from
RGB to CMYK, for example--has to make assumptions about the color
appearance represented by both the RGB and CMYK numbers. One ofthe
things that makescolor management acool tool is that, in color-managed
applications, you can easily see and control those assumptions.
All color-managed applications let you set default profiles, which they
use to interpret the numbers in untagged documents, so when you open
an untagged document, you /mowwhich profile the application is using
as the document's source profile, and you know what color that profile
makes the document display on your calibrated monitor.
It may not be a source profile that represents the intent of the person
who sent it to you, but you've gone from knowing nothing aboutthe color
to knowing how it looks in, say, Adobe KGB. Color management purists
might blow a gasket at the notion, but we'd say you've already started to
color manage the document.
The purist (Bruce says "fascist") view of color management is that its
goal and purpose is to fake an original image, and represent it as faithfully as possible on various outputs. Purists see color management as a
scdpel.We liken it more to a Swiss Amy Knife--it's neither as precise nor
as specialized as they claim. But it's a lot more useful.
In fact, it'simpossible to notuse color management in any application
that does color conversions. All you can do is hide the assumptions the
application makes when it does so, which we think is pointless. So as far
as we're concerned, color management starts as soon as you open a file
in a color-managed application.

The Limits of the Knowable

Of course, when you're dealing with mystery meat, there's always an element of mind-reading-because the person who sent you the mystery
meat hasn't (yet) learned to use color management to convey his color
meaning clearly, which is bad, but he also hasn't learned any other way
to do so, which is the real problem.
Color management lets you attach a specific color appearance to a set
of KGB or CMYK numbers. Getting from that appearance to the appearance the originator intended is not something that color management can

300

Raal World Cdor Management, W n d Edition

address-it'sapeople problem, not a technology problem. Color management can help by lettingyou tryout various plausible and implausible alternatives, but if people give you untagged files with no other information,
you simply have to use your best judgn~ent.Of course, you can always iry
asking why you're being given untagged files--see the sidebar, "Fear of
Embedding." on the Facing page.

Tip: Always Send Hard-Copy References. If you're dealing with someone whose workflow is obviously non-color-managed, and you're going
through multiple edit rounds, adumyssend a hard-copy reference such as
an inkjet print, cross-rendered to the output space, that clearly indicates
yourcolor meaning. It's unlikely that the other person will see the same
appearance you do when she opens your edited file. so the hard copy
~rovides
a less ambirrnous reference.
lfwe have anagenda in writing this book, it's to make mystery meat go
away. Lack of clear color communication is never a good idea-it simply
creates extra work for all concerned. Embedding profiles is one simple
way to remove ambiguity, but it's not the only way, and if you're dealing
with someone who is seriously terrified of color management, it may not
be the best way. So do whatever you can to eliminate mystery meat, but
remember that gentle persistence almost always works better than wildeyed evangelism, and in those cases where mystery meat is just a fact
of life, use the guidelines we provided above to render it somewhat less
mysterious.

Preparing Materials for a Non-Color-Managed

Print Environment
rhere's no part~cula~
mag~ctrickhere.Yousimply have to find out as much
as possible about the printingconditions, then convert to a CMYK profile
that bears some reasonable resemblance to those conditions.
If the job is large or critical, and you have profiling capabilities, you
maywant to consider profiling the pmofingsystem-lust have the printer
print the profiling target thesame waythey'd printyourjob, then measure
it and build a profile.
Failing that, try to find out the ink limits and anticipated dot gain on
press, andcreate your CMYKdocument usingaprofle that matches those
conditions Basically, you're aiming at the side of a barn-you want to try
to hit it somewhere close to the middle.

Fear of Embedding
may take persistence to make
the question percolate to a place
where it can do some good.)
Don't make promises. or offer any guarantee other than
that profile embedding won't
make things any worse. Just be
I don't know how to embed gently persistent. If the docuprofiles, or which profiles to ment originator obviously needs
Embedding profiles will change embed. It's hard to argue with help figuring out how to embed profiles, decide if it's worth
my color. Many people have had this one.
t h ~ sexperience when they sent
Very often, though, you'll find providing that h e l p y o u pmba job to someone who had color that people simply want to be ahly don't wan1 to wind up givmanagement configured incor- told what to do with color man- ing endless hours of unpaid
rectly, or who misunderstood its agement. Most applications tech support unless it's to a very
purpose. Of course, when color make it difficult to embed the important client.
Sometimes it's easier to simply
management is set to work right. wrong profile, and easy to embed
embedding a profile makes it the right one. So if you're tired of find out what profilethe originator
much less likely that the color cooking with mystev meat, it's uses for his CMYK or RGB working
will change. Remember-a single always worth asking the question. space. and simply apply it yourprofile simply describes. Itdoesnt "Why would embedding profiles self to incoming work. Profiles
make things worse than they are aren't the onlv way to convey coldo anything to the file.
or meaning, just the simplest and
now?"
Ifno good answer is forthcom- most automatic. But any small
My r m l c e provlda/color gnu1
IT guy/drlnklng buddy lold me ing. gently suggest that profile steps you can take to reduce the
not to. People generallv fear embedding might be a good idea amount of mystery meat in your
what they don't understand, and in this workflow. (If you're at the diet will make you a healthier and
bottom of a long food chain. it happier color management user.
t~ to make sure that no one else

Very often, the reason people

send mystery meat is because
they're afraid to do anything
else-specifically, they have an
irrational fear of embedding
profiles. This fear can stem from
many sources, but some of the
most common ones are:

understands it, either. In some

workflows, it makes absolute
sense to not embed profiles, but
if you're receiving mystery meat,
this workflow obviously isn't one
of them

Convert all your color to final CMYK,don't embed any profiles, and
submit the job.You may have to go through some rounds of proofingand
correction-that's the normal, expected workflow-but your first submission will likely b e in the ballpark
Some non-color-managed shops will claim that they can accept RGB
files a n d convert them to CMYK themselves. We usually treat such claims
~ l t suspicion,
h
t o say the least. If the shop can rell us what flavor of RGB
they expect, we may a t least go ahead a n d try a test. But if the response
makes it clear that multiple flavors of RGB is a concept they've yet to entertain, we'll walk away.

302

Real World Color Management, Second ldklon

However, if you submit RGB,with embedded profiles, on a CD or other

non-rewriteable medium, there can be no argument about what you've
submitted-and by accepting RGB files, the printer has taken the responsibility for the RGR-to-CMYKconversions.So if things go badly, you've at
least covered your bases.

Preparing Materials for the Non-Color-Managed

Internet
The vast majority ofweb browsers simply take the RGB values in files and
send them unmodified to the screen. So unless you go and calibrate the
monitor of every user who is likely to look at your site, you have no wav
of knowing exactly what they're going to see.
Various vendors have touted schemes for managing color on the Internet.They all work, up to a point, but they all do so by forcing theviewer to
do some kind of visual monitor characterization, which they then use to
alter the information that gets sent to that particular Web browser. Most
of these solutions are sufficiently complex and expensive that they can
only be implemented at the enterprise level, and by all accounts, no one
is making huge amounts of money doing so.
One day, all monitors will be self-calibrating and self-characterizing,
and all operating systems will use display compensation as a system-wide
feature. Until that lucky day arrives, we suggest that theonly practical solution is to aim forthe lowest common denominator. Fortunately, that lowest common denominator is exactly what thesRGB space wasdesigned to
represent-it purports to represent the "average" uncalibrated Windows
monitor. So our simple recommendation is to convert all your color to
sRGB, and then save without embedding a profile, before uploading it.
The only color-managed browsers we know of exist on the Macintosh.
For Mac OS 8.9 and X, it's Microsoft Internet Explorer. If you enable Colorsync in Explorer's Preferences, it will assume sRGB for untagged images
and use the embedded profile in all other images, using your monitor
profile as the destination profile. And on Mac OSX only, it's OmniGroup's
OmniWeb, which currently always assumes images are sRGB and uses
display compensation.
So Macintosh users who use a display gamma of 1.8 will at least have
a chance of seeing correctly. For most Whdows users, sRGB is at least in
the ballpark. And while gamma 2.2 images will seem darkand muddy on

an unmanaged Macintosh gamma 1.8 display, the Mac users should be

used to that.

Internet Color Workflow

When you prepare materials for the Internet, you'll probably have to use
a mixture of color-managed and non-color-managed applications. If you
aren't bothered by the color appearing different in color-managed and
non-color-managed applications, you don't need to do anything, but if
you're at all like us, you may find that disconcerting.
There are really only two solutions to the problem that make sense
to us:
b

If your work is exclusively for the Internet, calibrate your monitor to

sKGB-most monitor calibrators offer sRGB as a preset, and for those
that don't, use 6500 K as the white point and 2.2 for gamma. Then.
do everything in sRGB. Your calibrated color in the color-managed
applications will closely match your uncalibrated color in the noncolor-managed ones.
Use monitor RGB as the source profile for all your Internet work.
Color-managed applications will see that your RGB is already monitor RGB, so they'll just send the values in the file to the screen, the
same way non-color-managed applicarions do. Then, when the work
is complete, convert it to sKGB.

If you expect your target audience to be primarily Macintosh users.

you may consider targeting Apple KGB instead of sRGB. Neither solution
is ideal, but until such time as the Internet becomes something other
than a very large collection of random output devices. they're the best
we have to offer.

Understanding Workflow
In 1111schapter, we've tried to present the essential workflow concepts
and features that all color-managed applications share, no matter how
they're presented, while relating them to the fundamental concept that
color management does only two things-convey color meaning, and
convert dcvice values to preserve that color meaning.

304

Real World Color Management, kcod Edition

Once you grasp these basic concepts, you'll find that you can look at
just about any application and figure out what each color management
feature does, because they always boil down to some combination of
the two fundamentals. In the following chapters, we'll lookat the specific
ways color management is presented in someofthe most common colormanaged applications, but we can't cover them all, and it's often all too
easv to get bogged down in the details. So use this chapter, and the concepts it presents, to keep the biggerpicture in mind when you're grappling
with the minutiae ofthis or that application-theanswer is almost always
simpler than it might first appear.

Color
ianagement in
the Operating
System
Who Does What to Whom, When?
Apple and Microsoft tend to paint an overly rosy picture of color management as a panacea that will make whatever comes nut of your desktop
printer match your monitor, automatically.Anyone who has ever printed
from a desktop computer, or has a passing knowledge of the laws of physics, knows othenuise. The truth is that 0s-level color management does
less than the marketing hype, but it also does much more than the marketing hype.
It does less in that 0s-level color management can't change the gamut of
your printer to match that of your monitor-it can come reasonably close,
but there's simply no way you'll get that RO. GO, B255 blue out of your printer
(or,for that matter, on any other reflective hard-copy medium).
But it does more in that it's not just a color management system-it's
a whole architecture for color management systems to live in. OS-level
color management provides a whole slew of services that applications can
call to do all sorts of useful things, but in some cases makes us do more
work than should be necessary.

306

Real World Color Hanagcmenl, Second Edlflon

Color Management in OSs, Applications, and Drivers

One of the more confusing aspects of color management lies in tiwring
out who does what to whom. when. One might think that in the year 2004
we'd have applications, OSs and device drivers that could effectively talk
to each other and prevent obviously conflictingsettings from being used.
But those ducks aren't yet in the row. Therefore, ifyou want to ensure that
color conversions happen correctly,at the right time, and only at the right
time, understanding the various interactions yourself is critical.
0s-level color management is the generic term for Colorsync on Mac
OS, and ICM on Windows. In the past, we've thought of 0s-level color
management as something that gets invited to the party on request. and
on Windows. Mac OS 9, and Mac OS X up to version 10.2.x this is indeed
how 0s-level color management behaves-it's invited either by applications or drivers specifically written to ask the OS for a color conversion.
It often requires an explicit setting in the application or driver for it to be
used. There are only two actors to cause 0s-level color management to
occur-applications, and drivers.
But Mac OS X Panther (10.3.~)changes the rules, and comes to the
party whether invited or not. The operating system follows an "always
on" philosophy to color management, unless spec~ficsettings are used
to prevent it from being involved. So the possible actors in Panther are:
applications; drivers; and the OS itself. This approach has some potential
pitfalls that we discuss in detail in "Mac OS X Panther (10.3.x)," later in
this chapter.
The main graphic arts applications, such as those from Adobe, Quark.
Macromedia, and Corel. offer user-configurable color settings in the
applications themselves.We call these "color-managed applications." The
distinction between color-managed and non-color-managed applications is important because the operating systems donZ negotiate potential conflicts between settings in color-managed applications and device
drivers. So yon have to do so yourself.
Color-managed applications allow far greater control within the application than do the compararively rudimentary settings in printer drivers, for example. Professional users will likely want to rely on application
settingsandcommands for color management, andensure no additional
OS-level color management occurs.

Non-color-managed applications rely on driver settings, and a series

of assumptions (made by either the OS or the driver), to control colot
management, because that's all there is.
One of the keys to successful color management lies in making sure
that OS-level color management in the various device drivers cooperates
with the application-level color management done by color-managed applications, rather than injecting unexpected extra conversions into the
color production chain.
With color-managed applications,youcanuseeither application-level
color management or driver-level color management. but we recommend
using the application controls to manage color rather than relying on
limited printer driver options.You almost invariably want to avoid using
application-level and driver-level color management at the same time,
so much of this chapter is devoted to understanding how OS-level color
management functions so that you can make sure that i L doesn't do anything you didn't expect.

ColorSync and ICM

Apple's ColorS!ncand Microsoft's 1CM (ImageColor Management)are the
s part of the Mactechnologies that provide color management s e ~ c eas
intosh and Windows operating systems, respectively. As previously noted,
they're both less than and more than color management systems.

What Are They?What Do They Do?

1he overwhelming bulk of both ColorSync and ICM is comprised ofAPls,

or Application Programming Interfaces. These are chunks of code that

developers can exploit using relatively simple calls to the operating system. They provide a way for developers to include color-management
capability into their applications without having to write their own color
management sysrem from scratch, and they also provide an architectural
framework in which third-party CMMs can live.
For example, a programmer might use CMConvertXYZToLab to ask
ColorSync to convert CLE XIZ-based data into CIE LAB. There are APls
for almost anything you can think of: support for all of the profile classes.

308

Real World Color Manag-t,

Second Edition

conversions,status ofconversions reportingof profile locations and profile

information, and ways for applications to request user settings related to
s
color management-pretty much all the basic housekeeping s e ~ c erelated to dealing with profiles and conversions. ColorSync and ICM are primarily useful for programmers, whose lives are made easier by using APls
instead of having to write the code themsetves.
Users need relatively little interaction with ColorSyncand ICM. The one
user setting everyone needs to make is the display profile--most colormanaged applications ask the OS (using ColorSync or ICM) for that single
piece of user-supplied information. Other ColorSync and ICM user interface elements have less obvious effects.

Mac OS 9
On Mac OS 9. Colorsync3has two p i e c f f a system extension called "ColorSync Extension'' and a contml panel called "ColorSync." The only part
of the ColorSync Extension that might concern users is the Applescript
dictionary it contains, accessed by the ScriptEditor. (See Chapter 17. Alltomntion and Scripring.)
Thecontrol panel. on the other hand, is exclusively for users to manipulate (see Figures 11-1 and 11-21. Theconnol panel is misleading because
it implies that it will universally use the profiles specified under "Profiles
for Standard Devices" and "Default Profiles for Documents." Nothing
could be further from the truth. Tn fact, applications must be written to
request this information specifically from ColorSync. A few of the major
applications offer the option to use the "Default Profiles for Documents"
profiles, but we can't think of anythat do so by default, and we don't know
of any that use "Profiles for Standard Devices."
The ColorSync control panel represents a laudable attempt to create a
single point of access and consistent user interface for configuring color
management for all applications. It was great in theory but since all the
major applications aim for cross-platform parity, and ICM lacks any comparablefeatures, it didn't workout in practice. So youcan safebignore the
Profiles tab of the ColorSynccontrol panel entirely. However. two settings
are important.

Chapter 11: Color Management In the Operating System

ColorSync control panel:

Profiles for Standard
Devices

C0lorS)mc contml panel:

Default Proliles for
Documents

309

The only sofnuare we know

that uses these settings isAppleScript, specifically the scripts
that liue in rhe ColorSyncExtras:
AppleScripts folder Some of
these scripts let you use the
default profiles for devices as
variables, so you can quickly
change theproflles used bya
script by changing them here
rather than editing lhe script.
The "ColorSync Workflow"
option offered byAdobe
Photoshop, fllustrator;and
InDesign tlses these profiles,
as do theaforementioned
AppleScripts.We don't know
of anyrhingelse that doesso.

ColorSync Control Panel Settings

In t h e ColorSync control panel, under t h e CMMs tab, you'll find a single
setting for the Preferred C M M (see Figure 11-31,T h e default is Automatic,
which is functionally equivalent t o "random." I f you want subtly random
results, u s e this setting. Otherwise change it t o something else. We've
found t h e Apple C M M t o work well, and with t h e exception o fblack point

30

Real World Color Management, Second Edition

compensation, it provides similarresults to the Adobe ColorEngine (ACE)

found in Adobe applications.
If you're working in a cross-platform environment without the exclusiveuseof theAdobe applications (which can use ColorSync. ICM, or their
own ACE engine), you may want to standardize on a single CMM for all
systems. Since the Apple CMM doesn't exist on Windows, but the Heidelberg CMM exists on both. you can standardize on the Heidelberg CMM.
Likewise, if you have an important application or profile that uses the
Kodak Digital Science Color Management System, you can standardize on
the Kodak CMM. However, be aware that these CMMs don't exist for OS X,
so if an upgrade is in your future (and frankly, it's a matter ofwhen rather
than whether), you'll have to rethink your workflow for OS X, where the
only available 0s-level CMM is the one from Apple.

Monitor Settings
In the Monitors control panel, click on the Color bunon and you'll be
presented with a list of ColorSync profiles (ICC profiles) to select (seeFigure 11-4). Each display connected to your Macintosh can have a separate
display profile selected, but don't count on applications performing display compensation on anything other than the primary display. (Adobe
Photoshop 6 and7 are the only applications we know of that can perform
display compensation to multiple displays on Mac OS 9.)

This setting does two things:

b

If the profile contains the vcgt tag, which most do, the Display Managerupdates the video card LUTswith the correction needed to place
the monitor in the calibrated condition defined in the profile.

Chapla ll: Cdor Elknagement in the Opcratlng SWem

b

The selected profile becomes the displav profile for the current display

The effect of updating the video card LUTs happens immediately and is
usually visually obvious. Changing the display profile may have no visible
effect, but applications that query the system for the display profile use the
one you specify here for display compensation. Adobe applications are well
behaved in asking the svstem for the display profile, but most other applications fequire you to specify the display profile in their settings manually-watch out forthis potenlially major pidall.

Profile Locations
Profiles are usually 5toredinHard Drive:System F01der:ColorSync P~ofiles.
However, some older applications and printer drivers insist on finding
profiles in the old location. Hard Drive:System Fo1der:Preferences:ColorSync Profiles. Other applications may store profiles in subfolders in the
Hard Drive:System FolderApplication Support folder, though the ones of
which we're aware always place an alias in the ColorSync Profiles folder
to let other applications use the profiles. Since some applications don't
resolve aliases, and others don't look deeper than the root level of the
ColorSync Profiles folder, this doesn't always work, so ifyou have aprofile
that appears in some applications but not others, try placinga copy in the
root level of the ColorSync Profiles folder.

Location for CMMs

CMMs are system e.xk?nsions. You can find them in Hard Drive:Svstem
Folder:Extensions.

Mac OS X Panther (10.3.x)

ColorS!nc 4, thelarest version of ColorSync, debuted in a previousversion of
Mac OS Xand persists in OS 10.3 (Panther) and later. ColorSync 4 is aframework buried quite deeply within the ApplicationSe~ces.frameworkbundle.
Fortunately, this isn't a problem since, again, it's intended for programmers.
Color is handled in a very different way on Panther than on any other
operating system, and, l i e much of Mac OS X. is a work in progress. The
major applications still workas expected, but this operating system bhrrsthe

32

Real World Color Management, Second Editlon

lines between what is andis not a color-managed application. It doesn't blur
them heavily, but over time it will become increasingly easy for developers
to implement color management in their applications and printer drivers.
and increasingly hard for them to avoid doing so.
Mac OSX's color management rests on the premise that every window
is color-managed,whether the application that requests the window to be
drawn knows anything about ColorSync or not. This is currently achieved
through assumptions, and these assumptions are not always consistent.
Panther's color-management behavior in some cases seems more in
!lux and less stable than in previous versions. Some of these problems are
due to bad assumptions on the part ofApple,whileothers rest with device
drivers that don't work as expected. All we can do is report what we know
at the time we're writing this chapter, as it applies to Mac OS X 10.3.4.

Rules of Engagement
hlacOSXPantherhas always-on colormanagement, as previously stated.
What does thu mean? It literally means that the OS will perform a conversion no matter what, but the conversion could be what's called a null
transform. For display purposes, a null transform occurs when the OS
assumes the current display profile as the source profilefor an untagged
image or object, and also uses the display profile as the destination. Since
they are the same, no conversion occurs. Likewise when printing with
raster drivers. a null conversion occurs when Generic RGB is assumed as
source as well as destination.
Apple seems obsessed with the problems created by untagged files,
so they've embarked on a course of making assumptions on how to deal
wih them--even if the untagged file is a prematched print file on its way
lo the printer driver! Making matters somewhat worse, the application
of the assumptions in Panther is inconsistent, and even Apple's own applications don't always behave as expected. But here are some guidelines
to help put it into perspective.

Display. Non-color-managed applications have no say, so the operating

system assumes the display profile as both source and destination, which
means no display compensation occurr.
Color-managed applications query ColorSvnc for the current display
profile, and use it as a destination profile for display using either embed-

Chapter 11: C o b Management in lhe Operating System

313

ded profiles for tagged files, or a selected assumed profile for untagged
files as source. This is display compensation.
The key point is that the application determines the sophistication of
color management. At this point. ColorSync will bow out if it isn't otherwise asked to be involved by an application.

Input. For scanners and digital cameras, the profiles selected in ColorSync
UtiliQoDevices window are used only by applications that look for these
settings. Image Capture,which is both an applicationand a n N I provided
by Apple, looks to these settings as default Image Capture allows you
to override the default within the application itself. Your manufacturerk
scanner driver invariablyignores this, and for capture purposes ColorSync
is uninvolved, again, unless requested by the application or driver.

Print. When you print from any application using a raster driver, and
when you print from applications that do not generate their own PostScript stream using a PostScript driver, a PDF spool file is generated. The
application sends drawing commands to the OS, which in turn produces
the PDF spool file.
b

\men the applicationsends untagged RGB, the OS embeds theGeneric

RGB profile.
When the application sendsuntagged gayytheOS embeds theGeneric
Gray profile.

When theapplication sendsuntagged CMYK, it remains untagged, for

the momenl.

The behavior of the print driver becomes important at this time, because its settings determine the destination profile ColorSync will use.
PostScript drivers are fairly straightforward, and are discussed later in
this chapter.
Raster drivers are the main concern because ColorSync affects them
no matterwhat. Ifyou select aproprietary color management option ( i . ~ .
anything but ColorSync),then ColorSync assumes Generic RGB for the
destination profile. The result is a null transform when applications submit untagged RGR data.

S14

R u l World Color Management, Second Edition

If you select ColorSync, then theoreticallythe profile set for the printer
in the ColorSyncUtilitywillbe used as destination. More about this can be
found in the section on Raster Drivers later in this chapter. The key point
is that ColorSync invites itself at print time evenif neither the application
nor the print driver has explicitly asked for it.
The scary part is that the handling of prematched data from a colormanaged application such as Photoshop, set to convert your image to
SP2200 Premium Luster.icc, only works because of a sequence of events
that causes Generic RGB to be the source profile and the destination profile, and hence a null transform. The data is clearly not Generic RGB, and
the print behavior is clearly not Generic RGB either, yet that is the philosophy of null transforms on Mac OS X Panther.
Fortunately, color-managedapplications thal prematch data don't tag
files, which means the OS will tag it with Generic RGB. And so long as you
don't select ColorSync in the print driver, Generic RGB will be used as the
destination profile also. The end result is ColorSync is a oon-factor in
the conversion. But clearly Apple needs a strategy for explicitly handling
prematched print data, since all professional graphic arts applications
can produce it. If you're wondering about those generic profiles, see the
sidebar. "The Generics."

Displays Panel Settings

This panel is found in System Preferences, and is functionally equivalent
to the Monitors control panel on Mac OS 9. Click on the Color tab, and
select the profile for your displays here. As on Mac OS 9, multiple displays are supported by the operating system, but also as with OS 9, you'll
need to determine whether or not your specific applicationsupports color
management for two separate displays. (See Figure 11-5.1

Chapter 11: Cdor Management In the Opefalng Spiem

Mac OS 9 and X come with two

generic profiles: Generic RGR.
and GenericCMYK. Makeaspecial
notewhereveritwillstick, because
they are different!
On Mac OS 9 Generic RGB is
based loosely ontheP22 phosphor
set. a 9300 K whitepoint and a 1.8
gamma--the venerable Apple 13inch RGR display On Mac OS X,
Generic RGB is based exactly on

K whitepoint, and a 1.8 gamma.

On Mac OS 9, Generic CMYKiq
based on the Apple Color Laserwriter. Ever wonder why you'd get
such disgusting results using it as
sourceordestination for apress or
inkjet printer? Now you how! On
Mac OSX, Generic CMYK is based
on sub-sampled TR 001 measurement data, soit'sessentially SWOI?
PZ2 phosphors, a 6500

a5

Don't expect ir ro do a gear job on

a CMYK inkjet or color laser printer though, as the output will likely
look washed out. For press work,
its marginally acceptable, but we
recommend using the US.Web
Coated (SWOP) v2 profile that
comes with Adobe Photoshop.
InDesign, and Illustrator. Or visit
www.profilecentral.com for more
profiles based on TR001.

Profile Locations
Mac OS X stores profiles in a variety of locations, and in order for it to
make any sense, a bit of background on Mac OS X is necessary. Mac OS S
is designed as a multiuser operatings)stem.You may only have one user,
but the system is still designed for more than one. Thus, the locations
for profiles follow the general rule that there are five possible "domains":
Apple only all users, a specific user, a specific application, and network
For practical purposes we list just the first three:

/SystemlLibra~lColorSynclProfile~-All
userscan use profiles stored
here, but they can't be removed or supplemented unless you have a
special set of keys. Since this is Apple's domain, nothing except the
Mac OS installer should remove profiles from or add profiles to this
location

lLibnrylColorSpclprofilesAUusers can use these profiles, but only

Admin usersand authenticated installers w n addprofiles to or remove
profiles from this location. The preferred location for display profiles is
in the Displays folder contained within this location, although thev'll
wvorkif placed in the other legitimate locations.

/Usersl<usemame>lLibrary/ColorSp~/Profile~Profiles
in this location are only available when this user is logged in, and aren't available to other users; only thisuser can add profiles t o o r remove profiles
from this locarion.

3l6

Real World Color Management, Iscond Edition

More than likely you'll want profiles available to everyone, so we recommend placing them in the IUbrar)rlCololSynclProfilesfolder. If you're
working in an administered environment where you don't have permissions for this location. place them in your user ColorSync Profiles folder
instead.
Another major difference with MacOSXis that manufacturer-supplied
'.cannedmprinrer profiles don't normally go in any of the above locations.
Instead, they'restored asresources within the printer modulebundleand
as such aren't directly accessible by users. However, they are available to
all applications just as ifthey were stored in ILibrarylColorS~clProfiles,
but they don't clutter the various Profiles folders with profiles for which
professional users may haveno use. The ColorSync Utility, discussed later
in this section. has a Profiles feature that, among other things,shows locations of currently available profiles.

Locations tor CMMs

CMMs live in ILibra~IColorSynclCMMs.

ColorSync Utility
The ColorSync Utility has evolved from its lowly beginnings as a slightly
beefed-up system preference to a fairly handy toolkit. It incorporates the
old ProfileFirst Aid utility, and addssome nifty featuressuch as3D gamut
comparisons of profiles, and the ability to view the individual tags inside
profiles.

Preferences. The preference settings for ColorSync are no longer found

in System Preferences. Everythingis now rolled into the ColorSyncUtility
found in the lApplicationslUtilities folder (see Figures 11-6 and 11-7).
Here, you can (supposedly) specify default RGB. CMYK, and grayscale
profilesfor documents that don't contain embedded profiles.Youmay be
tempted to conclude two things from these settings: embedded profiles
are recognized svstem-wide, and documents without embedded profiles
will have the default profiles assumed as source profiles. You'd be wrong
on both counts.

C h a m lk C o b Management in the Operating System

317

QuickDraw, GDI, and Quartz

Each of the three operating systems discussed in this chapter, Mac OS 9, Mac OS X and
Windows, has its own display
and print engine. On Mac OS 9,
this is QuickDraw, on Mac OS X
it's called Quartz, and GDI is for
Windows.
QuickDraw and GDI have
something in common. They
have no idea what CMYK is all

n g u n 11-6

Mac OS X ColorSync
panel: Default Profiles tab

Mac 0s X ColorSync
panel: C M M s tab

about, so non-PostScript printer drivers for these operating

systems must receive RGB data.
Quartz is different. It's PDFbased, and as such it knows about
CMYK and LAB as well as RGB.
There's been some speculation
that Mac OS X printer drivers
could accept either RGB or CMYK
data. It's theoretically possible
starting in Mac OS X 10.3 (Pan-

ther), because its PDF to raster

RIP does support CMYK. However, manufacturer printer drivers that convert raster data into
printer commands for generating droplets of ink (screening algorithms) still expect to receive
RGB data. That may change in
the future, but if you're holding
your breath, we suggest you exhale now.

38

Real World Color Management, Second Edition

MacOSX, likeOS9 andwtndows, stiurequiresdevelopers tospecificallv

write code into their applications to make them recognize these settings,
as wellas to recognize embedded profiles. Anapplication that isn't written
this way ignores these preferences,and if it doesn't knowwhat an embedded profile is, it opens the image as though it were untagged.
Apple has claimed it was never their intention to imply that the Default
Profiles settings would be used system wide.Why they chose language that
pretty clearly indicates this would happen, we don't know, but it seems like
it's not a very good idea, and we hope the wording will finally change in the
next version of Mac OS X
The advice we give under Mac OS 9 for the setting of the Preferred CMM
applies here as well. If you like random results, use the Automatic option.
Otherwise, picksomething else (atthe timeofwriting, theonly something
else that's available is the Apple CMM). Interestingly enough. Apple also
recommends the Apple CMM in their "Color Management with Mac OS
X Panther" documentation, yet the default is set to Automatic.
Profile Errt Aid.This is a utility for verifying the integrity of K C profiles
stored on your system, and can repair most problems it finds. It looks for
aspects of the profile that don't conform to the ICC specification version
under which they were written [i.e.. it applies v.2 rules only toe2 profiles,
and v.4 rules to v.4 profiles). One of the most common side effects of a
non-conforming profile is that it doesn't appear in application pop-up
menus, which, of course, means you can't use i t
Running a repair session in Profile First Aid. then relaunching the application frequently solves such problems. We recommend running a
Repair session regularly because a surprisingly large number of profiles
don't conform full to the ICC spec. Most deviations are benign, but so is
running Profile First Aid.
Pro(ilrr.'l'he Profiles window shows currently installed ICC profiles. Here
you can view basic profile information to the right, and even click on the
blackarrow on the upper-left comer of the Lab Plot, to find more options.
You canview the colorspace in3D in a number of color spacesother than
LAB, as well as perform gamut comparison between profiles. (See Figure
11-8.) If youdouble-clickon aprofile, it opens in asepantewindow showing information about each tag in the profile. (See Figure 11-9.)

ColorSync Utility allows you to compare thegamuts of two profiles as a 3 0

graphic. The above enample shows sRGB (in color) plotted againstlapanWeb
Coated [MI in white.

mure n-9

ColorSync Utility:
Profile Info

Double-clickinga profile in the window

shown in Figure 11-8 opens the window
shown at left.You can click on each tag
to view information about it. Here, the
red channel of a display profile's 'vcgt'
tag [which loads calibration settings
into the video card) is shown.

Devices. Developers can write their device drivers to register themselves

as ColorSync devices. Doing so allows them to appear in this window,
with possible sub-categories such as media type, or color model. Each
sub-categorycan have a profile associated with it. Developers can specify
default profiles to use for their devices, and end users can select the arrow pop-up menu to change the Current Profile to a custom profile, for
example. See Figure 11-10.

UCN 11-10

Mac OS X ColorSync
Utility,Devices window

What does this do exactly?If the application or driver you use doesn't
specifically look to these settings, they do nothing. We can confirm that
changing the Current Profile for a display device has the same effect as
selecting it in the Displays panel in System Preferences. If you use Apple's
Image Capture application, changing the Current Profile for your listed
scanner or camera will change the default profile used for embedding in
captured images.
For printers it's dicey. The idea is that if you tell your printer driver to
use ColorSync, the Current Profile selected will be used as destination.
The problem is that not all drivers play along with this rule, which seriously undermines the usefulnessof print driver level ColorSync support.
To date we know of no Epson printer driver that honors custom profiles
selected in the ColorSyncUtility,and Apple was unable to provide us with
information regarding other vendors' level of compliance. See the section
on printing later in this chapter.
Filters (also known as Quartz Mters).We really hope Apple works on
this feature because it sounds like a great idea, but due to non-standard
terminology, an ill conceived user-interface,and bugs, its real-world useMness is lacking. Filters can be used directly in ColorSync Utility itself,
or in the standard print driver under the ColorSync>Quartzfilter pop-up
menu. See Figure 11-11.

Chapkf ll:Cdor Management in the Operating System

3P1

For example, Chris created a filter set to "Convert color data to," which
is Quartz Filter terminology for "Destination profile." Opening a PDF in
ColorSync Utilitywith this filter, and checking the Preview checkboxin the
upper-left comer of the resulting window showing the document, would
make ColorSync convert all objects in the PDF to SWOE and then display
them on-screen. Similarly, when printing from an otherwise non-colormanaged application, selecting this Filter in the ColorSync portion of the
print driver would first convert the document to SWOP, and then ColorSync color matching would convert again to the profile for the printer
you're printing to. In theory.
In practice, as previously mentioned, many inkjet printer drivers don't
play along well, which makes hard proofing a game of chance. Other

329

Real World Color Managanenl, Second Edition

drawbacks include the lack of black point compensation and rendering

intent control to the would-be proofing device. Known bugs at the time
of writing include:
The default tab specifies profiles for untagged objects. but no matter
what you select only the Generic profiles are used.
b

At least two and possibly three destination profile pop-up menus are
possible.
The clauned PDFIX-3 support doesn't pass preflight in Acrobal6. Pitstop Pro. PDFIX Inspektor,or PDFIX Checkup: and the OS ignores the
OutputIntent in existing PDFIX-3 documents so you can neither soft
proof nor hard proof them. We aren't sure where or how PDFIX-3 is
supported in OS XPanther.

Windows
ICM 2.0 (ImageColor Management) is built into Windows 98 and higher,
includingWindowsXP.Like ColorSync, ICMisasetofAPIs intended for use
by programmers. Outside of printer drivers, and applications specifically
written to m e ICM 2, there's only one user setting, but it's an important
one to which you should pay special attention.

Setting the Active Color Profile for the Display

Each device on Windows can have multiple profiles associated with it, as
wellas adefault profile. Only thedefaultprofle is the active profile for that
device and is the one that gets used. To associate profiles with a device,
and to set the default profile, go the device's Properties panel, click the
Advanced button under Settings, then click on the Color Management
tab. In that window, you can associate profiles with the device and also
specify one of them as the default profile.
For example, to set the display profile, go to the Display control panel,
click on the Settings tab, then click the Advanced button. In the resulting
window, click on the Color Management tab, click Add to open the Add
Profile Association dialog box, then choose the profile and clickAdd.To set
the profile as the default, select it in the Color Management tab, and click
Set As Default (see Figurr 11-12 and Figure 11-13).

Chapter H: Color M s l l p w t in the Opaallng S@em

Apple's Preview application is useful for Previewing TIFF.JPEG, and

PDE. It honors embedded profiles
as source profiles, in file types that
support them, but it actually converts files to Monitor R G S t h e
current dsiplay profile-upon
opening them. While the files
aren't saved to disk at this point,
as far as Preview is concerned, it's
working on data that has already
been converted.
To make matters worse, this
problem is compounded when

Figure n.m
Display Properties:
Settings tab

Figure n-u
Display Properties,
Advanced: Color
Management tab

you save the document from Review, such as you might do when
you wanted to change formats.
TIFF and JPEG are left untagged.
while PDF is tagged with Generic
RGB, and JPEG 2000 is tagged
sRGB. The idea of converting the
data automatically on opening is
had enough, but to then embed
obviously incorrect profiles is
even worse. AU these behaviors
have been confirmed as bugs.
For untagged documents, Preview assumes Monitor RGB (i.e.,

341

the current display profile) as the

source profile. When displaying
the document, this means the
destination profile, also Monitor
RGB, is the same as the source,
and no transform occurs. When
printing the document, the full
sequence of transforms (it can be
more than one) depends on your
print settings.
Preview can preview, but if you
care about quality, don't rely on it
to save any documents that you
care about!

3%

Real World Color Management, Second Edltlon

On Windows, you need to be extra careful in setting the active display
profile. Unlike the Mac OS, Windows doesn't get calibration information
from the profile to update the video card LUTs, so it's all too easy for the
monitor to behave differently than the way the active profile describes.
Display calibrationlprofiling packages should set the display profile automatically when it's created, bur if in doubt. double-check to make sure
that it has in fact done so.

Locations tor Profiles

Where your profiles live depends on what version of Windows vou're using, and whether or not you chose the most typical installation [on a C:
drivel. You don't need to worry about these locations except for doing
housecleaning, because vou can right-click on a profile in any location
to get a contextual menu containingan option to "Install Profile," which
will place it in the appropriate directory

Windows 2000,XP--C:\W1nNT\System32\SpooI\
Drivers\Color

NOTE: Windows NT doesn't support ICM 2,and has no built-in color

management at all. Any color management happening under Windows
NT is done entirely by applications, and is hardwired into them. So the
location for profiles underwindows NT is merely for reference.

Color Management in Applications

In the introducrion to this chapter, ive ment~onedthat color management
can be requested by either applications or printer drivers. This section
looks at colormanagement strictly from an application point of view, covering both color-managed and non-color-managed applications.

Color-Managed Applications
Color-managed applications knowwhat ICCprofilesare, anduse them to
at least some degree. Most of these applications use either ColorSync or
ICM to do color-management tasks such as color space conversions.

Chaptu n:Color Management In Lhc Opmtlng Syslcnr

395

Some color-managedapplicationsimplement their own color engine.

such as the Adobe applications, all of which use ACE (the Adobe Color
Engine). Within each application. ACE produces the exact s a m e results
between its Macintosh a n d Windows counterparts. However, w e n when
usingACE, they still make minimal use of ColorSync and ICM to requesr
the current display profile and to find profiles o n the system. All conversions are performed, bv default by ACE. It's possible to select ColorSync
or ICM (and thus any available CMMs) in lieu of ACE in most Adobe applications. More detail o n this can be found in Chapter 12, The Adobe
Comnwn Color Archifectzire.
Note that color-managed applications don't all have identical capabilities. While they'll support colorconversions in one way oranother, not all
of them support color management to the display (such as many scanner
drivers1 o r fully support embedded profiles.

CUPS
CUI'5 stands for the Common
Unix Printing System. Apple
ha%touted CUPS to the graphic
arts community as a good thing,
but hasn't done a very good
job of telling us why we should
care. The printing architecture
designed for Mac OSX as embodied in Mac OS 10.0 and 10.1, had.
let's just say,lots ofproblems.
Sa it got gutted in favor of CUPS.
CUPS is primarily a scheduling
and queuing system, handing off
print jobs from applications to
appropriate filters, drivers, and
printer modules. It's the center of
the print architecture, and therefore defines i t
Nowwehavetwoldndsofprinter
drivers. We have the original architecn~re'sPrinterModules,forwhich
Apple created a bridge so that

CUPS could use them without tequiringdevelopers to revvrite their

drivers from scratch lhese Printer
Modules can be ColorSync aware.
and do driver-level color management But now there are also CUPS
printer driversand filters.
Filters do conversions between
formats. For example,a PostScript
to PDF filter converts PostScript
intoPDE and a PDF to Raster filter
converts PDF into aTlFF or other
bitmap. A PostScript RIP filter
would convert PostScript into a
bitmap too, lhnugh we've vet to
see one.
CUPS drivers are responsible
for communicating with a particular printer in a language it
can understand. An Epson Stylus
Color 3000 printer driver would
convert raster bitmap data into

the lnt~guagethe printer needs.

accounting for media, resolution
and uther settings.
The good news? CUPSefictiwly
turns any printer into a Postscriptand PDF-based printer.Theplug-in
filter-based architecture lets us insertdiffe~ntkindsofRIPsfromvarous wndors, and use these RIPs for
any raster-based printer comected
to the host computer.Stay tuned for
moredevelopmentson this front as
CUPS matures and printer wndors
decideto huildCWSdriwminsread
of Printer Modules.
Interesting news? Starting with
Panther, CUPS drivers automatically register with ColorSync, so
CUPS printers' "Current Profile"
settings In the ColorSync Utlllty
are honored by ColorSync,and will
beused asadestination profile.

396

Real World Color Managcmcnt, Second Edition

The sure way to distinguish a color-managed application from a noncolor-managed application is to snoop around in the program itself.
Color-managed applications have preferences for configuring source,
destination, and possibly display profiles (ifthey don't request thedisplay
profile from the operating system).
So where is the color management actually happening? From a programming point of view, it's in the application. While ColorSync or ICM
may actually be doing the work. they're APls, not standalone programs.
Think of it like this: ifyou drive a friend to the airport, did you or the car
do the work?Wen, both.

Source profiles. Color managed applications can hand off the source
profilefor yourdocument to the operatingsystem at print timealongwith
the data to be printed. but with the exception of the Adobe applications,
this seems to be fairly rare. The point of doing so is to provide the source
profile(s),and let the printer driver and operating system do any subsequent color management. However, color-managed applications can
already do the conversion to print space whiie producing the print stream,
so most of them don't include source profiles in the print stream.

Nan-Color-Managed Applications
Previous examples of non-color-managed applications are Microsoft
Word, Excel, and Powerpoint, as well as most Web browsers. However,
this has become a little more blurred on both operating systems' latest
versions of these applications. Microsoff Office applications honor embedded profiles for preview. Some applications pass source profiles onto
the OS at print time: others convert to Monitor RGB first. then submit the
print job tagged with the current display profile.
Non-color-managed applications send the RGB values in the file directly to the monitor, so the sensible assumption for a source profile for
filesgenerated by these applications is the display profile,since it defines
the color the monitor produces in response to the RGB values.
To colormanagefiles From these applications,export them asTIFE JPEG.
or as a Postscript file produced by printing to disk, then import them into
a color-managed application, assign the current display profile, and resave
with the display profile embedded. From that point onward, the files are
color managed.This technique works for screen shots, too. (Mac OSX Panther screen shots are embedded with the current display profile.)

Chapter 11: Color Management In the Opentlng System

397

Color Management in Printer Driven

At first glance,it mayseem that the printerdriveris whereallcolormanage-

rnent destined for print was intended to occur. Applicationswould manage

color within the application and pass on a source profile for their documents at print time, then the printer driver would specify the destination
profileand the operatingsystem would do the necessaq7conversion.
The truth is that printer drivers have no knowledgeof the applicationthat's
sending the data, so they have to handle both tagged and untagged files.
Howwer,tbe assumptionsthatdifferent printer drivers make as to the source
profile van; widely,and in our experience, they're usually incorrect.
Meanwhile, color-managed applications generally want to color manage the content of their documents completely before passing it off onto
the operating system for printing, sending "pre-matched data." But the
printer drivers still expect to do color management at their end, and to
date there isn't a mechanism for the application to tell the printer driver,
"Hey, buddy don't do anymore colormanagement!"That means you, the
user, have to ensure that when you print from applications that do their
own color management, either the printer driver color management is
turned off or theapplicationdoesn'tconven to printer space before handing off the data to the printer driver.
With non-color-managedapplications,you can use printer driver color
management to perform the conversion to print space. but doing so involves some pitfalls that we discuss below.

Default Behavior
Each rnanutacturer seems to decide a different default behaviot lor their
printer driver, and sometimes they decide on different defaults for different models orchange the default behavior betweenversions of the driver.
Most raster drivers default to proprietary color management rather than
to ColorSync or ICM.
PostScript printers also vary their default settings, but we know of none
that default to Colorsyncor ICM-at least not directly.They almost d l use
Postscript color management when receiving RGB data, to convert it into
CMYK in the printer itself. Some also use Postscript color management
when receiving CMYK data. The printer driver option that controls this
is usually called something like color correction color rnnmgement, or
simrtlation.

328

Real World Color Management, Second Edition

When you use driver-level color management, the settingsyou usewhen

you print the target to profile the device are very important. For example.
when you use ColorSync or ICM in Epson printer drivers, what actually
goes on under the hood is that the driver uses the "No Color Adjustment"
setting, and asks ColorSync or ICM to do theconversion to the print space.
If you print the profiling target using the "Photo-realistic" setting to take
advantage of its better linearization and gray balance (putting up with the
smaller gamut), the resultingprofile won't workwellwth ColorSyncor ICM
selected because the printer driver won't use the "Photo-realistic" setting
when you choose ColorSync or ICM.

Source Profile Assumptions

When theapplicationdoesn't include thesource profilein theprintstream,
and you use ColorSync or ICM in the printer driver, a source profile has
to be assumed. Two things can make the assumption: the printer driver
itself, or the operating system.
In Windows. the assumed source profile is always sKGB, whether the
printer driver orwindows itself is doing the assuming. .4pplications that
don't generate theirown Postscript must use GDI, which has no ideawhat
CMYK is. So the application must convert CMYK content into RGB for
either displayor for print. Afew applications use GDI+, which will assume
a default CMYK SWOP profile.
OnMac OS9, the operating systemassumesthe currently selected display profile, although an increasing number of printer drivers tell ColorSync to use sRGB as the source instead. To find out which assumption
your printer driver makes, print an image with a display profile selected,
then select asubstantiallydifferent display profile (orperhapsevenanondisplay profile such as Wide Gamut RGB if you have Photoshop installed)
and reprint the image. If the test prints look the same, something is being
assumed over which you have no control, probably sRGB. If they'redifferent, the driver is assuming the display profile. CMYK behavior is similar
to Window.
On Mac OS X Panther, Generic KGB is assumed system-wide for untagged RGB at print time. As we discussed earlier in the chapter if the
application does not specify a profile, Mac OS X embeds the PDF print
spool file with Generic RGB. For CMYK, the waters are muddy because
legacy applicationsstill use QuickDraw,so they still convert their content

Chapter ll: Color MmsgrmenlIn L c Operating System

329

to RGB at print time, and this conversion is up to the developer. However,

applications using Quartz can send CMYK data for display or print. Currently Generic CMYK is the assumed source.

Raster Drivers
Usingprinter driver color managementwith raster printers and non-colormanaged applications boils down to choosing a destination profilethe
source profle is largely out of your control as it's assumed by either the
application, printer driver,or operatingsystem.Where you choose the destination profile depends on the operating system.
We can't possibly cover all printer drivers, so let's use the example of
printing to an Epson Stylus Photo2200from Microsoft PowerPoinl, atypical non-color-managed application. You should be able to find similar
terminology in other printer drivers. If you don't find a ColorSync or ICM
option in your printer driver, then it almost certainly doesn't support ICCbased color management.

Mac 0 s 9. When you open the printer driver, the default behavior is set
to Automatic. To access the ColorSync option, you must select Custom in
the Mode portion of the driver, and click the resulting Advanced button
(see Figure 11-14).

Mac OS 9 Epson Stylus

Photo 2200 main dialog

in the resulting dialog, look in the Color Management section.The default setting is for the Color Controls radio button, so select ColorSync
here (when you're printing from an application that does its own color
management, you'd select the setting you used when making the profile
For this printer, usually No Color Adjustment).See Figure 11-15.
Epson drivers select aprinterprofile based on the mediaand resolution
settings you've selected. Some printer drivers allow you to select a specific

330

Real World Color Management, h

d Edition

profile here manually but in nearly all cases it must be the canned profile
supplied by the manufacturer. We haven't had much success in getting
custom profiles to appear in Mac OS 9 printer drivers, even by replacing
the profile with an identically named profile.

Mac OS X Things are quite a bit different in Mac OS X, starting with the
print dialog (see Figure 11-16),The third pop-up menu itemlists different
views, and the one you want is Color Management. Select ColorSync,and
you'll see something like Figure 11-17.
~

EMac OS X Epson Stylus

Photo 2200 Print dialog

~~~

~-

-.

'
me: ! **"dad
*1
hlna.:

-!?&

--*

Tip: Use No Color Adiustment.When printing from applications that let

you select a destination profile for your print, choose No Color Adjustment (Epson) or equivalent in your print driver to ensure double-color
management does not occur. If you are printing from a non-color-managed application, use either ColorSync or the driver's proprietary color
management, whichever gets you results you prefer.

Chapter n:Color Managsment In the Operating System

Figure 11-17

33l

Mac OS XEpson Stylus j

Photo 2200 Print dialog:
Color Management
options

Where are the printer profiles? OnMacOSX, each device registers a device profile. Some printers, such as the Epson Stylus Photo 2200, register
a profile for each media type they support. You don't need to do anything
unless you don't want to use the registered profile for the media you're
using. The ColorSync Utility, discussed earlier in this chapter, is how you
theoretically can override the registered profiles. In some cases this works,
and in other cases it doesn't. With native printer drivers, it's hit or miss,
and with Epson drivers it appears to be most often a miss. If you are using
PostScript driversthat registeryour printer, or Gimp-Print drivers, custom
profiles set in the ColorSync Utility generally seem to work.
Mac OS X 10.3 print dialogs contain a new "ColorSync" menu item,
where you can access Quartz filters, mentioned earlier, as well as a Color
Conversion menu. For raster printers, the only option available is Standard. See the "PostScript Drivers" section for more information on this.
The version of OS X that was current when we wrote this lacks any
mechanism for selecting arenderingintent to use with printer driver color
management. Once you've selected the desired profile for the intended
media type in the ColorSync Utility, you simply print your document.

Windows XP. Windows printer drivers share similarities with both Mac
OS 9 and Mac OS X drivers. Using the Epson Stylus Photo 2200 as an examole again, in the main mint dialoe. click on the Advanced button (see

332

Rul World Color Hanagancn(, Second Ednlon

Figure 11-18). The resulting window contains a Color Management section where you select ICM (see Figure 11-19),As in Mac OS X, there are no
options--they're determined from alist of associated profiles.We like that
the driver displays the profile that will be used by the OS as the destination profile. Also, note that onWindowsXP you need to first select ICM if
you wish to select No Color Adjustment, such as when using application
level color management.
ngure n-18

Windows XP Epson Stylus

Photo 2200 Print dialog

Window6 XP Epson Stylus

Photo 2200 Print dialog:
Advanced

To associate profiles with a device, go to the Start menu>Printers and

Faxes. Here we right-click Epson Stylus Photo 2200 and select Properties,
as shown in Figure 11-20, which brings up the Properties dialog box as
showninFigure
11-21.
Here we click on the Color Management tab to open the dialog box
shown inFigure 11-22. These profile associations,and the default profile,
are device-specific, not media-specific.

Chapter lk Cdor M
a
-

Windows XP Printers and

Faxes dialog box

n g u 11-91
~

Epson Stylus Photo 2200:

Properties,General tab

nwre n-PP
Epson Stylus
Photo 2200: Properties,
Color Management tab

In the Operating S p k m

333

334

Real World Color Management, Second LdlIlon

PostScript Drivers
ColorSync and ICM can affect PostScript output, even though neither actually knows what PostScript is. Non-color-managed applications don't
generate PostScript themselves--they send drawing commands to the
operating system, and the PostScript printer driver turns those QuickDraw, GDI, or Quartz commands into PostScript. The operating system
can color manageQuickDraw,Quartz, and GDI/GDI+routines, then produce the PostScript stream already color managed. The latest Mac OS and
Windows operating systems can also produce CSAs (Color Space Arrays,
which are PostScript source profiles). Note that applications that generate their own PostScript (which virtually every color-managed application does) aren't affectedby ColorSynclICMsettings, so PostScript printer
driver color management simply doesn't work with these applications.

Mac 05 9. The MacOS 9 PostScript driver is afrequent and fruitfulsource

of confusion due to the ColorMatchingsection,which contains ColorSync
Color Matching (and, for some printers. PostScript Color Matching). See
Figure 11-23.

Mac OS 9 Laserwriter8:
ColorSync Color

Matching

Bothoptions only workwith applications that don't generate their own

PostScript.Why a mechanism was never developed for the application to
better communicate its capabilities to the printer driver isn't something
we can explain, so let us just stress that with all of the major applications
we cover in laterchapters, these printer driver options don't apply because
all of the color-managed applications generate their own PostScript.
But if you're printing from an application like Powerpoint or a Web
browser, you can use the ColorSync Color Matching option, which works

Chapter ll: Color Management in the Operating -5

335

the same as for raster printer drivers, with the exception that the current
display profile is always the assumed source profile. You simply choose
a destination profile, which normally should be the profile for your intended destination PostScript device.
PostScript Color Management is an option we don't recommend, both
becausewe know of very few people who have madeit work, and because
there are precious few tools to help anyone do so. The way it's supposed
to work is that the assumed source profile (the current display profile)
gets converted by ColorSync into a Postscript CSA (Color Space Array).
The selected destination profile gets converted into a PostScript CRD
(Color Rendering Dictionary),and both are sent with the 0s-generated
PostScript print stream to the printer. The Postscript RIP uses the CSA
and CRD as source and destination profiles, respectively, and makes the
conversion. If you get it to work and it doesn't cause you confusion or
misery, feel free to let us know.

X (lO.9). The PostScript driver on Mac OS X has absolutely no

Mac
user options for color management, apparently by design. Applications
that generate their own PostScript simply bypass any additional influence
by the operating system, anyway. For all other applications, a PostScript
CSA (a Postscript source profile) is placed in the print stream by the operating system based on the foUowingconditions:
b

Tagged images.The embedded profile is usedas the source profile, so

it becomes the CSAsent to the PostScript printer.

Tip: It's Only Tagged If ihe AppllcaMon Knows It's Tagged. ~fthe application being used to print doesn't know what an embedded profile is,
the image isn't considered tagged by the operating system.
b

Untagged IDeviceRGB. IDeviceCMYK, IDeviceGray.Withapplications

that generate device-dependent color, a CSA is generated from the
Generic RGB. Generic CMYK, or Generic Gray profiles.

Untagged lCalRGB or ICalCMYK. Applicationsthat produce calibrared

color get a CSA generated fmm the calibrated color tables submitted
by the application.

3%

Real World Color Managcmcnl, kcond Edition

There are two gotchas inherent in printing under Mac OS X 10.2 and
earlier from applications that don't generate their own PostScript to a
PostScript printer. First, there's no way to avoid sending a CSAto theprinter, so there's no way to avoid PostScript color management (unless your
RIP simply ignores CSAs, which is relatively rare). So don't print test files
From such applications. Second, the destination profile is almost always
the printer's built-in CRD, which is usually a generic placeholder rather
than an actual description of the printer's behavior.

Mac OS X 00.3).Mac OS X Panther has added a ColorSync menu option

to all printer drivers. It contains theaforementioned Quartz filters menu,
and also a Color Conversion menu. Forraster drivers, the%-Printer" option is grayed out because it can't occur in the printer itself.
The "Standard option is grayed out with PostScript drivers when you
access the driver-level ColorSync menu item from an application that
generates its own PostScript. This is because the OS can't color manage
application-level generated PostScript. Finally, there is some sensible communication between applications, OS, and drivers. See Figure 11-24.
Figure n.4

Mac OS X PostScript
driver: ColorSync menu's
Color Conversion options

Chris wondered if the dog, the pigand the pony would have to squaredance under a full moon and shake hands afterward to start seeing such
examples of cooperation. However, technically the "In-Printer" option
should not apply either unless PostScript color management is selected in
the application. There's still room for improved communication between
the parties, but rest assured that the OS will not modify PostScript print
streams from applications that generate them.
For PostScript printers the option is either "Standard" or "In-Printer."
In-Printer is the behavior we were stuck with in 10.2, with one exception.

Chaptar 11: Color Management in the Oparatlng System

If you have untagged CMYK content, also known as IDeviceCMYK, the OS

will allow that to pass through without Postscript color management.
Standard causes ColorSync to do all conversions prior to Postscript
being generated, andthen sends essentially untagged IDeviceCMYKdata
to the printer. The destination profile is set in ColorSync UtilitpDevices
window. Ifyour printer registers itself with ColorSync, this method of color managing output from otherwise non-color-managed applications will
be faster, and superior in quality than relying on Postscript color management used with the "In-Printer" option.
Eventuallywe expect Apple to provide a mechanism for sending CRDs
based on ICC profiles,which should provide reasonable results from PostScript Level2 or later RIPS.But forthe time being, ifyou want good output
from a PostScript device, print from an application that allows you to
control the source profile, destination profile, and rendering intent.

Color Management in Scanner Drivers

Most scanner drivers implement their own proprietary and rudimentary
color management by default, and some also tie into the operating system. Moreover, scanner drivers that do offer color management exhibit a
Fairly wide range of behaviors and capabilities. See Figure 11-25.
Figure n-PS

Epson Perfection 3200

Color Configuration

The best drivers let you specify a source and a destination profile,
convert the data from the source to the destination during the scan, and
embed the destination profile in the scanned image on save, whiie also

338

Real World Color Management, kcond Wilion

using the display profile to display the image correctly on the monitor.The
worst ones have a single button marked ColorSync or ICM that converts
from a hidden generic scanner profile to thedisplay profile, and produces
untagged RGB as output. Most fall somewhere between the two extremes,
but the most common problem is the inability to let the user specify a
custom scanner profile as the source profile. In most cases, we find that
the easiest coulse of action is to find out how to turn off both color management and any autocorrection features in the scanner driver, treat it as
dumb and raw, bring the scans into Photoshop, and assign the scanner
profile there. Some scanner drivers make this procedure more difficult
than others, but it's rarely impossible.
If you decide to use scanner driver color management. you need to
confirm whether the driver embeds the scanner profile, or uses it as a
source profile for a conversion to some destination profile. If the latter,
what is thedestination?Ifthere7sno explicit setting for the destination, it's
prohablythe display profile, which the scanner driver grabs automatically
from the operating system. That means that allyour scans get clipped to
the gamut of your displa): which is far from ideal.
Scanner drivers that operate as Photoshop plug-ins need special attention, because theAPI that lets thedriver passasourceprofile to Photoshop
along with the pixels hasn't been implemented in any scanner software
we've encountered, so the images always come in as either untagged or
working space RGB. If the scanner driver actually converted the pixels to
working space RGB, that's fine. If it didn't, you need to assign the scanner
profile manually.
Using the Epson Perfection 3200 Photoshop plug-in driver as an example, note that by default proprietary color management is used. If you
select ColorSync the driver will ask ColorSync to use the specified source
and target profiles to convert the file. Remember, after conversion what
was selected as target (destination profile) becomes the source profile
for the scanned data. Be aware many of these Photoshop plug-in drivers
do not inform Photoshop of the proper source profile.They may open in
Photoshop "Untagged"or tagged with the RGB WorkingSpace profile. Either way this is incorrect,you will need to manuallyassign the proper protile if you scanner software behaves this way. See Chapter 12. The Adobe
Common Color Architecture, for more information on assigning profiles.

Chapter 11: Color Managementin the Operating Syrtmr

339

East Is East, West Is West

Tlie old saw has it that, "East is east, west is west, and never thetwain shall
meet." It's generally a good idea to enforce a similar separation between
OS-level and application-level color management as much as is feasible,
<andto ensure that when they do meet, they do so gracefully rather than
colliding. It's not only possible, but distressingly easyto have both the application and the printerdrivermanagingcolor behind each other's backs,
so you need to keep track of who is doing what to whom. When you work
with color-managed applications, the easiest course is almost always to
turn off color management in the various device drivers and simply let
the applications handle the color management.

This page lefi blank. ..

A--a
The Adobe
'ommon Color
Architecture
I

Color Management in
Adobe Photoshop,
InDesign, and Illustrator
On our bleaker days, we wonder why all vendors that implement color
management feel they have to do it their way, with their own unique
user interface and their own unique terminology. So we derive a certain
amount of comfort from the fact that Adobe Systems has at least attempted to standardize much of the user interface andmost of the terminology
between its three main color-managed applications. Adobe Photoshop,
Adobe Illustrator, and Adobe InDesign.
We think that Adobe has done more than any other vendor in finding
rational ways to present color-management options to the user, and the
integration between the three applications generally makes life easier.
However, since the three applications do rather different things, some
of the apparent similarities are misleading-we'll point these out as they
arise. And, sometimes, the more obscure consequences of some of the
settings may not be particularly intuitive, so we'll likewise flag these when
we get to them.
We'll start by looking at the settings that really do work identically in
all three applications.

342

Real World Color Management, Second Edition

Color Settings-Command Central for Color

-

Adobe Photoshop, Illustrator, andInDesign all share averysimilarand somewhat formidable-looking dialog box called Color Settings (see Figure 12-1).
Figure 19-1

Color Settings

Photoshopk
Color Settings
L
-

- L -~ . ? ~ ~ ~ * * ~

15- ,.,..".Mrn.-L.
1

6"

5 r*-raa.u*r-.
-

-"
:rm-7
~-

p..<---.~

a m "

m
-w.r**d"t".p..,

",--

%.-d.d---

Color snft

--

9 m a b k color ~

Sewings

'

I.11 W O M Cobr

@ ~ d v a m c dMode
Working sm<n
-

InDesignk Color Settings

Illusnatork Color Sem'ngs

Chaptar If The Adobe Common Color Architecture

343

For now, let'signorethe differences,and concentrate on thesimilarities

All three applications' Color Settings have areas labeled Working Spaces
and Color Management Policies, though the options are slightly different
in each one. Let's first look at what they have in common.

The Senings menusimply containssavedpresets for theColorSettings dialog box. Settings files created in one of the threeapplications can he used
by the other two, which makes synchronizing the behavior of the three
fairly easy. However. Photoshop has more options than InDesign, which
in turn has more options than nlustrator-so if you want to synchronize
Photoshop's setting witheither or both of the other applications, save the
settings from Photoshop, and if you want to synchronize InDesign and
Illustrator, save the settings from InDesign.
Saved settings shordd appear automatically in the Settings menu, and
in the majority of cases, they do. If you find they don't, check the location
of the settings files-the file extension is .csf. On Mac OS 9. they're stored
in the Hard Drive:System FolderApplication Support:Adobe:Color: Settings folder: on Windows they're stored in the Program Filev\Common
Files\ Adobe\Color\Settingsfolder.
The somewhat anarchic situation that applied to previous versions of
the applications under Mac OSX has been addressed in the CS release:
Photoshop CS, InDesign CS, and Illustrator CS all default to saving them
inthe lUserslcusemame>ltibrarylApplicationSupportlAdobelColorlSettings folder, though you can manually load .csf files saved in the main I LibrarylApplication SupportlAdobelColorlSettingsfolder,too. If you want to
make the color settings available to other users, the simplest solution is
create aliases of your settings in the main Library folder.
Macintosh users get one extra preset. labeled ColorSync Workflow.
When you choose this setting, the application uses the RGB, CMYK,
and in the case of Photoshop, Grayscale profiles specified in the Default
Profiles for Documents panel of the ColorSync control panel (Mac OS 91,
or under the Default Profiles tab of the Preferences panel in ColorSync
Utility (Mac OS X). Also, on both operating systems the application uses
the CMM chosen in ColorSync-see "Engine," later in this chapter, for
further ramifications. Unless you really know what both you and the OS
are doing, we recommend that you avoid this option. which may well
disappear in the future.

344

R u l World Color Management, Second Edition

Default Profiles-WorkCng

Spaces

The WorkingSpaces section lets you specifi.default profiles (called"Working Spaces" in Adobe's terminology) for RGB and CMYK. The exact role
that these profiles play depends on the settings you enter elsewhere in
the dialog box, but they always represent the default profile--either assumed or assigned-for untagged RGB and untagged CMYK, and for new
documents.
When the'ildvanced" checkboxis unchecked, your choices are limited
to a recommended set of profiles for RGB and CMYK. But you can control which profiles appear in the lists when'ildvanced" is unchecked. The
lists simply show the profiles in the Recommended folder, so any profile
you install there will show up in the list. This mechanism provides an
administrator aconvenient way to limit the choice of profiles in the three
applications to "safe" ones.
On Mac OS 9, they're stored in the System Fo1der:Applicatinn Support:
Adobe:Color:Profiles:Recommendedfolder, on Windows they're stored in
the Program Files\Common Files\Adobe\Color\Profiles\Recommended
folder, and on Mac OS X they're stored in the ILibrarylApplication SupportlAdobelColorlPmfileslRecommendedfolder. (In all cases, the folder
is aliased to the main Profiles folder, so any profiles you install in the
Recommended folder are available to those other applications that search
more than one level deep for profiles in both the main Library and the
user Library.)
If you turn on the"Advanced"checkhox, the RGBWnrkingSpace menu
lists all the installed bidirectional RGB profiles, and the CMYK Working
Space menu lists all the installed bidirectional CMYK profiles-you can't
use unidirectional input profiles that only convert from device values to
PCS as working spaces. (In the case of RGB, you probably don't want to
use adevice space at dl-see "Intermediate Spaces." in Chapter 10. ColorManagement Workfloru.)

Color Management Policies

The most critical choices you make in the Color Settings dialog boxare the
Color Management Policies,which control the applications' behavior when
they open tagged and untagged documents, when you create new documents. and when you move selected elements between documents. They
don't affect the handling of imported objects in InDesign documents, just

Chapbr 1% The Adobe Common Color Archlteclan

nativeelements.The policies control the applicalions' defaull behavior for

assuming or assigning profiles to untagged documents and for handling
tagged documents by either honoring the embedded profile, converting
from that profile to another, or ignoring the embedded profile and assuming a different one. All three applications let you set separate policies for
RGB and for CMYK (Photoshop lets you set a third policy for Grayscale).
A complicating factor is that InDesign documents have two default
profiles, one for RGB and one for CMYK elements, which can muddy the
distinction between tagged and untagged documents. It's possible to have
an InDesign document whose native RGB elements are tagged and whose
nativeCMYKelementsareuntagged, orviceversa.\Ve71 leave you to make
the call as to whethersuch a document is half-tagged or half-untagged!
The three Warning options provide manual overrides to the policies by
displaying dialog boxes that let you take an action different from the one
dictated by the current policy. Before we look at the warnings, here's how
the policies behave when the warnings are turned off.
O((.This is the probably the most misleadingly labeled option of the three.
You can't turn color management off in these applications-thev always
convert ram a source profile (either embeddedlassigned or assumed) to
themonitor profile fordisplay, and they always use profiles to convert from
RGB to CMYK.
For new documents. the Off policy makes the applications assume the
working space profiles for all native RGB and C M elements, and treat
the documents as untagged-if you change the working spaces, the documents take on the new profiles, and the appearance changes.
For documents that are tagged with a profile orhw than the working
spnceprofile, the applications discard the embedded profile, assume the
working space profile, and treat the documents as untagged.
However, xvhen the application opens a document that's tagged with
the current working space profile, all sense goes out the window. It treats
the document as tagged with the embedded profile. so any subsequent
changes to the working space have no effect on that document. Effectively, adocument with an embedded profile that matches the working space
profile is treated with a "Preserve Embedded Profiles" policy (discussed
next). We think this is counterintuitive, to say the least.

346

Real World Color Management, S m n d Edltlon

When youmovenativeobjects from adocumenl in onespaceto another
in a different space by copy and paste or drag and drop, the application
simply moves the numericalvalues in the object.

P m m Embedded Profiles. Unlike the previous option, this one does

what it says. Tagged documents get opened in the embedded profile's
space (in the case of InDesign documents, which can have both RGB and
CMYK elements, each keeps its definitions). New documents use the nrrrent working space(s1and are treated as tagged. Untagged documents stay
untamed-vou can thinkof this as "preserving" their untagged statusand
use the working space profiles as assumed profiles.
When you move native RGB objects from a document in one space to
another ina different space by copy and paste or drag and dmp, the application performs a conversion from the source to the destination and moves
the color appearance. When you move native CMYK objects from a document in onespace to another in adifferentspace by copyand paste or drag
and drop, the application transfers the numerical values in the object.

Convert to Working Space. This policy is best thought of as an automation feature.W%en the application opens a tagged document, it performs
a conversion from ihe embedded profile's space to the current working
space. When it opens an untagged document, it uses the working space
as an assumed profile and keeps the document untagged.
When you move native RGB or CMYK objects lrom a document in one
space to another in a different space, the application always performs a
conversion from source to destination, preserving color appearance and
changing the numbers.
As a general rule. we set all ourpolicies to Preserve Embedded P r o f i l e s
that way, we at least get to evaluate the image in its profiled space before
deciding what to do next.

Warnings-Manual Ovemdes
'The Missing Profile warnings in Photoshop and InDesign and the Profile
Mismatch warningsin all three applications let you manually override the
default behavior dictated by the color-management policy currently in
effect. The policies determine which radio button is checked by default

Chapter 14: The Adobe Common Color Architeelure

347

in the warning dialogs, but the same options are available n o matter
which policy is in effect. However, since each of the three applications is
designed to do different things, the warnings are presented slightly differently in each one.
The Missing Profile warnings appearwhen you open anuntagged document, and offer the options listed below.

Missing Profile warning in Photahop. The Missing Profile warning,

when enabled, appears whenever you open an untagged document (see
Figure 12-21,

Photoshop Missing
Profile warning

.
The warning offers four options:
b

Leave as is [don't color manage) tells the application to assume the

current working space profile and treat the document as untagged.

Assign working RGBlCMYKIGrayscale tells the application to assign

the working space profile and treat the document as tagged.

Assign profile lets you tell the application to assign any profile that's
applicable to the document's color mode (you can only assign RGB
profiles to RGB documents and CMMC profiles to CMYK documents)
and treat the document as tagged.

Assign profile, then convert to "working RGBICMYK" lets you tell the
application to assign any profile that's applicable to the document's
color mode, then convert from that protile to the Working Space for
that color mode, and treat the document as tagged.

348

Real World Color Management, Setond Edition

Miring Profllt warning in InDesign. While Photoshop documents only

exist in a single color space, InDesign documents can contain RGB and
CMYK elements governed by diierent profiles, called Document profiles-so
lnDesign has not one, but two Missing Profile warnings, one for RGB,one
for CMYK [see Figure 12-31. Note that this dialog only pertains to native
InDesign content, not to placed objects. If the document has profiles embedded, but placed images do not, you won't see this dialog. However, it's
also important to realize thatuntagged placedimages will usethe Document
profiles (for RGB and C W as assumed source profiles.

InDesign's Missing Profile warnings provide the same three options

[with slightly different labels), first for RGB,then for CMYK:
b

Leave as is (usecurrent working space) tells the application to assume

the current working space profile and treat the RGB or CMYK elements
in the document as untagged.

Assign current workingspace tells the application to assign the working

space profile (which becomes the RGB or CMYK Document profile for
this document) and treat the RGB or CMYK elements in the document
as tagged. Untagged placed objects assume these profiles as source and
remain untagged.

Chapter 19: The Adobe Common Color Arehltcclure

b

349

Assign pmfile lets you tell the application to assign any RGB profile in
the firsr warning, and any CMYK profile in the second one. Untagged
placed objects assume these proNes as source and remain untagged.

Missing Profile warnlng in Illustrator.Unlike the other two applications.

Illustrator doesn't let you turn offthe Missing Profile warning-it always
displays the warning when you open a document without an embedded
profile (see Figure 12-4).(This can be annoying, since Illustrator always
looks for profiles in EPS documents, but can't embed profiles in EPSs on
save.)
~ ~ I I W194

~ b l i y ~ k

nlustrator's Missing
Profile warning

Illustrator's Missing Profile warning provides the same options as

InDesign's-the only difference is that Illustrator documents are always
defined as either RGB documents or CMYK documents, so you only get
one warning:
b

Leave as is (don't color manage) tells the application to assume the

current Working Space profile and treat the document as untagged.

Assign current working space tells the application to assign the working space profile and treat the document as tagged.

Assign profile lets you tell the application to assign any RGB profile to
an RGB document, or any CMYK profile to a CMYK document.

The Profile Mismatck Ask When Opening warning appears when you
open a document with an embedded profile that's different from the current working space.

350

Real World Color Management, Second Edition

Photoshop's Embedded Profile Mismatch warning. When enabled, this

warning displays the dialog box shown in Figure 12-5.

C19-5
as

Photoshop's Embedded
ProfileMismatch warning

.B

."

emkdbcd cola
,.a<.

.ron1.

*a,*#".

The warning offers the followingthree options:

b

Use the embedded profile (instead of the working space) preserves the
embedded profile and treats the document as tagged. The embedded
profile is used as the source for all conversions.

Convert document's colors to the working space tells the application

to perform a conversion from the embedded profile to the working
space profile, and treats the document as tagged with the working
space profile.
Discard the embedded profile (don't color manage) tells the application to discard the embedded profile, assume the working space
profile,and treat the document as untagged.

InDesigth Embedded Profile Mismatch warnlng.Again, since InDesign

documents can contain both RGB and CMYK elements, it's possible to
have a profile mismatch in either or both. If the Color Management Policies are set to anything other than Off, InDesign will embed the Document
profile for that mode. If neither mode is set to Off,InDesign will embed
both Document profiles (RGB and CMYK), regardless of the contents of
the document. When enabled, this warning displays the dialog boxes
shown in Figure 12-6.

Chapter 119:The Adobe Common Color Architecture

3fi

ngure 19-6

InDesign's Embedded
profile Mismatch
warnings

'
-me
Embdded

Imn*

Ismatch

doolmmvr embddrd r o t ~ f~ i l de m nm match m~ cunmt

CUYK uorkinq lm<e

Embedded R*IP17MaxK-R-S-ZllO

Both warnings offer the following three options:

b

Use the embedded profiles (instead of the working space) preserves

the embedded profiles and treats the RGB or CbM( elements in the
document as tagged. The embedded profiles become the Document
profiles for this document, which means they are used as the source
profiles for all conversions of native objects, and as the assumed
source profiles for untagged placed objects.

Convert the document's colors to the current working space tells the
application to convert the color from the embedded profiles to the
working space profiles (which become the Document profiles for this
document), and treats the document as tagged. It doesn't convert
placed objects, but untagged objects use the new Document profiles
as assumed source profiles.

Discard the embedded profile (and use current working space) tells
the application to discard the embedded profile, assume the working
space profile, and treat the RGB or CMYK elements in the document as
untagged. Untagged objects use the workhg space profiles as assumed
source profiles.

352

Real World Color Managrmml, Second Edition

Illustrator's Embedded Profile Mlsmatch warning. Illustrator'swarning

is functionally identical and cosmetically very similar to Photoshop's (see
Figure 12-71,
Figure 19-7

Illustrator's Embedded
Profile Mismatch warning

It offersthe following three options:

Use the embedded profile (instead oftheworkingspace) preserves the
embedded profile and treats the document as tagged. The embedded
profile is used as the source for all conversions.
b

Convert document's colors to the current working space tells the application to perform a conversion from the embedded profile to the
working space profile, and treats the document as tagged with the
working space profile.
Discard the embedded profile [don't color manage) tells the application
to discard the embedded profile, assume the working space profile,
and treat the document as untagged.

Annoyingly, both InDesign and Illustrator take the warning a step further by displaying an AssignedProfile Mismatchwarningshould you have
the temerity to assign a profile other than the working space in the Missing Profile dialog box. This is the kind of behavior that gives color management a badname. Photoshop sensibly refrains fron doing so, and we hope
the other applications will follow suit in a future release!

Profile Mhmatch: Ark When Pasting warning. The last set of warnings
lets you override the default behavior when you move elements (or in
the case of Photoshop, selected pixels) from a document in one space to
another in a different space, by copy and paste or drag and drop. It only

Chapter 1% The Adobe Common Color Archiledure

353

applies when both source and destination document arein the same color
mode--RGB to RGB or CMYK to CMYK. If you move elements from one
color mode to another, a conversion always occurs-you can't paste RGB
values into a CMYK document or vice versa. The warning offers the same
pair of options in all three applications (see Figure 12-8).

Paste Pmfile Mismatch

warnings

The IOUrce data's cob, P.ORIC dm5 not matrh the deW"atlon
documenl'l d o 1 vroRIe.

sourre:

I ~ O -

RGS 119981

bflnflon: Wofklng RCB PmAlMD Ua

ZnDesignk Paste Profile Mismatcl~a~arnir~g

Illustratork Paste Profile Misnzntch umrning

In this situation, the only choices are to move the RGB or CMYKvalues,
in which case the appearance will very likely change, or move the color
appearance, in which case the RGB or CMYK values will change.

354

Real World Color Management, Second Edltlon

Convert (preserve color appearance) makes the application convert

the object from the source document's assigned or assumed profile
to the destination object's assigned or assumed profile, preserving the
color appearance.
b

Don't convert (preserve colornumbers) moves the numerical values in

the object to the destination document, where theywilllikely generate
a different color appearance.

We usually leave all the warnings turned on-they provide a useful

reality check. Ifwe know that we'll be dealing with alarge number of documents in a space other than our current working space, we'll change the
working space rather than turn off the warning.

Advanced Option+Default

Conversion Settings

When you enable the 'Rdvanced" checkbox, new controls appear that let
you control how the applications perform conversions, including not only
the ones discussed above, but all other conversions that don't present a
user interface with a choice of conversion options. This includes choosing a different mode from Photoshop's Image>Modemenu, Illustrator's
File>Document Color Mode and Filter>Colors>Convert to RGBICMYK,
or InDesign's Color palette pnp-out menu (see Figure 12-91.

PflotoshopS Conoerriori Opriorrs

InDesigrlk Conr'er-sioilOptions

Chapter 19: The Adobe Common Color Archllectun

355

mine. This option allows you to choose a specific CMM from the ICCcompliant CMMs installed on your computer.We're big fans of the Adobe
[ACE) e n g i n e i t seems to be remarkably bug-free and accurate. Its only
major downside is that it's only available inside Adobe applications, so if
you need to make sure that you're getting exncriy the same conversions
inside and outside Adobe applications, you'll need to settle on a CMM
that's installable as a standalone.
Most ofthe items on the menu are self-explanatory-it's simply a list
of different CMMs-but Mac users get one extra item, Apple ColorSyc,
that's a bit more mysterious. When you choose Apple ColorSync, the application uses the CMM specified in the ColorSync control panel (Mac OS
9) or the ColorSync panel found in System Preferences (Mac OSX). If you
choose Automatic, each profile looks for its preferred CMM, which then
gets used for the conversion between that profile and the PCS. The positive aspect of this is that you ensure that any profiles containing "secret
sauce" that's reliant on a particular CMM will get used to their fullest potential. The negative is that it's veryunlikely that you'll know at any given
moment which CMM is doing what to whom!
Intent. This option lets you choose the default rendering intenl. which
is used forall conversions that don't let you specify a rendering intent at
conversion time. For those into trivia. it also affects the LAB values in the
Info palette. so ifyou arewondering why the LAB values in the Info palette
don't match up to actual measurements, it's because this setting isn't set
to absolute colorimettic.
Note that the default rendering intent is relative colorimetric luith
black point cornpermtion, which is not at all the same thing as relative
colorimetric without black point compensation--see "Use Black Point
Compensation." below.Weleave this setting at the default. but ifyou find
that youwant tousesome other rendering intent more than half the time,
you may want to change it to the one you use most often.

Use Black Point compensation. This proprietary Adobe feature plugs a

hole in the ICC prof le spec. It ensures that black in the source is always
mapped to black in the destination, so that the entire dynamic range of
the input is mapped to the entire dynamic range of the output (seeFigure
12-101.

356

R u l World Wor Manawnnnt, ktond Edltion

Black point
compensation

black point compensation off

Black point compensation ensures that black in the source is always
translated to black in the destination, prewntingthe loss of shadow
detail that's evident in the above right image.

When "Use Black Point Compensation" is turned off, it's possible to get
either of two undesirable outcomes:
b

If the source has a lower black point than the destination, all values in
the source that are darker than destination black get clipped to black,
destroying shadow detail.

If the source has a higher black point than the destination, the converted color contains no true blacks, so the result appears washed out.

"Use Black Point Compensation"avoids both these problems.The only

reason we can see for turning it off, other than to see what it does, is if
your workflow depends on having exactly the same conversions available inside and outside the Adobe applications-and quite honestly, we
find it so valuable that if you do have such a workflow, you may want to
consider changing it.

Manual Assignments and Conversions

The Color Settings dialog box is mostly concerned with the behavior of
newly opened or newly created documents, but the three applications
also let you perform profile assignments and conversions while you're
working on a document.

Chapter 19: The Adobe Common Color Arthltecture

357

Like the Color Settings, the controls described in this section operate
at the document level. They don't affect the behavior of linked graphics
with embedded profiles in InDesign, although they clo affect the assumed
profile for linked untagged graphics.

Assign Profile
All three applications offer anAssign Profile command. It letsyou do three
things:
b

Untag any document (don't color manage). Working spaces are assumed as source.

Assign theworkingspace profile to a document. If the document was

untagged, assigning the profile turns it into a tagged one.
Assign any profile to a tagged document, overriding the one that was
previously in force, or to an untagged document, thereby turning it
into a tagged one.

In Photoshop, Assign Profile is found on the Mode submenu under the

Image menu. In InDesign and Illustrator, it's on the Ed~tmenu.The Assign
Profile dialog boxis virtually identical in Photoshop and Illustrator. Since
InDesign documents can contain both an RGR and a CMYK document
profile, the InDesign Assign Profiles dialog box lets you make separate
assignments for RGB and CMYK (see Figure 12-1 1).

Assign Pmfile(s)
dialog boxes

Clockwisefrom top:Photoshop'sAssign Profile,

IllustratoriAssign Profile, and inDesign5
Assign Profiles dinlog boxes.

358

Real World Color Management, kcond Edition

-

--

Tip: Don'l Trust Illustrator'sDocument Info Palette. One Illustrator 10

bugthat hasmade its way into Illustrator CS is that the profile information
in the Document Info palette usually doesn't update until you close and
reopen the palette. At the time of writing, the current version of Illustrator
is 11.0.0, and the bug has persisted in all versions of Illustrator 10, so we
can only conclude that fudngit isn't a high priority. The Document Profile
display from the pop-up menu at the lower left of the document window,
however, operates as expected.
InDesign's Assign Profiles command is different in other ways, too.The
profile assignments apply not only to native objects, but also to linked,
imported objects that are untagged. To make matters a little more confusing, the rendering intent choices apply to both native and imported
objects, whether tagged or not, by default. You need to use the Image
Color Settings option, discussed later in this chapter, to specify a different
rendering intent.

Convert to Profile
Photoshop and InDesign both offer a command labeled Convert to Profile,
but due to the different nature of the two applications, they dovery different things.

Photoshop's Convert to Profile. Photoshop documents are always governed by a single profile, so assignments and conversions are both relatively straightfomd. Photoshop's Convert to Profile command is found
on the Mode submenu of thelmage menu.The dialog boxlets you choose
a destination profile, aCMM (or Engine, in Adobe terminology),a rendering intent, and a "Use Black Point Compensation" checkbox (see Figure
12-12).

Chapter 1P:The Adobe Common Color Archil&ure

359

The conversion applies to the entire document. In the case of layered

files, the dialog also offers the option to flatten the document-it's usually a good idea to do so, particularly when the conversion also includes
a mode change, because layers may blend differently in the destination
space, changing the document's appearance. The Preview checkbox lets
you preview the result of the conversion.

InDesign's Convert to Profile. InDesign's Convert to Profile dialog box

looks fairly different from Photoshop's, since InDesign documents have
two profiles--one for RGB, one for CMYK (see Figure 12-13).

InDesign's Convert to
Profile dialog box

InDesign's Convert to Profile converts all native InDesign elements

from the Document profiles (the current spaces) to the selected destination spaces. The destination spaces then become the Document profiles
for that document. As a result, they also become the new assumed profiles
for untagged placed objects. The operation has no effect on tagged placed
objects.

Illustrator convefslonr. Illustrator only allows you to perform one type of

conversion--a mode change between RGB and CMYK. The source profile
is always the working space for the document's color mode in the case of
untagged documents, or the Document profile (which may be the working
space, or may be a different assigned profile)in the case of tagged ones. The
destination is always the working space for the destination color mode.
The Document Color Mode command (on theFile menu) lets you convert an entire document from RGB to CMYK or vice versa.

360

Real World Color Management, Second Wltlon

Color Managing Imported Graphics

All the controls we've discussed so far operate primarily on native objects

in adocument, not on imported graphics such as placed images. But we have

to leave in the weasel word, "primarily,"because thecolor-management policies dohave animpact on how imported graphics are handled, and these
interactions can be subtle and quite often counterintuitive. Moreover, all
three applications have quite different controls.

Placed Graphics in Photoshop

Photoshop is pretty straightforward. It has a Place command, but any
placed elements end up as part of the document, and are governed by
the document's assigned or assumed profile. Photoshop always places
the numerical values contained in the placed document, so its embedded profile is ignored. Basically, once an object is placed in Photoshop, it
ceases to be a placed graphic.

Placed Graphics in InDesign

When vou place a graphic in InDesign, the numerical values in the file
are always presen'ed-InDesign never changes these numbers except
as part of the print stream. It lacks the ability to go out and change the
source file. All it can do is take the numbers in placed graphics and put
different interpretations on them by assigning or assuming profiles. The
profiles are used for display and for conversion to output space at print
time (and for exporting to PDF, discussed in Chapter 16). InDesign also
dowsyou to specify rendering intents forplaced objects. both as defaults
and on an object-bv-object b a s i s t h e rendering intents areused only for
the conversion to output. Conversion to the display rendering intent is
controlled by Color Settings.

Placing tagged graphfa. InDesign always tracks embedded profiles in

placed graphics, even if you choose the Color Management Off preset
in Color Settings, but it doesn't necessarily use them. To make lnDesign
ignore an embedded profile when you place a graphic, you can do any of
the following:
r Set the Color Management Policy in Color Settings for the graphic's
color mode to Oft

Chapter 11P:The Adobe Common Color A~h(tectun

361

t Uncheck"Enab1eColor Management" in the Color Settingsdialog box;

or
t Check the "Show Import Options" checkbox in the Place dialog box,

then turn off the "Enable Color Management" checkhox in the Image
Impon Color Options (see Figure 12-14).

Image Import Options

nwn12-14

Since this is a book about color management, we assume that you're

unlikely to disable color management by any of the aforementioned
methods, but if you do, note that placed objects aren't color managed
at all, even to the monitor. There's no display compensation, and CMYK
imagestypically look hideous. If you subsequently enable color management for that object, InDesign will revert back to the embedded profile if
present; otherwise it will assume the Document profiles, if present, and
the working spaces if not. We take theview that placing tagged files in InDesign, then tryingto make it do something other than honor the embedded profile, is at best a recipe for confusion, and more likely a fast ticket
to the funny farm (assuming you're not already there).
The key thing to realize here is that the only way to make InDesign
ignore an embedded profile in an image and assume some other profile
in order to get on-screen previews is to use the Off policy. Even then, the
only profile that will be assumed is the Document profile for the image's
color mode, or, if the document is untagged, the workingspace profile for
the image's color mode.

Placing untagged graphla. The key to understanding the behavior of

placed untagged graphics in InDesign is knowing whether the host InDesign document is itself tagged or untagged. Placed untagged graphics
always use the Document profile for their color mode as the assumed
source profile.

369

Real World Color Management, Second Edition

F

Tagged host documents have two Document profiles assigned, one for
RGB, one for CMMC Changes tothe working spacesdon't S e c t thedocument or the Document profiles. and therefore don't affect the placed
graphic. If you change the Document profiles [byusing Assign Profiles
or Convert to Profilel, then the placed untagged graphic is affected.

Untagged host documents have Document profiles that aren't assigned

to them, but are assumed from theworkingspace profiles. Ifyou change
the working spaces, you also change the Document profiles, and therefore the behavior of untagged graphics.

Thesourceprofilesfor untagged placed graphicsare the Document profiles:

AssignProfileand Convert tol'mflechangethe Documentprofiles,and hence
the source profiles for untagged placed graphics. InDesign documents can
also be half-tagged (or half-untagged) ifa profile is designated for one color
mode but not the othecThe above rules apply to that situation too.

Image Color Settings. The Image Color Settings command lets you assign a different profile to placed graphics on an object-by-object basis
[see Figure 12-15).

lmage Color Settings lets you assign a specific profile and rendering
intent to a placed object.If you assign a specificprofile, ithecomes theassigned profile for that instance of the placed object, and remains in effect
until you change it by assigning a different profile through Image Color
Settings. If you choose Use Document Default in Image Color Settings,
you're in effect asking InDesign to treat the image as untagged, which
means the Document profiles will apply.

Placed Graphics in Illustrator

We often finding ourselves wishing that Illustrator would decide what it
wants to be when it grows up. It's true that it's an amazingly flexible tool
that$ used for everything from page layout to Web page design, but all too

Chaplo IP: The Adobe Common Cdor Architecture

363

often the processwindsup beinga bit like doing brain surgerywith a Swiss
Army knife, and we've seen Illustrator's approach to color management
reduce grown men to tears. This is particularly true in the case of images
placed in Illustrator.
The first key point to understand is that Illustrator always includes a
copy of the placed image in EPS files. Illustrator itself doesn't seem to be
aware of this--if you don't specifically include the linked files when you
save an EPS from Illustrator, and reopen the EPS in Illustrator, it looks for
the linked He, and if it can't find it, it asks you to replace it. But if you place
the EPS in a page-layout application, it will display and print correctly.
But this doesn't happen with native Illustrator filesif you want these to
be self-contained,you need to save using the "IncludeLinked Files" option,
or uncheck the " L i e checkbox in the Place dialog when you place images.
The second key point lo understand is that Illustrator color-manages
eitherthe copy that's always embedded inEPSs,or the copy that's embedded in native Illustratorfiles, either by explicitlyembeddingorbyrefreshing the link.Illustrator never converts the linked image.
The third key point to understand is that Illustrator always converts
placed images to the document color mode. If you place an RGB image in
a CMYK Illustratordocument, the copy that goes in the Illustrator file will
be converted to CMYK, and ifyou place a CMYKimagein an RGB Illustrator document, the copy that goes in the Illustrator file will be converted
to RGB. You may get error or warning messages that would lead you to
believe otherwise, but we've done enough testing to be pretty confident
that they're bogus.
AU conversions of placed graphics use the CMM and rendering intent
specified in Illustrator's color settings, and the destination profile is always the document profile. So the only real questions are
t Does a conversion occur?

r If so, what is the source profile?

Pladng antagged flle~in Illustrator. When you place untagged files in

an Illustrator document, one of two things happens:
t lfthe placedfile is in thesame color mode as the Illustrator document,
the numerical values in the placed file are transferred to the lllusttator
document with no conversion.

365

Real World Color Management, Second Witlon

b

If the placedfile is in a different color mode from the illustrator document, the copy in the Illustrator file is converted using the Working
Space profile for that color mode as the assumed source profile.

For example, if you place anuntagged RGB image in a CMYK Illustrator file, illustrator will use the RGBWorking Space profile as the assumed
profile forthe conversion to document CMYK.

Pbdng tagged filein Illustrator. Ifyou considerplacingimagesin Nustrator documents as a pasting activity. things make more sense, including
the occasional Paste Profile Mismatch dialog box you get when placing
images. If the placed file is in a different color mode from the Illustrator
document. Illustratoralways performs a conversion using the placed file's
embedded profile as the source. If the placed fileis inthesame color mode
as the Illustrator file, Illustratorworks the same way as Photoshop's pasting behavior, which depends on the Color Management Policy selected.
b

OfE The numerical values in the file get placed (pasted).

Preserve Embedded Profiles: Placing (pasting) RGB in an RGB document preserves the color appearance (Illustrator converts using the
embedded profile as the source and the Document profile as the destination). Placing CMYK in a CMYK document preserves the color
numbers and doesn't do a conversion.

Convert to Working RGBICMM(: Illustrator always preserves (pastes1

the color appearance-it converts from the embedded profile to the
Document profile.

Ofcourse, ifthe source and destination profiles are the same, no conversion occurs. Ifyou've checked %k When Pasting" fm Profile Mismatches
in Color Settings, you'U only get a Paste Mismatch dialog to override the
above default behaviors if the Link checkbox is unchecked in the Place
dialog. If you choose to Link the file to be placed, you won't get a Paste
Mismatch warning.
If you find this collection of paste behaviors counterintuitive, you're
not alone!The onlysilver lining is that they're consistent in all three Adobe
applications.

Chapln 1% The Adobe Common Cdor Archltdurc

365

--

Simulations and SoftmProofingOne of thenlost valuahle capabilities ot the BigThree Adobe applications
is their ability to simulate, on the monitor, the result of conversions to
other profile spaces, or, in the case of Illustrator and Photoshop, theresult
of sending the unconverted fileto different outputs.
All three applications let you invoke a soft proof by choosing Proof
Colors from theview menu. Soft proofs are window-specific,so you can
view the same document with different simulations by opening multiple
windows and assigning different soft proofs to each one.
The default simulation is for the CMYK working space, but you can
change and control the simulation by choosing Proof Setup from theview
menu. Atthis point, however, the applications diverge in the controls that
they offer.We'll start with illustrator, since it's the simplest.

Simulations in Illustrator
Illustrator's Proof Setup dialog boxis relatively simple (see Figure 12-16).

Ill~~wrator's
Pmof Setup

I
Illustrator's Proof Setup offers only three controls:
The Profile menu lets you choose a destination protile for the simulation.
b

The "Preserve Color Numbers" checkbox tells the application to simulate what would happen if you printed the numerical values in the file,
so it's only available when the profile selected for the simulation is in the
same color mode as the document-you can only send RGB numbers
to RGB devicesand CMM( numbers to CMYKdevices.Thevisual result
is the same as assigningthe selected profile, but it's a simulation rather
than a permanent profile assignment. If you enable the checkbox, the
third item, Intent, becomes dimmed and unavailable-rendering intents aren't applicable since you aren't requesting a conversion.

366

Real World Color Managcmcnt, Second Edition

The Intent menu lets you choose a rendering intent for the conversion
from document space to simulation space, allowing you to preview
the effects of different renderings.
Illustrator doesn't let you control the rendering from the simulation
space to the monitor space--it's always relative colorimetric, with black
point compensation if it's checked in Color Settings,and without it if not.
Note also that the Color Settings "Use Black Point Compensation"setting
also affects the conversion from document to simulation space.

Simulations in InDesign
InDesign's Proof Setup dialog box is rather different from Illustrator's. In
InDesign, rendering intents are applied to individual document elements
rather than to the document as a whole,so the ProofSetup dialog box has no
rendering intent control for the conversion from the varioussource spaces
used bythe document elements to the simulation space. Instead, each element is rendered according to its specified intent. As with Illustrator, the
Color Settings "Use Black Point Compensation" setting applies globally to
the conversions from all the source spaces to the simulation space.
Unlike Illustrator, InDesign offers control over the rendering from the
simulation space to the display (see Figure 12-17).

u
InDesign's b o f Setup

- 1

dialog box

InDesignrsProof Setup dialog box contains three controls, the Profile

pop-up menu, and two Simulate checkboxes:
c The Profile menu lets you specify the destination profile forrhe simulation.
b

The Simulate checkboxes let youcontrol the renderingfrom thesimulation space to the monitor:

"PaperWhite,"when checked, produces an absolute colorimetricrendering from the simulation to the monitor, showing the color of the

Chapter 1% The Adobe Common Color Architecture

367

paper, and its influence on the rest of the color. When it's checked, the
"lnk Black" checkbox is automatically turned on and dimmed.
t "InkBlack,"when checked, turns offblack point compensation for the

rendering From simulation to the monitor, so if the simulation space

black is lighter than the monitor black, "Ink Black" will show you the
washed-out blacks you'd get on output-it's most noticeable when
you're simulatinglow dynamic range processes likenewsprint.
When both "lnkB1ack" and "Paper White" are unchecked, which they
are by default, the rendering from simulation to display is relative
colorimetric with black point compensation, which means that the
simulation white is displayed as monitor white, and the simulation
black is displayed as monitor black.
For a more in-depth discussion on the pros and cons of the different
renderings to the monitor, see "Soft-Proofing Practices and Pitfalls," later
in this chapter.

Simulations in Photoshop
Of the three applications. Photoshop offers the most complete set of soft-

proofing controls. Photoshop's Proof Setup is shown in Figure 12-18.

Photoshop's Proof Setup

dialog box

Photoshop offers the unique ability to name and save different proof
setups for fast recall. The Setup menu lets you recall proof setups that
you've saved in the special Proofing folder. (On Mac OS 9. this is the Systern FolderApplication SuppoltAdohe:Color:Proofingfolder. On Mac OS
X it's the ILibrary/Application Support/Adobe/ColorlProofingfolder, and
inwindows, it'sthe Program Files\Common Files\Adobe\ Color\Proofing
folder.) You can save pmof setups anywhere on your hard disk by clicking

368

Real World Color Manasnncnl, Second Edition

Save, and load them by clicking the Load button, but the setups you save
in the Proofing folder appear on the list automatically. (Even bener. they
also appear at the bottom of the Proof Setup submenu, where you can
choose them directly.)
Photoshop is also unique in offering a live preview-the window updates while the dialog box is open-when the "Preview" checkhox is
checked. We find that this makes it much easier to compare the effects of
different rendering intents.
The remaining controls operate like those in the other applications.
The Profile menu allows you to choose the destination profile for the
simulation.
b

The "Preserve Color Numbers" checkboxtells the application to simulate what would happen if you sent the numerical values in the file,
so it's only availablewhen the simulation and document profile are in
the . w e mod+RGB to RGB or CMYK to CMYK.

t The Intent menu allows you to choose a rendering intent for the con-

version from document space to simulation space, lettingyou preview

the effects of different renderings.
TheMUseBlackpoint Compensation"checkboxletsyou choose whether or not to apply black point compensation in the rendering from the
document space to the simulation space.
The Simulate checkboxes cone01 the rendering from the simulation
space to the monitor.
"Paper White," when checked, produces an absolute colorimetric rendering from the simulation to the monitor, showing the color of the
paper and its influence on tbe rest of the color. When it's checked, the
"Ink Black" checkbox is automatically turned on and dimmed.
t "lnkB1ack." whenchecked, turns off black point compensation for the

rendering from simulation to the monitor, so if the simulation space

black is lighter than the monitor black, "Ink Black" will show you the
washed-out blacks on output-it's most noticeable when you're simulating low dynamic range processes like newsprint.

Chapter 1% The Adobe Common Color Arthllcctun

b

369

When both "Ink Black" and "Paper White" areunchecked, which they
are by default, the rendering from the simulation space to the display
is relative colorimetric with black paint compensation, which means
that the simulation white is displayed as monitor white, and the simulation black is displayed as monitor black.

Soft-Proof ng Practices and Pitfalls

Using soh proohng successfully requires a little forethought and a linlc
howledge. First, it's useful to distinguish between "Preserve Color Numbers" simulations, and simulations of actual conversions.We use Preserve
Color Numbers in two different scenarios:
F

With RGB, we use Preserve Color Numbers in conjunction with Web

graphics. ChoosingtheWmdowsRGB and Macintosh RGB settings lets
us see approximatelv how Web graphics will appear an uncalibrated
Mac and Windows systems.

With CMYK, we use Preserve ColorNumbers to see how existing CMYK

files will work on different printing processes. For example, we may
use Preserve Color Numbers to decide whether we can get away with
using the same file for several slightly different printing conditions, or
ifwe need to create a separate file for each one.

Most of the time, though, we use Proof Setup to preview conversions to

final output space. But no matter whether you're previewingstraight output
or conversions, there are a few things you need to know about InDesign's
and Photashop's "Paper White" and "Ink Black" simulation settings.
You might expect the absolute colorimetric rendering ofthe simulation
to the display that you get fromchecking"PaperWhite"to be the most accurate of the three possibilities. The problem, though, is that if you have
any white user-interface elements-lnDesign's Pasteboatd, or any menus
or palettesdisplayed on the screen, your eyeadapts to that white, so the
simulated paper white looks wrong.
In Photoshop, it's relatively- easy to hide everything except the image--we hide all the palettes, and use full-screen mode with the menu
barhidden-and get an honest absolute colorimetric soft proof. In 1nDesign, though, it's just about impossible to do so. Hence in InDesign. we
only use paper whitesimulation when we're dealing with papers that are

370

Real World Color Wanagemml, kcond Edilon

a long way from white or when we're comparing the screen display with
actual hard copy. (See"Viewing Environment" in Chapter 9, Evaluating
nnd Editing Profiles,for a thorough discussion of the pitfalls of screen-toprint comparisons.)
With both Simulate checkboxes unchecked (and always in Illustrator
since it lacks them), the rendering to the screen is relative colorimetric
with black point compensation, whichmeans that paper whiteis mapped
to monitor white and ink black is mapped to monitor black. We find that
this is generally the most useful view for making overall judgments on
tone and color. Its only flaw is that it provides a samewhat optimistic
rendering of low dynamic range processes such as newsprint or inkjet on
uncoated papers.
In those cases, we'll use the ink black simulation to make quick reality
checks on shadow detail. but we still rely primarily on the other two settings to make critical judgments.

Tip: Change the Default Proof Colors from Working CMW. The default setting when you choose Proof Colors is for Working CMYK, using
the rendering intent and black point compensation settings specified in
Color Settings, with [in InDesign and Photoshop) simulate Paper White
and InkBlack turned off. Ifyou want to change thedefault, simply choose
Proof Setup with no documents open. enter the settings you want as the
defaults, and clickOK.The application willuse the new settings whenever
vou choose Proof Colors.
Although Proof Setup is primarily for on-scteen simulations, it also
plays into the applications' printingarchitectures, which are the penultimate stop on our journey through Adobe's color-management features.

Photoshop, Illustrator, and InDesign can all perform a conversion on the

data that gets handed off to the printer driver.
If you use these features-and we encourage you to do so-the one
major pitfall to avoid is also having color management performed by the
printer driver. It's quite easy lo set things up so that the driver does corrections tothe already correcteddata. and the results usually aren't pretty.
See Chapter 11. Color Management in tlw OpermingSystem.

Chmpfu 119:The Adobe Common Color AwhltccWn

37I

The features in these applications' printing controls allow you to do

essentially four things:
Send the numbers in the document directly to the printer driver.
b

Send the numbersin the document, alongwitha profile that describes

their meaning, to the printer driver (with the assumption that the
printer driver will be able to interpret the profile and do something
useful with it).

Convert the numbers in the document to the printer space, and send
the converted numbers to the printer driver.

r Convert the numbers in the document to the space you specified

in Proof Setup, then convert those numbers to the printer space, so

that the resulting print is a hard-copy simulation of the Proof Setup
space--this is sometimes called cross-retldering. (This option is not
available in Illustrator.)
Let's look at how the controls operate to let you accomplish these various ends. The layout of the dialog boxes is different in each of the three
applications, but the controls have the same labels. In InDesign, they're
in the Color Management panel of InDesign's Print dialog box. In Photoshop, they're in the Colot Management section of Photoshop's Print with
Preview dialogbox. And in Illusbator, they're in Illustrator 10 options for
PostScript driversandas an extension to tasterprinter dialogs.The dialog
boxes are shown in Figure 12-19.
Each dialog has a section labeled Source Space and a section labeled
Print Space. Source Space has only two options (except in Illusuatorl:
b

Document uses the Document profile (or profiles, in the case of InDesign) as the source profile for any conversion specified in the Print
Space section.

Proof Setup (Proof in InDesign) makes the application perform a conversion from the document profile(s1 to the profile specified in Proof
Setup. In Photoshop, it uses the rendering intent specified in Proof
Setup. In InDesign, each object in the document has its own rendering
intent. The results of this conversion are passed to the Print Space
section of the dialog.

37P

Real World Color M a n a m , Second Edition

n ~ 1P.n19
Print Color Management
in Photoshop, Illustrator,
and InDesign

Photoshop's Print
with Preuiew
dialog box

The Color Management

panel ofIZlushntor's
Print dialog bar

Chaptn 1P:The Adobe Common Color Architecture

373

Print Color Management

in Photoshop, IUustrator,
and InDesign, continued

7'heColor Management
panel ofInDesigrzkPrint
dialog box

Print Space has only three options:

b

Same as Source passes the numbers that result from the choice made in
Source Space to the printer driver (InDesign doesn't have this option).
Printer Color ManagementlPostscript Color Management passes the
numbers that result from the choice made in Source Space to the printer
driver, along with the profile or profiles that describe the meaning of
these numbers. The assumption is that the printer driver will then use
the profile(s)as source space(s)for a conversion to the print space.

Choosing a specific profile instructs the application to perform a

conversion from whatever numbers result from the choice made in
Source Space to the selected profile, using the rendering intent you've
specified in Print Space.

InDesign has a couple of extra wrinkles. First, if you choose Document

in the Source Space, you don't get to choose a rendering intent in Print
Space-instead, each object in the InDesign document is rendered using
its own specified intent. (If you choose Proof in the Source Space, you do
get to choose a rendering intent, since the conversion is now from one
single space--the Proof space-to another-the Print Space.)

374

Real World Color Management, Second Edition

Second, the choice of profiles in inDesign3sPrint Space is governed by

thechoices youmakein thecolor menu ofthe Output panel ofInDesign's
Print dialog box (see Figure 12-20). If you choose Composite RGB, only
RGB profiles are available, while if youchoose CompositeCMYKorSeparations, only CMYK profiles are available.
While the numberof possibilities that exist is extremelylarge, the printing controls offer easy support for three fairly obvious workflow options:
In an early-binding workflow, where all the color is already in final output space, setting the Source Space to Document and the Print Space
to Same as Source lets you print the document *dues unchanged to
the final output device.
You can also print proofs to some other device by selecting its profile
in Print Space. If you choose absolute colorimetric rendering, the
proofing printer will lay down ink in the paper white areas to match
the paper white of the final output, while relative colorimetric will
scale the final paper white to the paper white of the proofer.
t

In a late-binding workflow, where the document is in a space other

than the fmal output space, choosing the profile for the final output
device in Print Space makes the application convefl the color to the
final output space--you'd use this when you'reactually printing to the
final output device.

t To generateproofsinalate-bindingworkflow, you can set Proof Setup

to the final output space, then select Proof in the Source Space, and
the profile for your proofing device in Print Space. This instructs the
application to first convert the color from the source spaces to the final
output spacespecifiedin Proofsetup, then to convert that color to the
spaceyou'vespecified in Print Space.Theresult is that the printer sirnulates the final output. Again, absolute colorimetric rendering makes
the proofer simulate the final paper white, while relative colorimetric
scales the color to the paper white of the proofer.
One caveat: InDesign only applies absolute colorimetric rendering to
the contents of frames. It leaves the paper white areas paper white,
which defeats the point oftheexcercisesince our eyes promptly adapt
to the blank paper white. An easyworkaround is toplace a blank frame
on the background of the master page.

Chapla 19: The Adobe Common Color Archltetture

375

InDesign's Print Output

options

)10

~,,,.c;
Cnmtm,:eI+.w

unrhsnp",

cornwrite cr*

The choice you make

in the Outpltr pop-rip
menu determines
which profiles become
nr~ailablein the Color
Management panel
shorun in Figure 12-19.

Flip

',#>?

Adobe Camera Raw

The only completely new feature inthe Creative Suite that has colormanagement implicationsis the Camera Raw plug-in (a version was made available
for Photoshop 7 towards the end of that product'slifetime,but Camera Rawis
built into Photoshop CS).h a nutshell, Camera Raw is a universal converter
for raw captures from digital cameras. It also takes a unique approach to
digital camera color management, one that we find extremely powerful and
interesting. But before we discuss it in detail, it's worth spending a little time
on understanding the nature of digital raw captures.

Digital Raw
Raw format is to some extent a misnomer, because it isn't a single file
Format. Each camera vendor has their own proprietary file format, and
in some cases has different file formats for different cameras, but they
all share a common characteristic. Raw files from digital color filter array
cameras are linear-gamma grayscale images that require processing by a
raw converter (such as Adobe Camera Raw) to produce a color image.
A typical Bayer-pattern color filter array (CFA) might look something
like the one shown in Figure 12-21. The array contains twice as many
green-filtered sensors as it does red and blue, because our eyes are most

376

Real World Color Managsmnt, Second Edltlon

sensitive to the wavelengths we interpret as green. Other arrangements

are possible--some cameras use CMY filters instead of RGB, because they
transmit more light, while still others may add a fourth color.
In a BuyerPattern colorfilter
array, each photosensor is
filtered so that it captures only
a single color of light: red, green.
or blue. Twice as many green
filters are used as red or blue
because our eyes are most
sensitive to green light.

ngum 19-91
Bayer pattern
Color Filter Array

But no matter how the color filters are arranged, each element in the
sensor simply records a brightness value in the raw file. To produce color,
the raw image must be "demosaiced" to interpolate the missing channels
for each pixel.

Camera Raw and Color

The demosaicing process necessitates some assumptions about the colorimetric values of red, green, and blue. Most raw converters use a camera
profile to encode this assumption. Adobe Camera Raw takes a different
approach.

Camera Raw Input. For each supported camera, Camera Raw contains
not one but two built-in profiles, one for D65, the other for Illuminant A.
Camera Raw's White Balance controls-color temperature and tint-let
youinterpolate between, or even extrapolate beyondthese two profilessee Figure 12-22.
On the dozen or so cameras we've tried, we 6nd that the Camera Raw
approach allows us to get good color much more easily than we can with
any of the raw converters that rely on a single static profile-most cameras
respond very differently under tungsten and daylight, particularly in the
blue channel, and the dual-profile approach addresses this issue capably.

Chapter 19: The Adobe Common Color Archltecture

377

However, the profiles built into Camera Raw are generic profiles for
a specific camera model, and some cameras exhibit a great deal more
unit-to-unit variation than others, so the default behavior may not work
optimally with your specific unit. You can't replace the generic profiles
with custom ones, and it might not be a good idea to do so even if you
could. The Camera Raw profiles were built by Thomas Knoll using his own
proprietary profiling technology, and simply swapping in profiles built
with different technology could hurt as easily as it could help.
Instead, Camera Raw features Calibrate controls that let you edit the
built-in profiles so that they better reflect the behavior of your specific
camera. We discuss the procedure in detail in "Calibrating Camera Raw"
in Chapter 9 , Evaluatingand Editing Profiles.

Camera Raw Outpul. The destination spaces for Camera Raw are also
hard-wired. Camera Raw offers the choice of sRGB. Colormatch RGB,
Adobe RGB (1998). and ProPhoto RGB. Unless you're preparing images
for the Web, in which case sRGB is the obvious choice, we recommend
using ProPhoto RGB, because the other spaces can easily clip colors,some
of which are printable, that ProPhoto RGB does not. Figure 12-23 shows a
ProPhoto RGB image plotted against sRGB and Adobe RGB, both ofwhich
clip a significant amount of color.

378

R u l World Color Management, Second Edition

Thegarnut plots below, produced using

Chromix ColorThink, plot color in Lab
space. You're lookingat a side elevation
of the color space, with the Lightness axis
running vertically. Th.e aa*mis,fromred
togreen runs almoststraight toward
you out of the page; the baaxis,fromblue
to yellow, runsfrorn lefr to right.

Euen an innocuous image

like the one at right can
contain colors thatlie well
outside the range that
either Adobe RGB (1998)or
sRGB can represent.

The image above plotted

(as squares) against the
mlorgarnut ofAdabe RGB
(1998)(shaded solid)

These dark yellows and

oranges lie outside the
gamut ofAdabe RGB 11998)
or sRGB.

The image above plotted

(as squares) against the
color gamut of sRGB
(shadedsolid)

If you really need camera raw output in a space that isn't supported by
the Camera Raw plug-in, set it to produce a 16-bitlchannel ProPhoto RGB
image, then convert that image toyourworkingspace of choice. It's highly
unlikely that ProPhoto RGB will clip any visible colors, and any loss in the
conversion from 16-bitlchannel ProPhoto RGB to another space will be
visually (and probably numerically) insignificant-see Figure 12-24.
If it strikes you as odd that anyone as fanatical about color management as we admit to being would advocate a solution that relies on generic profiles, bear in mind the following.

Chapla 19:The Adobe Common Color A r c h i i u n

Camera Raw
worknow controls

379

Camera profiling is really a very special case. Most devices that we

profile havea fixed, unambiguouscolor gamut, whilecameras do not.
(Scanners don't have a fixed unambiguous color gamut either, but
since the materials we scan DO have a fixed color gamut, it's a moot
point.) Cameras have to deal with the entire range of tone and color
represented by the real world.
The Calibrate controls in Camera Raw do a great job of letting you
tailor the built-in profiles to match your camera's behavior, and the
process, while detailed. is a great deal simpler than building camera
profiles from scratch.
Camera Raw has many other strengths besides its color-management
feature--Brucehas written an entire book (RealWoridCnrnera Raw with
AdobePl~oroshopCS about it-but we think that the way Camera Raw
handles color is nothing short of brilliant. In fact, the profile-tweening
strategy is one we'd like to see extended to output. We can envisage, for
example, being able to interpolate between two press profiles, one with
heavy black generation, another with light black genaration, to optimize
the black plate for a specific image. All we have to do now is to convince
someone to build such a system.

But Different
Adobe deserves kudos for at least attempting to provide a consistent and
rational user interface for color across its main applications, but as we've
pointed out throughout this chapter, sometimes the similarities are misleading. Part of this is an inevitable consequence of the different capabilities of each application; part is an example of goad intentions gone
astray. The moral is to make sure that you understand the often-subtle
differences between apparently identical features in each of the applications. and use them wisely.

This page lefi blank. ..

Color
Management
n Macromedia
Creerland 10 and 11
Capable but Quirky
Macromedia must have a whole bunch of customers with an interesting
variety of legacy needs, because Macromedia FreeHand doesn't have just
one color management system. Exactly how many it has depends on your
point of view, but we'll concentrate on the ones that use ICC profiles,with
a brief overview of the others--if only so that you can wonder, as we do,
why they're even there.
FreeHand has many strengths, but we can't count color management
as being among them. With most OSs, FreeHand works (or fails to work)
as described in this chapter. But at the time of publication, FreeHand 10.0
and 11.02 both exhibit bizarre behavior that defies detailed explanation
when running under Mac OS X Panther (10.3.~).RGB to CMYK conversions at output time, and the onscreen previews of such conversions, are
intermittently unreliable and more often unreliable than not.
We can report that an early binding worldlow-placing CMYK images
and usingan assumed profile set in the Separationsprinter pop-up menuprovides decent soft proofingof placed and native content, but if you want
FreeHand to perform conversions at output time, don't trust them without
a lot of testing. FreeHand honors embedded profiles only in imported RGB
TIFF images, and lets you assign profiles to imported RGB TIFF and JPEG
images,and all CMYKelementsare invariably assumed to be output CMYK.
If your workflow is to design in RGB and convert to CMYK at output time,

383

Real World Cdor Msnagmwnt, Second Wition

FreeHand's color management can serve you well. Likewise, if you do all
your work in final CMYK, FreeHand can provide you with good on-screen
previews, and will pass the CMYKvalues unchanged to output.
For any other workflow, all bets are off. And even in these two simple
worldlows, FreeHand offers plenty of opportunities for mistakes-so in
this chapter, we'll tryto steer you clear of the shoals of confusion and guide
you to the safe harbor of smoothly functioning color management.

FreeHand's Color Preferences

The Preferencesdialog box has a series of categories on the left-hand side.
When you select Colors, the dialog box shown in Figure 13-1 appears.
F@nla-1

FreeHandColor
Preferences dialog box

Color Management Types

Onwindowssystems,the dialog box offers fourpossibilities:None, Adjust
Display Colors, ColorTables,and KodakDigitalScience. Mac OS offers the
same four and adds a fifth option, Apple ColorSync (see Figure 13-21,We
strongly recommend that you ignore the first three options and choose
either Kodak Digital Science or Apple ColorSync, which offer identical
options and functionalityin FreeHand But for those who must know, here
are the ramifications of the first three options.

C h a p t u 0:Cdor Management in Macromedia htcHand 10 and ll

-re

383

l3-

Color management types

None. This option offers no user control for either output conversions or
on-screen previewsews The conversion from CMYKtoRGBfor either displayor
for RGB output devices is controlled by a built-in and nonmodiftabletable,
as is the mnversion from RGB to CMYK output triggered by the "Convert
RGB Colors to Process" option in FreeHand's print dialog (discussed later).
Needless to say, we don't recommend this type of color management.

Ad~ul(
~ d o r rhavetwobroblemswiththis
~ e
option-ne Philosophical, one practical. It lets you change the behavior of your monitor,
for FreeHand only, in an attempt to match printed output-the approach
that Bruce d l 8 "messingup your monitor to match the print'' (though he
usuallyuses a stronger term than "messing").
The philosophical objectionisthat it negates one &the major strengths
of colormanagement. BackinChapter3wepointed out that color management reduces the number of device-to-device linksfrom nxm to n+m.This
approachgoesbackto then~mmethod,becauseyouneed to mess upyour
monitor in a different way each time you change output processes.
The practical objection is that it simply doesn't work very well. Even
when we adjust the display to match printed versions of the swatches,
we find that the swatch colors are just about the only things that match
between display and print--everythingelse is off, somerimes a long way
off. And to add insult to injury, this method only compensates for native
PreeHand elementsimported images preview inaccurately.

384

Redl World Color Management, Second Edltion

Color Tables. This type of color management is pointless. It depends on

Kodak Digital Science or Apple ColorSync, and builds ColorTables based
on existing ICC profiles. Since you have to base the tahles on ICC profiles
anyway, you might as well just learn to use ICC color management. We
see no advantage to this method, and would be glad to see it removed
from FreeHand.

Kodak Digital Science and Apple ColorSync

Referring back to Figure 13-1, we need to cover two checkboxes before
getting to the Setup dialog, the meat and potatoes of this course.
b

Color manage spot colors lets you color manage just their on-screen
preview. You don't manage their CMYK values for output, because
they're hardwired based on FreeHand's built-in Pantone tables. (In
FreeHand 10. theseare the olderpre-May2000 tables. wWeFreeHand
11 uses the newer post-May 2000 ones.) See "Named-Color Systems"
in Chapter 18, Building Color-Managed Workfows

Rebuild Color Tables uses the settings in Setup to build color tables
for use with the Color Tables type of color management previously
mentioned, but since we told you not to use this feature, let's move on
to the contents of the Setup button.

The Color Management Setup dialog box offers seven options (see
Figure 13-3). FreeHand 10 isn't particularly assiduous in looking for
profiles--on Mac OS X, it only looks in the tLibrarylColo6ynclProfiles
folder. On other platforms, it ignores both subdirectories and aliases in the
usual directories for profiles-so it doesn't, for example, find display profiles
stored in the Displays subfolder in the Mac OS 9 ColorSync Profiles folder.
FreeHand 11 successfully finds profiles in an legitimate locations.
Monitor.This is where you select your current display profile-FreeHand
doesn't get this information automatically From ICM or ColorSync.
Monltor tlmulates.The options offered are None, Composite printer, and
Separations printer. "None" literally means do no display compensation. If
you use this option, neither embedded profiles nor assumed profiles get
used for on-screen display, though they may be used for output.

Chapter 1%Color Management In Macmnmlia h d a n d10 and 11

385

If you use either of the other options, RGB images are displayed by
converting from their embedded profile or the "Default RGB image
source" profile to the Composite or Separations printer profile (depending
on which one you choose in the Monitor simulates pop-up), then to the
display profile.
CMYK images, however, are always converted for display using the
Separations printer profile as source, even if Monitor simulates is set to
"Composite printer." If you set Monitor simulates to "None." FreeHand
uses its built-in unmodifiable table as the source for display conversion
of CMYK instead.
Separations printer. This setting lets you choose a CMYK profile, which
becomes the assumed source profile for all CMYK content, including imported images, even if they have an embedded profil-FreeHand
simply ignores embedded profiles in imported CMYK. As the CMYK source
profile, it affects on-screen preview of CMYK native elements and placed
objects. If you print from FreeHand to an RGB device, the profile you
select here is used as the source profile for all Cm content. The only
CMYK-to-CMYKconversionFreeHand performsis to a composite CMYK
printer when "Compositesimulates separations" is turned on-again, the
Separations printer profile is used as the source.
IntenLThis pop-up lets you specify a rendering intent for all conversions.
It affects RGB-to-CMYK conversions at print time, and also affects RGBto-RGB and CMYK-to-RGB conversions when the Composite printer is
an RGB printer (and the output device is non-Postscript).This is the only
rendering intent control FreeHand offers.

386

Real World Color Mmagemen4, Second Wltlon

Composite simulator separations. This checkhox only affects the output,
not the on-screen preview. It makes the composite device simulate the
separation printer by converting all non-CMYK content to Separations
printer ChOK, assigning the Separations printer CMYK profile to all
CMYK content, then converting the resulting Separations printer CMYK
to the composite printer space.
All the conversions use the rendering intent you specify under Intent.
so it's impossible to use perceptual orrelativecolorimehicrendering to go
from the source profiles to Separation printer CMYK, then absolute colorimehic rendering to go from Separation printer CMYKto the Composite
printer space. If you want the composite printer to produce a reasonable
simulation of the final separations, we recommend that you set the Intent
to Relative Colorimetric.
Composite printer. Here you may select an RGB or CMYK profile for a
composite printer. If you select anRGB profile, the "Convert RGB to process" checkbox in the FreeHand print dialog is ignored, though it isn't
grayed out. But selectingan RGB profile here is quite dicey whenit comes
to printing-see "Printing," later in this chapter, for more information.
To select a profile here, you must check the "Compositesimulates separations" checkbox previously described. There's no logical reason for this;
that's just the way it is. You can temporarily check the box to change the
profile, and then uncheck it to ensure separation simulation does not
occur. Even though the selected profile is grayed out when this box is
unchecked, it's still set as the Composite printer profile.
Default RGB Image wum.The profile selected here is arrtomatical/yassigned to unragged imported RGB images rather than simply acting as the
assumed profile.We make this distinction because images imported while
profile%" isselectedwillretainpro6le'Xastheirsource ifyousubsequently change the default RGB image source to profile "B." Only subsequently
placed images will use profile "B,"Ifyou choose"None," the displav profile,
set in the Monitor pop-up menu, is assigned as the source.
The profiles automatically assigned to imported RGB images are referenced in the saved FreeHand document. but not embedded. If you open
the Rle on another workstation that doesn't have the profiles installed,you'll
get a cryptic warning dialog listing the missingprofiles when you open the

document (see Figure 13-4). The dialog says that the default RGB image
sourcewillbeused instead-that means whicheverdefaultRGBimagesource
profile is selected in Preferencesat the time the document is opened.
If.w)u get this warningdialog,
the prudent thing to do is make
a note of the missingprofiles, click
the Cancel button, thengofind the
missing profiles and
install them. Once you relaunch
FreeHand and reopen the
document, the warning will
no longer czppear:

Note that this setting applies only to imported graphics. Native RGB
elements are always treated as untagged in FreeHand, assuming the display profile as their source profile. This is annoying because the same
native elements on two different workstations have different RGB source
profiles assumed, and will print differently. To avoid major differences
when printing native elements, you need to calibrate all monitors to the
samestandard. k e n then, there'stypically enoughvariation from display
to display that you'll still get minor differences.
Imported images will display and print the same from multiple workstations, but native elements probably won't. It's amajor gotcha andoversight by Macromedia.

Manual Assignments
FreeHand offers only one manual control, for assigning profiles to imported images, and it applies exclusively to imported RGB images.

Object Palette
The Object palette--which youopen by choosingobject from the Inspect o n submenu of the Windows menu-shows you the file type and color
space, asweU as a pop-upmenu showing the source profile, of a currently
selected imported RGB graphic (see Figure 13-5).

388

Rwl World Color Management, Second Edition

Notice that the Object palette example for RGB indicates the Image
source is "- SMPTE-C." The "-" in front of the profile name indicates that
this profile was embedded in the imported image. FreeHand always uses
the embedded profile in supported RGB image formats unless you change
it in this pop-up menu (which we don't recommend).
If you do change the profile associated with an object, FreeHand immediately forgets theembedded profile. Ifyoulater decide to use the embedded profile instead, you'll need to re-import that object.
Note that the Image source pop-up menu is grayed out for the CMYK
example. Embedded profiles for CMYK images are ignored, and it's impossible to assign profiles manually. As previously discussed, the Separations printer profile specified in Color Management Setup is the assumed
source profilefor all CMYK elements, including imported images.

The undocumented rules that governhow document content is converted

~
by FreeHand's Export command to various file formats cause B N C to
muse, "Where am I going, and why am I in this handbasket?" As far as
we've been able to determine, they spell disaster. The export commands
have few options pertaining to color conversions; everything else is controlled by the color management settings in Preferences. There are so
many permutations that we confined our testing to just Apple ColorSync
behavior as it pertains to EPS, PDF, and TIFF exporting. We expect the
behavior when using Kodak Digital Science to be the same.

lip: Create Output-Ready CMYK PBF W a . To create a CMYK printready PDF file, print a Postscript file from FreeHand to disk-which will
make FreeHand color manage the output following the rules for printing
(see "Printing," later in this chapter)-then process the Postscript file in
Acrobat Distiller and make a PDEVoilB, output-ready PDF that you can
place in your page-layout application.
But for the brave or foolish. we'll document what we've learned thus
far in the hopes of shaming Macromedia into making FreeHandSsExport
features more rational in a future version of the application.

TIFF
TIFF export is always to untagged RGB. It doesn't matter if the entire
content of your document is CMYK-it will end up as an RGB TIFE The
contxols in the Export dialog box for TIFF are limited, and don't include
options for CMYK export or profile embedding.
AU CMYK elements and images get converted from the Separations
printer profile to the display profile. usingthe rendering intent set in the
Intent pop-up in FreeHand's Color Management Setup.
RGB behavioris controlled by theMonitor simulatessetting. Ifit'sset to
"None,"RGBobjects aren't color managed on export-the raw RGB values
are simply written to the TIFE If it's set to "Separations" or to "Composite," RGB objects get converted to the Separations or Composite printer
profile, respectively, then to something we've been unable to determine-perhaps a hardwired internal space--and finally, to monitor RGB.

EPS and PDF

Both EPS and PDF exporting have a Setup button to access their options.
and both offer a "Convert colors to" pop-upmenu containing threechoices: CMYK, RGB, or CMYK and RGB. The similarities end there.

EPS. The three options produce EPS files on Export as follows:

b

CMYK. This option produces an all-CMYX EPS. All CMYK objects.

whether native or imported, simply have their raw values written into
the EPS file. RGB native elements use the Monitor profile as the assumed source, and RGB imported images use their assigned profile.
Both are then converted to an unknown CMYKdestination.We haven't
been able to figure out what it is, or get an answer from Macromedia.
htlt i t dpfinitplv isn't the Gnnmtions nrinrer nrnfilp

390

Real World Color Management, kcond Mltlon

t RGB. This option produces an all-RGB EPS. Native and imported

CMYK elements use the Separations profile as the assumed source,

and are converted to an unknown RGB destination. Imported RGB images use their assigned profile as source, and are also converted to an
unknown RGB destination. RGB native elements are left unconverted,
and the raw RGB values are exported to the file.
t CMYK and RGB.This option produces a mixed-mode EPS. Imported

CMYK images, as well as RGB and CMYK native elements, are left
unmodified. Imported RGB images areconverted to CMYK using their
embedded or assigned profile as source, and an unknown CMYK destination profile.

PDP.The three options produce PDF files on Export as follows:

t CMYK.You get amixed-modePDE Imported CMYK images and CMYK

native elements are unmodified,so their raw values are exported. Imported RGB images are exported unmodified with raw RGB values
intact. RGB native elements use an unknown source and unknown
destination, and are converted to CMYK.
F

RGFJ.You also get a mixed-mode PDE Nothing is converted-raw values are exported for all images and elements.

t CMYKand RGB. Same as for RGB.

If anyone can provide a rational explanation for this set of Export behaviors, we'd love to hear it!

Printing
Printing from FreeHand using color management is relatively straightforward. The main thing you need to do at print time is to tell FreeHand
whether you're printing to the Composite printer or to the Separations
printer (see Figure 13-6).
In the FreeHand portion 01 the print diiog is a Separations checkbox
If you check it, the Separations printer profile you selected in Color Man-

Chapter D: Color Managcmcnt In Maonnedla heeHand 10 and 11

39(

FreeHand mint
dialog box

agement Setup is used as the destination profile. When it's unchecked, the
Composite printer profile you selected in Color Management Setup is used
as the destination profile. To use the "Composite simulates separations"
feanue, you must uncheck the Separations checkbox, and have previously
checked the "Compositesimulatesseparations"checkboxinColor Management Setup.

Gotchas
Didn't think it was going to be that simple, did you? Certainly not after
the hell called Export.
"Convert RGB to process" must be checked (it is by default), or else
RGB content isn't converted to CMYK at output time. This checkbox
has two locations-you can check either one. One is reached through
Output Options on the File menu. The other is in the Output Options
section of the Imaging panel in the Print Setup dialog box, which is
reached by clicking the Setup button in the Print dialog (see Figure 13-71.
F

When you print to a Postscript device, FreeHand wants almost everything to be CMYK. Even if you uncheck the "Convert RGB to process" option and select an RGB profile for the Composite printer, only
imported RGB images get converted to the destination Composite
printer profile. Everything else is CMYK, and there appears to be no
way around this, so you may have significant problems printing to an
RGB Postscript RIP.

Output Options and Print

Setup: Imaging tab

On Mac OSX, the default printer, set in the Print Center utility, must be
a PostScript printer. Otherwise, FreeHand assumes that you're printing to an RGB raster device and produces exclusively RGB Postscript,
so you'll most likely end up with a mess.

Be Careful
If you're careful, you can produce good results using FreeHand's color
management.You can either work entirelyin final CMYK, or work in RGB
and separate to final CMYK on output. If you work in mixed RGB and
CMYK, make sure that any imported CMYK is already in the final print
space. If you need to export CMYK from FreeHand, the only way that
works is to use the rather laborious process we describe in the Tip, "Create
Output-Ready CMYK PDF Files," earlier in this chapter. FreeHand's Export
features are currently unsafe at any speed!

Color
Management
It Manages Everything
But Its Own Files
CorelDMW has a rather clever user interface for color management that
makes the way it functions relatively obvious, but its use of icons instead
of menu options makes it a little hard to describe in writing.
Like Macromedia FreeHand, CorelDMW sees CMYK as press-ready, so
it doesn't do any CMYK-to-CMYK conversions except when you make a
composite CMYK printer simulate the Separations printer. Also like FreeHand, CorelDRAWonly performs conversions on output (with one exception-see "ManualControls,"later in this chapter) or on export. It supports
embedded profiles in RGB images for import and export ofTIFE JPEG,and
.PSD (Photoshop) formats, but it doesn't support profile embedding in its
own document format.
CorelDRAWsettingsare application-wide,so they apply to all open documents. CorelDRAWdocuments are always governed by only two profiles,
one for RGB and one for CMYK, but the gaping hole in CorelDRAW's color
management is that it doesn't embed these profiles in its own native files,
anddoesn't offer any othermechanism for recording the settings that apply
to a specificdocument. So the only way to color manage native CorelDMW
files is to note the settings manually, and associate them somehow (textfile,
Mac OS Finder Info) with the document-a kludge at best.

393

394

Real World Colw Management, Second Edition

Color Management Preferences

You access the Color Management Preferences by going to Color Management on the Edit menu in Mac OS 9 and X, and on the Tools menu
in Windows. The resulting dialog box shows a series of icons and arrows
representing workflow (see Figure 14-11.
figure 14-1

CorelDRAWColor
Management dialog box

CorelDRAW'sColor Management dialog box uses icons and

arrows to represent the color management workflow.

This dialog box works a little differently than most. As you move your
mouse around, the cursor periodically changes into a hand. Ifyou briefly
pause, a description of what the hand is pointing to appears (see Figure
14-2). Each icon has a description, as does each of the arrows that serve to
enable or disable color management in the direction of the arrow

Mum 14-%
Mouse wer in Color
Management dialog box

TI

When you mouse over the icons

and arrows, rollover help boxes
appear.

*rn.Ttl.a.d*

,,htb".-lo.dlkrr

To select a profile, click once on the black arrow below the icon for
the deviceifeature you want to change, and a pop-up menu of available
profiles appears-CorelDRAW finds ICC profiles in their usual locations
on each platform. In addition, three of the icons have advanced options
that are accessed by clicking once on the icon. If you click on an icon and
nothing happens, that icon doesn't have advanced options.
At the very bottom is a Style pop-up menu. These are presets-once
you've configured CorelDRAW to your liking, you can click the "+" button
to save your settings as a preset. The ColorSync Workflow preset grabs
settingsfrom ColorSync, such as the RGB default. CMYKdefault,and the
current display profile. At least on Mac OS X, we've found the behavior to
be inconsistent, so don't rely on making changes in the ColorSync panel
without also going into Corel's Color Management dialog to make sure
the profiles you want are selected.
We'll describe each settings icon-including any advanced featuresstarting with the center icon, followed by the monitor, then clockwise
from there.

Internal RCB
The center icon represents the RGB editing space used in CorelDR4W. Bv
default it's effectively disabled since its profile is the same as the Monitor
profile, so no display compensation is in effect, although impon, export.
printer, and scanner conversions can still happen.
It's important to understand that Intemal RGB is the assumed RGB
source profile for all RGB content, both native and imported. It doesn't
get embeddedinto CorelDRAWdocumentswhenyou save, so you need to
picka profile and use it consistently. If you share documents with others
who also want to use color management, you'll need to tell them what
profile you used for Internal RGB. If you change the lnternal RGB profile
while working on a document, the color behavior of all RGB content is
updated immediately to reflect the new profile. This isn't a conversion.
but a change to theassumed profile-so the numericvalues for RGB content don't change--buttheircolor meaning, and hence their appearance.
does.
When you click once on the Intemal RGB icon itself, you get an Advanced Settings dialog box (see Figure 14-3).

396

Real World Colof Managemenf, Second Ldklon

n p r e 14.3

The Advanced Settings dialog box for

Internal RGB lets you set the rendering intentand CMM for all conuersions.

Internal RGB: Advanced

Settings dialog box

Rendcrlns Intent. Rendering Intent affects all conversions except those

going to the display profile for on-screen preview purposes-Relative
Colorimetric is probably the best choice considering the lack of per-object
control.

Color Engine. Color Engine lets you choose a CMM. On Mac OS, if you
choose ColorSync, CorelDRAW uses the CMM you've selected in the CoIorsync control panel (OS 9) or System Preferences (OS 10.2 or older) or
ColorSync UtilitpPreferences (OSX 10.3).and ifyou chooseKodakDigital
Science CMM, CorelDRAWusesthat CMMforall conversions. On W~ndows,
the options are Microsoft ICM 2.0 CMM or KodakDigital Science CMM.

Monitor
This is where you select the profile for your display. When you click on
Monitor, the Advanced DisplaySettings dialog box appears (see Figure 14-41,
If you have activated either of the arrows coming from the Composite
or Separations printers, then the "Highlight display colors out of printer
gamut" option becomes available. It does exactly what the name implies-if the color can't be printed on the output device the display is
simulating, CorelDRAW superimposes a chosen warning color instead.

Monitor: Advanced
DisplaySettings dialog

box

TheAdvanced Display Settings dialog

box letsyou set thegamur warning
and control the display of spot colors.

Chapter 14 Color M.nag.mmt In CorclDRAW10

397

What determines whether the on-screen simulation is based on the

Composite or Separations printer profile is whether the arrow to the
Monitor icon comes from the Composite or Separations printer icon. If
it comes from the Internal RGB icon, the software just performs display
compensation from the Internal RGB profile to the display profile.

Map spot colon into CMYK gamut. This option displays spot colors using their Pantone-specified process equivalents (which are pre-May 2000
in CorelDRAW 10) piped through the Composite or Separations printer
profile (used as the source profile],and then to the display profile (used as
the destination profile),to showhow those CMYK builds would output on
the simulated printer. It doesn't create custom CMYK builds for Pantone
colors-it's an on-screen simulation only.

Composite Printer
This can be either an RGB or CMYKprofile.CorelDRAWsupportsfull RGBto-RGB and CMYK-to-RGB workflows. To make the Composite printer
simulate the Separations printer, click on the large curved arrow at the
bottom of the Color Management dialog, which then deactivates the arrow between Internal RGB and Composite printer.

ImportExport
By default, no profile is selected for thisicon; clickon it once and you'll get

theAdvancedImport/ExponSettingsdialogbox(see Figure 14-5).Aboved.

note that this dialog pertains only to RGB. On CMYK import, CorelDRAW
ignores embedded profiles and uses the Separations profile instead.
-

Advanced Import/Fxport
Settings dialog box

The Aduanced lmport/ExportSemmngs

dialog box lers you control profile
handling on import and export.

398

Real World Color Manrgnmnl, kcond Edilion

The first options in the import and export sections ("Use emhedded
ICC profile" and "Embed Internal RGB profile," respectively) are the settings we recommend. The others are potentially dangerous.

Import: Use embedded ICC profile. This option makes CorelDRAW use
the emhedded profile as the source profile for tagged RGB images, and
the profile that you choose here as the source for untagged RGB images.
From there. CorelDRAW converts images to the Internal RGB profile. If
you choose "None" then it assumes the Internal RGB profile for untagged
images, and no conversion occurs.

h p o k Ahraw convert using.Thisoprionignoresthe emhedded profile,

substitutes theselected profileas thesource profile, and converts theimage to thelnternal RGB space. Needless to say, wedon't think this is a good
idea. If there's a problem with the embedded profile in an image, correct
it in the originating application.

Import Ignore embedded ICC proflle.This option ignores embedded

ICC profiles, doesn't convert the image, and assumes the Internal RGB
profile.

b r t : h b c d internal RGB profik.This option embeds the Internal

RGB protile in exported RGB files.

Scanner/Digital Camera
Select a profile here to be used as the source profile for acquiring images
from the File menu's Acquire Image command. All images that come in
through Acquire Image are converted to Internal RGB when Internal RGB
and ScannerIDigital Camera are set to different profiles.

Separations Printer
This setting accommodates only CMYK profiles. The profile you choose
here is usedas thesource profile forall CMYKcontent.It's also used as the
destination profile for the conversion that takes place when you print to
the Separations printer. (Which profleis used at print time-this profile
or the Composite printer--depends on settings in the printer driver,
which we discuss in "Printing," later in this chapter.)

Tlp: Check Your Output Conwrrbm. Make absolutely certain that the
arrowbetweenInternalRGB andseparations Printer is turned on, or CorelDRAWwon't use ICC color management for its RGB to CMYKconversions.
Instead, itwill use abuilt-in table whose separationsare unconaollable and
usually pretty hideous. For the Composite printer, make sure that either the
arrow from Internal RGB or the large curved arrow from the Separations
printer is activated, for the same reason

Importing
Importing files into CorelDRAW is fairly straightforward once you know
the rules. The Import dialog box appears when you choose Impoa from the
Fie menu (see Figure 14-6). When importing a fiIe with an embedded
profile, the "Extract embedded ICC profile" checkbox is enabled. If you
check this box, then immediately after clicking the Import button for
this dialog, you're presented with a Save dialog box that lets you save the
profile embedded in the image.
UgunU.6

Import dialog box

400

Real World Color Management, kwnd Wltlon

CorelDRAW floes see embedded profiles in CMYK images, as the example figure shows, but it doesn't retain them upon import. Instead, it
always assumes the Separations printer profile as the source for any imported ChlYK, ignoringthe embedded profile.
For RGB importing. CorelDRAW honors embedded profiles and uses
them as the source profiles (if you've configured the Advanced Import1
Export Settings to use embedded profiles) in the following file formats:
TIFE JPEG, PSD, and CPT (Corel PHOTO-PAINT).We've testedTIFE JPG,
and PSD, and we're confident that those formats am reliable.
TaggedRGB images then get converted on import to Internal RGB. Unragged RGB images use Internal RGB as their assumed profile.

Manual Controls
CorelDRAW has no manual controls for per-object profile assignment.
However, it does let you do mode changes on already-imported bitmap
images, such as from RGB to CMYK, CMYK to RGB, RGB to RGB, and even
from RGB or CMYK to LAB.

To convert RGB to CMYK, select an RGR image, then choose CMYK

Color (32-bit) from the Mode submenu on the Bitmap menu. The image gets converted from the Internal RGB profile to the Separations
printer profile.
b

To convert CMYKto RGB,select a CMYK image. then choose RGB Color (24-bit) From the Mode submenu on the Bitmap menu. The image
gets converted from the Separationsprinter profile to thelnternal RGB
profile.

To convert RGB to RGB, select an RGB image, then choose Apply ICC
Profile from the Mode submenu on theBirmap menu, which opens the
Apply 1CCProfile dialog box (see Figure 14-7. The profile you choose
in the pop-up menu is used as the source profile for the conversionthe destination profile is always the Internal RGB profile.

Chaptm IS: Colw Management In ConlDRAW 10

Apply ICC Profile

dialog box

401

The Apply ICC Profile dialog box lets you clioose a

source profile for conuersions to Internal RGB.

Exporting
The previously described Color Management Advanced ImportlExport
Settings dialogbox implies that ifyou select "Embed Internal RGB profile,"
CorelDRAW will embed the Internal RGB profile in the supported formats,
and this is true.
At export time, the Bitmap Export dialog box appears, which contains an
'Rpply ICC profile" checkbox (see Figure 14-61.This control has varying
effects depending on just what you're exporting. When exporting RGB, it
seems to have no effect.
ngure 15-9

Bitmap Export and EPS

Export dialog boxes: Color
Management options

Color management in
Bim~apand EPS Export

The EPS w o r t dialog box contains a Color Management section (see

Figure 14-61. If you don't check 'Rpply ICC profile," you get a hardwired
separation instead of one that's ICC based. You can choose the Composite
or Separations printer profile as the destination using the radio b u t t o n s
CorelDRAW also displays the name of the specific profile.

404

Rwl World Color Management, W n d Edition

When you select CMYK Color from the Color pop-up menu, the document is converted to CMM(usingInterna1RGB as the source andtheSeparations Printer profile as the destination-butonly if Xpply ICC profile"
is checked. Otherwise you'll get a remarkably ugly bardwired conversion
to CMYK.

Profile Embedding
The documentation states that the supported export formats for embedding profiles areTIFF; JPEG, EPS, PDE PSD (Photoshop),CPT (Corel
PHOTO-PAINT),and CDR (CorelDRAWl.However, EPS files onlpseem to
contain a reference to the profile, not the profile itself.
TheCDR format doesn't even appear as an export format option. Since
we get no profile mismatch warning when opening an exported CDRfile,
even with an Internal RGB profile that doesn't match what should be embedded in the document, we can't tell whether or not a profile is really
embedded.
The PDF format isn't an export format option, but CorelDRAW offers a
Publish To PDF command on the File menu. CorelDRAW doesn't embed
ICC profiles in PDE at least as far as Photoshop and Enfocus PitStop Pro
are concerned.
We can confirm that CorelDRAW does embed profiles in RGB TIFE
JPEG, and PSD files. It never embedsprofiles in CMYKexports.

Printing
For color-management purposes, the options in the Print dialog are fairly
straightforward.The first option selects the Composite printer profile or
Separations printer profile as the destination for RGB-to-CMYK or RGBto-RGB conversionsat print time (seeFigure 14-9). In the Print dialog, under the Separations tab, the first option on the upper-left corner is "Print
separations." When this is unchecked, the Composite printer profile is
the destination, and when it's checked, the Separations printer profile
is the destination. To produce simulations of the Separations printer on
your composite printer, uncheck"Print separations." and select the large
curved arrow that goes from the Separations printer to the Composite
printer in the Color Management dialog box. Piece of cake.

The"Pn'ntseparations"
checkbox controls
whether print is
converted to the
Separations or the
Composite printex

Print dialog bcac

Separations tab

The Misc tab of the Print dialog box contains two options that are easy
to overlook.
Fist, the "Apply ICC profile" checkbox is available and checked by default in two situations:
r The arrow from Internal RGB to Composite printer is enabled in the
Color Management dialog box, and the "Print separations" checkbox
is unchecked.
b

mure l4-10
Print dialog box:Mise tab

The arrow from Internal RGB to Separations printer is enabled in the

Color Management dialog box, and the "Print separations" checkbox
is checked (see Figure 14-10).

404

Real World Color Management, h n d Edition

Second, you must pay attention to the "Output color bitmaps as" popup menu ifyou want RGB output. Even if you select an RGB profile for the
Composite printer and it appears under "Apply ICC profile," if you don't
select RGB from the "Output color bitmaps as" pop-up, your images get
converted to CMYK or grayscale instead.

Strengths and Weaknesses

CorelDRrlM'lets you work entirely in CMYK, using color management for
on-screen previews and proofing. It successfully color manages exports according to the Color Management settings and a simple set of rules for a
limited set of file formats.You can print to RGB printers, but native elements
built in CMYK may not turn out the way you want. The biggest plus for
CorelDRAW is its Color Management dialog box user interface.The glaring
weakness that it doesn'tembed theprofiles thatdefine RGB andCMYKinto
its own documents. So it's relatively easy to export color managed content
from CorelDRAW,but the only way to manage CorrlDRAW fdes themselves
is to keep track, manuallv,of the RGB and CMYKprofilesyou used to create
the document.

Color
Management
Incremental Improvements

In print production, QuarkXPress is still very much the 800-pound gorilla,

so we can't ignore it, but we find a certain irony in the fact that despite being one of the first applications to offer color-management capabilities way
back in the early 1990s,those capabilities have improved surprisingly little
in the intervening decade. In this chapter, we cover QuarkXPress 3.32r5
through QuarkXPress6 because many sites stilluse older versions of it, and
the third-party solutions we discuss work with all three versions.
QuarkXPress 3 effectively lacks built-in color management, and the
QuarkXPress 4 CMS lies somewhere between almost-usable and totally
maddening. QuarkXPress 5 offers color-management capabilities that are
generally useful, with some annoying limitations. The color management
capabilities in QuarkXPress 6.1 are essentially unchanged, though some
particularly annoying bugs have been fixed. Nevertheless, we have effective solutions for all three versions.
So fear not. In this chapter we'll look at solutions for the most common
QuarkXPress output workflows: CMYK output destinations, RGB output
destinations, the Web, proofing, and miscellaneous-grayscale and duotone images, spot colors, and spot-to-process, simulatingspot colorswith
RGB or CMYK builds.

406

Rral World Color Management, Second Edition

Turning It Off
The simplest way of doing something is usually the safest. and often the
most efficient. Ifyou use early-bindingworkflows. thesafe and simpleway
to handle color management in QuarkXPress is to turn it off.
All the elements you import into your QuarkXPress documents-images, vector art, etc.-must be "print-ready" and already converted to final
output CMYK. You need to do this anyway with grayscale, duotone, tritone, and quadtone images, all of which must be targeted specifically for
the output method in question before you place them into QuarkXPress.
Any CMYK colors specified in the Color palette must also be outputready, because their actual color appearance depends on the output
device--a CMYK build won't look the same on newsprint as it will in a
magazine. Consider creating your CMYK builds in an application such as
Photoshop or Illustrator. then simply using the same numbers in QuarKPress, and not worrying about display discrepancies-they'll print the
same even if they look different.
Note that with color management turned off in QuarkXPress,you can't
use RGB images and native elements-becausewhile they'll often look OK
onscreen, they'll either separate poorly or wind up entirely on the black
plate. CMYK images, on the other hand, print according to the numbers
in the file, but often look like neon onscreen.

Turn Off QuarlcXPress 3 CMS (Mac OS Only)

-

QuarkSP~~ss
3.3.215, the most recent and lastwrsion ofQuarWress 3, included a color management system called "EFIColor." It predated the KC
profile format, so it used its own proprietary profiles, never worked well,
and is best avoided-try finding software for making EFIColor profiles if
you enjoy exercises in futility.
Turning color management off is the only practical solution in QuarkXPress 3, unless you resort to a third-party XTension such as Compass R o
XT (which we discuss later in this chapter). To do so, you need to delete
three components on the Macintosh: the EFIColor system extension, the
EFIColor folder, and the EFlColor XTension.

Chaptcr 1%Color Manwmcnt In OwrkXPna

407

Turn OH OuarkXPress 4,5 and 6 CMS

The first method is to close alldocuments, then go lo the Edithferences:
Color Management Preferences window and uncheck "Color Management Active." It's wry irrzpormnt that you do this while all documents are
closed. otherwiseyou only turn off color management for the foreground
document.
The absolute, sure-fire method of disabling color management in
Quar!&Press 4 and 5 is to physically remove the Quark CMS XTension
from the XTensions folder, found inside the QuarWress application
folder.

Built-in Color Management

-

We've already told you the bad news about EFIColor, the built-in CMS
in QuarkXPress 3 for Macintosh-get rid of it before it hurts you. The
Windows version has no built-in color management whatsoever. Essentially, there's no built-in color management in QuarWress 3.32 on either platform. In a way, this isn't a bad thing because you can be certain
QuarWress 3.32 will send the RGB or CMYK values in the document as
is, without massaging the data first.

OuarkXPress 4
QuarkXPress 4 includes the Quark CMSXTension, which brings a certain
level of ICC support. It works in conjunction with ColorSync on Mac OS,
and ICM 2 on Windows 98 and higher.
You need to be either brave or nuts before deciding to use the Quark
CMS in this version of QuarkXPress. In theory, it will color manage any
RGB or CMYK TIFF and native colors from the Color palette, and let you
convert them to a CMYK or Hexachrome destination (with no control over
rendering intents). It sounds limited, and it is, but the Limitations aren't
the issue. The real problem is that QuarkXPress 4.1 1 has enough bugs to
qualify as a lifetime employment program for the Orkin man. QuarKPress4.04 and 4.1 don't haveas many problems,but they still haveenough
to be considered dangerous.

408

Real World Color Management, lacond Edition

Here's a short laundry list of QuarkXPress 4.11 CMS problems:
b

QuarkXPress 4 sees, and claims to use, embedded profiles in RGB and

CMYK images, but it doesn't actually use them. Instead, it uses the
Default Source Profiles. To make matters a littlemore complex, when
you change the Default Source Profile, imported images with embedded profiles [which QuarkXPress ignores) continue to use the Defnrrlt
Source Profile that was in effect at the time t h q were placed as their
source profile. Untagged images, however, will use the new Default
Source Profile.

Separations from RGB to CMYK produce different results depending

on whether the source profile is embedded. or manually selected in
the Profile Information palette.
In the Print dialog box, in the Profiles tab, the "Composite Simulates
Separation" option implies that QuarkXPresscan cross-render final
CMYK to acompositeprinterfor proofing.Itcan't.Whethertheoption
is checked or not, it makes no difference in output.

A cautious person might be able to incorporate all of the limitations

and bugs into an effective workflow using the Quark CMS, hut we're
hard-pressed to think of any real benefits to doing so, considering the
potential for disasters. If you need built-in color management for specific
features such as soft proofing, we recommend reading the next sections
on QuarkXPressS and 6's CMS, and the coverageof InDesign CS in Chapter 12. The Adobe Common Co1orArchitechrre,then upgrading to either
of those packages.

QuarkXPress 5
'I'heQuarkCMSin QuarkXPress 5 is, according to Chris, "nearly usable." It
doesn't support color management of EPS, unlike InDesign CS: however
like InDesign CS it also doesn't color manage grayscale files. These are
limitations of which you need to be aware.
If you fmquentlv use EPS or grayscale images, you can? really use the
"Composite Simulates Separation" feature for proofing, because some
content proofs correctly while other content is left untouched by Quark&
Press. And while QuarkXPress 5 supports RGB output device profiles in

the Composite Output pop-up menu. QuarkXPress disables "Composite

Simulates Separation" when one is selected. Therefore, built-in proofing
is only available for CMYK Composite devices. For such worMows, a solution that effectively handles all content is n e e d e d s u c h as Compass
Pro AT,discussed in this chapter, or some of the solutions we discuss in
Chapter 17. Automation & Scripting.
The biggest question mark with the version 5 Quark CMS is in regards
to RGB-to-CMYK conversions. Cenain combinations of profiles produce
discrepancies of 5%-10% compared to using the same profiles. CMMs,
and other settings in Photoshop. In our experiments, the most likely suspect to trigger this problem is when the RGB image uses a wide gamut
RGB profile--suchaswide Gamut RGB. Prophoto RGB, and somescanner
profiles--as the source profile.
Because of the uncertainty that surrounds this particular issue, we can't
recommend using the Quark CMS for RGB-based workflows. But at the
same time, we can't notrecommend using it at all. The following features
all appear ro work reliably:
t Soft proofing: The Quark CMS properly displays RGB and CMYK im-

ages, including the ability to simulate the output from the composite
or separation printer onscreen.
t Hard p m o h g : The Quark CMS "Composite Simulates Separations"
feature lets you produce proofs on a composite PostScript printer, but
only using relative colorimetric rendering. The option isn't available
Tot non-Postscript printers.

However, a major practical problem is that by default placed CMYK

images are set to no color management to CMYK destinations. That
includes proofing. Unless you change this setting prior ro image placement in preferences, or when importing the image, CMYK images are
not color managed. After image placement the only option is to check
"Color Manage to CMYK Destinations" in the Profile Information palene for each image. This is wry tedious and probably impractical for
many workflows unless the job was originally created with color management for CMYK content in mind.

410

R r l World Color Management, Second Edition

CMYK-to-CMYKconversions:
used for repurposingimages for output
processes other than the ones for which they were originally separated.
RGB-to-RGB conversions: for RGB workflows with RGB source images
outputting to RGB destinations. CMYK-to-RGB is also supported.
Embedded Profiles: QuarkXPress 5 both recognizes and uses embedded profiles in TIFF only. Embedded profiles in other formats are ignored, which is obviously potentially dangerous. Once ignored the
Default profiles apply instead.
The most useful feature by far is soft proofing. With the caveat that
onlyTlFE IPEG. PICT, andnative colors used in the Color palette are color
managed, it can be helpful to many workflows. No third-party product is
capable of bringing customizable soft proofing to QuarkXPress; however
Enhance Preview XT does produce a close SWOP simulation.

OuarkXPrw 6
The QuarkCMS in QuarkXPress6 has essentiaflythe same limitationsand
bugs as QuarkXPress 5. QuarkXPress 6.1 fixes at least one problem. Profiles are now found in all of the locations in which profiles can be stored
on Mac OS X, including subfolders, and it correctly resolves aliases of
profiles. One other problem that seems to be fixed is that RGB-to-CMYK
conversions now match those of Adobe applications, minus the effect of
Black Point Compensation, with all the profiles weGe tested.
QuarkCMS continues to be relatively simple to use ifyour requitements
are simple because by default it treats untagged and tagged CMYK the
same. Neither arecolor managed. It's also easy to configure for soft proofing only, if you have a CMYK workflowand that's all you care about.
But QuarWLPresshasn't lost its ability to annoy the color management
aficionado either. Besides glaring problems like ignoringembedded profiles in formats otherthanTIFF,even seeminglysmall things like automatically grabbing the display profiie from the OS are still not implemented.
QuarkXPress PDF export has the same color management capabilities as
printing, but it can produce neither PDFIX-la nor PDFIX-3-you'll need
a utility that can bring the appropriate PDFlX compliance level to your
PDFs.

Color Management Preferences

Pay particular attention to the behavior of the Color Management Preferenceswindow.To settheapplication-levelcolor-managementpreferences,
you must choose Edit:Preferences:Color Management with no documents
open. On QuarkXPress6 on Mac OS X find the QuarkCMS option in the
QuarkXPresssPreferences.The title bar then reads, "Color Management
Preferences." This application-level setting serves as the default settings
for any newly created documents. If you have one or more documents
open, the preferences apply only to the foreground document, and the
title bar reads, "Color Management Preferences for cdocumentname>."
The Color Management Preferences window is where you configure the
Quark CMS (see Figure 15-1). The interface is the same in QuarkXPress 5
and 6. We'll now decode it for you:

QuarliYPress Color
Management Preferences
Undo
XTenrions Manager
Font mpping
avenuc.qumrk
Fik List
Dzfault Path
Full RI Pmiew

When you choose Color

Management Preferences
with no documents open,
you set theglobal
application-level colormanagementpreferences.
lfyou choose Color
Management Preferences
when one or more
documents are open,
you set the preferences
for thefo~eground
document instead

414

Real World Color Management, Second Edition

b

Color Management Active checkbox:This turns color management on

and off.

r Destination Proliles, Monitor: This is where you set your current display profile. The Quark CMS doesn't grab your display profile from the
operating system automatically. This may not sound like a problem,
but ifyou open alegacy document three months from now, it will have
its own color management preferences associated with it. including a
display profile that probablv bears no relationship to the current one.
b

Destination Profiles, Composite Output: This is the profile for

your composite printer. It can be an RGB or CMYK printer. In most
worMows, it's the local color printer used for proofing, but in nonpress worMows it may be your final output device.
In workflows with a contract proofing device that simulates the press,
you'll want to specify the press profile here as well as in the Separation Output pop-upproofing systems and presses are effectively the
same device in these workflows.

Destination Profiles. Separation Output: This is the profile for a separation device, such as a printing press. The only profiles that appear
in this pop-up menu are CMYK and six-channel profiles. The way it's
actually triggered for making conversions at print time is by checking the "Separations" checkbox in the QuarkXPress Print dialog--see
Pigure 15-6. When unchecked, the Quark CMS uses the Composite
Output profile.

Defaulr Source Profiles, RGB, CMYK, and Hexachrome:Within the respective tabs, you can select default source profiles for images and
solid colors. By default, they affect untagged images and all solid colon. For images, you canuse either theGet Picturedialog box or Profile
Information palette to specify a profile other than the Default Source
Profile. QuarkXPress doesn't let you select RGB output device profiles
as Default Source Profiles.
?he Rendering Intent pop-ups serve as default rendering intents when
you use "Get Picture" to place images. We usually set all of the tabs to
Relative Colorimetric for RGB and CMYK images. Relative Colorimetric
for RGB Solid Colors, and Absolute Colorimetric for CMYK Solid Colors.

Chapter 1%Color Management In Q ~ r k X R r n

413

Solid Colors: These are colors specified in the Color palette used for
text, shapes, or backgrounds. The models affected are RGB, JAB, and
CMYK only. If you add Pantone. Toyo. Trumatch, or DIC colors, color
management doesn't apply-instead, you get the hardwired RGB or
CMYKvalues specified in those palettes.

Colot Manage RGBSources to RGB Destinations/Color Manage CMYK

Sources to CMYK Destinations:These options let you allow or disallow
RGB-to-RGB or CMYK-to-CAWconversions. Checking the box lets
you place images targeted for one kind of color space or device and
repurpose them for a different output device. For example. placing
all images separated for SWOP, then selecting a newspaper profile,
repurposes the images for newsprint.
In another example, you could assume all CMYK images are "press
ready." Unchecking the box in the CMYK tab prevents CMYK images
From being converted by the Quark CMS. At the same time, RGB-toCMYK conversions for output and onscreen simulations are still allowed to happen.
Conversely, in RGB output worMows. you can disallow RGB-to-RGB
conversions, assuming all RGB content is output-ready, while still allowing CMYK-to-RGB conversions.

But, ifyou uncheck"Color ManageRGBSources to RGB Destfnatlons,"

it completely disables Display Simulation for those images as well.
Therefore, they don't preview as they'll print to either RGB or CMYK
output devices.The monitor isan RGB destination, but this checkbox
would be a lot more useful if its effccts were confined solely to RGB
printers, rather than to any RGB destination, including the monitor.
And there's an annoying bug. "Color Manage CMYK Sources to CMYK
Destinations," when unchecked, passes CMYKvaluesdirectly to CMYK
output with no conversion, as one would expect But they preview as if
they tvozrM be converted, so the onscreen simulation is incorrect, and
the setting also prevents placed CMYK images from proofing to CMYK
destinations.

414

Real World Color Manageml, Second Edilion

Hexachrome tab: A couple of aspects of the Hexachrome tab are a

bit misleading. First, it should be named "Six-Color" because it will
actually use any six-color ICC profile--not just Hexachrome, which
is a proprietary six-color ink set developed by Pantone. Second, we're
confused by the presenceof a Default Source Profiles tab to begin with
since Q u a r m s s doesn't color manage DCS 2.0 files, which are the
only way we know of to save six-channel images and get them into
QuarkXPress in the first place!
Display Simulation:This option tens QuarkXPresshowthedisplaysimulation should hmction. Nonemeans exactlythat: don't simulateanything
on the display. Monitor Color Space means convert from the source
profile to the display profile for display only. Composite Output Color
Space makesthe display simulatethecomposite printer.Separation Output Color Space makes the display simulate the Separations printer.
Note that Display Simulation isn'twired to the "Separation" checkbox
in the Print dialog that controls whether the Quark CMS converts to
the Separation Output or the Composite Output profile. So it's possible for the display to simulate one device when you actually intend
to print to the other.
The Quark CMS only looks at the root level of the ColorSync profiles
folder, so profiles stored in subfolders within the ColorSync Profiles folder aren't visible. This includes the Profiles and Recommended aliases to
folders containing Adobe profiles such as ColorMatch RGB, Adobe RGB
(19981, and U.S. Web Coated (SWOP) v2, among others. If you want access to those profiles from inside QuarkXPress, you need to put copies of
them in the main ColorSync Profiles folder-aliases of ICC profiles don't
work either.

Manual Controls
QuarWressoffers only hvo manual controls. The controls in Get Picture
apply when you're opening images for import, while the Profile Information palette applies to images that have already been imported.

Chapter 8: (dor Malugcnrmt In QuclrkXPrar

415

Get Picture
The Get Picture dialog box appears when you use the Get Picture command from the Fie menu (see Figure 15-2).When color management is
active for the current document, an extra tab called Color Management
appears at the bottom of the Get Picture dialog box.
mure 15-9
Get Picture dialog box

The Color Management

tab lets you set the
assumed source profile
and the rendering
intent for the image.

Perhaps it's a small bug or user interface oversight, but the Color Management tab contents remain grayed out unless you check the "Preview"
checkbox in the upper-left comer of the dialog.
The Profile pop-up menu snaps to Embedded if the image you select
has an embedded profile, or to Default if the image is untagged.You can
override these settings by selecting another profile from the pop-up list,
which then becomes the assumed source profile for the image.
The Rendering Intent pop-up snaps to the intent specified under Default Source Profiles in Color Management Preferences.You can override
this here, if you wish.
The "Color Manage to RGBICMYK Destinations" checkbox reads RGB
when you click on RGB images, and CMYKhen you click on CMYKimages.

416

Real World Color Management, Second Wltlon

It has the same functionasthe checkboxfound in Color Management Preferences in the Default Source Profilessection, and it uses the settingsspecified
there as defaults. You can override this here, if you wish.

Profile Information
You open the Profile Information palette by choosing Profile Information
from theview menu with Quarmress 5 and earlier. 1nQuarlSPressG.You'U
find Show Profile Information in thewindow menuits controls offer identical functionality to the Get Picture dialog's Color Management tab (see
Figure 15-31, It applies only to the currently selected image. As previously
mentioned, the "Color Manage to RGBICMYK Destinations"checkhox has
some bugs.We recommend you review them, and be careful.

ProGle Information
palette

The Profile Infbnnatiorl

palette lets you change the
assumed source profile
and rendering intentfor
the selected image.

Printing
The Print dialog box, which you open by choosingmint from the File menu,
has some settings that apply whether color management is active or not.
and others that only become available when color management is active.

With or without color managemni.TheOutput tab of the QuarkXPress

Print dialog box deserves a brief explanation (see Figure 15-4l.When
color management is inactive, the Print Colors pop-up menu only shows
the supported modes for the currently selected printer. If it's a nonPostScript inkjet printer, for example, Composite CMYK isn't an option.
If it's aPoslScript printer, both CompositeRGB and Composite CMYKare
available oplions. This is normal.
When color management is active, however, the Print Colors pop-up
menu grays out either Composite CMYK or Composite RGB depending on
the profile selected in the Composite Output pop-up menu. If you select
a CMYK profile, Composite RGB is grayed out, and if you select an RGB
profile. Composite CMYK is grayed out. This is also normal.

Print dialogbox:
output teb

With color managernentacfive, the Print

Colors pop-up menu
only shows the mode
that matches de profile
specifiedfor the
Composite printer

With color manrg.ncnt. The Profiles tab of the Print dialog box is a
shortcut to the Composite and Separation Output profiles, and is only
available when color management is active (see Figure 15-5).If you
change them here, they only change for the current document, and the
change is reflected in the document's Color Management Preferences
as well.

Print dialog box:

Profiles tab

Profile selections you

makein the Proples tab

of the Print dialog only

apply to the current
document, and the
change is then reflected
in the document'sColor
ManqgementPreferences.

418

R u l World Color Management, kcond Editlon

You invoke XPress's built-in hard proofing by checking the "Composite

Simulates Separation" checkboxin the Profiles tab of the Print dialog box.
It always uses relative colorimetric rendering, so simulating the source
white on yourproofisn'tpossible. The potential gotcha is the "Color Manage CMYK Sources to CMYK Destinations" checkboxassociatedwith each
image. When this is turned off, the CMYK image isn't color managed at
all, including to the Composite Output device. The CMYK values in the
image pass straight through to the composite printer, so no simulation
takes place.
Last, and certainly not least, you may have noticed there are two possible output profiles: Composite Output, or Separations Output. Which
one is used as the destination? In the Layout tab of the Print dialog box
is a "Separations" checkbox When unchecked, the composite protile is
used, and when checked the separations profile is used (see Figure 15-6).
If you print composite Postscript to your imagesetter or platesetter RIE
you need to select your final output profile as Composite Output when
it's time to print.

b u m 154
F%int dialog b m
Layout tab

..

The '-checkbox,
shown in red for e m r ~ h ~ c o n n o l s
not only whetheryou print compositeors&rakd output, but also the
destination profile for all conversions.

Hard Proofing
-

QuurLXPress3 has no press simulation features, and the ones in QuarkX

Press 4 don't work, so hard proofing using built-in color management
isn't applicable. QuarkXPress5 and 65 built-in method of producing hard
proofs on composite devices has some limitations: if you useTIFF images
only: can always use the "Color Manage CMYK Sources to CMMC Destinations" option: and don't need to proof Pantooe, FocolTone, or other
named color systems, the built-in method works quite well. But if pour
worktlow relies on EPS files, or Pantone. Focoltone, orother named color
systems, vou'U need to look for a third-party solution.
Third-party proofingsolutionscome in several forms. Praxisoft's Compass ProXTXTension,colot servers, and self-contained proofingsystems
all have strengths and weaknesses, including cost, support, and deployment issues. We recommend you see a demo of any products you're interested in. to be sure that they'll meet your needs. See the next section
on Compass Pro XT for more information.

Vector Ro.Compass Pro XT's companion application. Vector Pro. can

simulate Pantone or Focoltone colors using hand-tuned RGB. CMYK, or
multichannel builds. Vector Pro then outputs these customized, outputdevice-specific colors asvector Pro palettes, which you can import into
Compass ProXT. Withoutvector Pro. Compass ProXTdoesa decent job of
simulatingsolid colon, but if you're looking for an extra level of accuracy.
you'll want Vector Pro.

Pro XT
Compass Pro XT, discussed in more detail below, is an XTension for
QuarkYPress 3 , 4 , 5 and 6 that lets you turn any profiled printer with a
sufficiently large gamut into a proofing device, and offers an assortment
of other useful features. It's published by Praxisoft (www.praxisoft.com).
and the currently available versions are: v2.2 for QuarkXPress 3 and 4, v5
for QnarkXPress 5, and v6 for QuarkXPress 6.

420

Real World Color Manilgrmant, W n d Mltlon

Pros. For individuals and relatively small organizations, this is probably

the least-expensive way to produce proofs with existing equipment. If
you can also exploit other features offered by Compass Pro XT,its value
increases exponentially.

Cons. This product is feature-rich, and despite decent documentation

and a fairly straightforward user interface, it has a learning curve-it's
not a point-and-print kind of system. In larger organizations, even with
the price breaks P d s o f t offers for purchasing multiple quantities, you'll
quickly find yourself in a price range comparable to that of the othersolutions.
Compass Pro XT can convert documents for any RGB or CMYK destination, be it the Internet. a large-format inkjet, or a printing press. It can
also convert documents forproofingpurposes, soyour lower-cost printer
can simulate find output. It's llexible enough to be used in virtually any
workflow calling for this kind of functionality from within QuarkXPress.
The file types it supports include not only the usual TIFE PEG, and
PlCT suspects, but also EPS (including IUustrator EPS). It also manages
QuarkXPress Color palette colors (in RGB. LAB, and CMYK modes), as well
as Pantone and Focoitone colors. It doesn't, however, support PDE
Compass Pro XT offers three unique features:
r it's totally happy allowingmultiple source profiles in a document: you

can place images directly into your documents whether they're from
Photoshop, digitalcameras, scanners, or stockphotog~aphy,each with
its own source profile.You can mixand match RGB and CMYKimages
into the samelayout, and Compass ProXT converts nll images (except
for grayscale, which it leaves untouched) properly to the desired destination.
b

It's the only way to color manage EPS files inside a page-layout program. Not only does it color manage RGB. L4B. and CMYK colors inside EPS files, but also Pantone and Focoltone colors. I t even color
manages EPSs or TIFFSembedded within an imported EPS file.

It supports RGB output device profiles. so in an RGB-destination

workflow, it lets you color manage content to RGB output devices in
all versions of QuarkXPress, including both RGB final output devices
and RGB proofers.

Chap-

15: C o b Management in QuarkXPrm

bm

Another useful feature is the ability to make a duplicate of an entire

QuarkXPress document. images and all, converted to a specific output
profile-so you can rework those files instead of the originals, or send
the press-ready duplicate back to the originator, to the service bureau,
or to the printer.
One last note is that the features in Compass Pro XT are extremely customizable. Even hard-core CMYK-only output wor!dou.s going to known
specific presses can take advantage of this XTension. For example, you
can disable all CMYK-to-CMYK conversions. but enable RGB-to-CMM(
conversions to catch the occasional RGB image that slips through. You
can also take advantage of Compass ProS solid-to-process conversions,
which use builds calculated for vour specific press condition instead of
the generic ones published in the Pantone Solid to Process Guide. Get a
copy for each workstation that does preflighting to weed oul RGB images
that haven't been separated yet, and rebuild solid-to-process colors correctly as a standard operating procedure.

QuarWress still dominates the page-layout market despite some stiff

competition from Adobe's InDesign, but its color management capabilities, though greatly improved in version 5 and 6, still have rather more
than their fair share of quirks. Nevertheless, as long as you're aware of the
limitations and occasional oddities, you can make the Quark CMS work
well in many workflows. For those worktlows that need a more capable
solution, m e s s is well supported both by XTensions such as Compass
Pro XT and by external color servers.

This page lefi blank. ..

Color

and PDF
The Wave of the Future

"Once upon a time, there was a great ... no wait, wrong story" is often
Chris's initial thought when the subject of PDF arises. So much promise,
so many problems (just likecolormanagement, sometimes), but anything
worthwhile usually involves a fair hit of work. Over the years. PDF has
been touted as the solution to so many different problems--as Bruce likes
to say, "it's a dessert topping and a floorwax"-that it$ often been hard to
figure out just what it is good for.
PDF workflows can bring numerous benefits and numerous pitfalls.
most of which we'll ignore in this chapter. We won't attempt to compete
with the variety of excellent books and primers on PDF: instead we'll simply give you a status report on the interaction between PDF and color
management. We can tell you a lot, but there are still a huge number of
unknowns, so you're doubly on the bleeding edge when you deal with
color management in PDE
Why a chapter on PDF and not on Acrobat? First, PDF is much bigger
than just Acrobat-many PDF workflows don't even use Acrobat-so the
real story is about PDE The Portable Document Format (which is what
PDF stands for) is designed as a platform-neutral container for all types
of digital content, allowing that content to be distributed without requiring the recipients to possess, or even have any knowledge of, the various

49.4

Real World Color Management, Second Edltlon

applications that generated the various components in the document.

This flexibility is a mixed blessing for print publishers, so much of the ongoing effort in developingstandards for PDF workflows involves defining
subsets of PDF features that define what is and is not allowable in PDFs
designed for print workflows.
It's important to keep in mind that there are two halves to PDF: the file
format half of PDF that gets most of theatrention, and the equally important but often-overlooked usage half.We intend to focus primarily on the
usage of PDF in a color management context, but to do so, we first need
tn talk ahout the various flavors of PDQ that interest us.

The Flavors of PDF

Unlike IJostScript, the Portable Document Format is an open oneAdobe publishes current versions of the specification so vendors can
decide which aspects of each version they want to implement. Sounds
like a recipe for chaos, and in some ways it is, hut there's also a method
to the madness.

PDF Versions
There are four basic versions of the PDF specification: we're primarily
interested in only one of them, but we mention the others for the sake of
completeness.
b

PDF 1.2. This is the version of PDF produced by Acroba~3,and it's

completely KC-unaware, lacking any mechanism for supporhg ICC
profiles.

PDF 1.3.This is the highest version of PDF produced by Acrobat 4, and

the default for Distiller5. It supports the embeddingof KC profileson
a per-object basis.This is the version of PDF that's most interesting to
the color management world.
PDF 1.4. This is the version of PDF produced by Acrobat 5, and the
highest version produced by Distiller 5. It too supports ICC profiles.
along with transparency and a host of other features that lie outside
the scope of this book.

Chaptu 1C Color Marugemnt and PDT

b

445

PDF 1.5. Acrobat 6 Standard and Professional will produce PDF conforming to this specification. It supports ICC profiles, transparency,
better compression including JPEG2000,enhancements to encryption, and again all sorts of things outside the scope of this book.

Embedded prdla. PDFsupports multiple objects (images-both vector

and bihnap-as well as text, shapes. backgrounds,etc.).PDF can contain
objects in different color modes, and each object can have an ICC profile
associated with it as long as that mode supports ICC profiles.You could.
for example, have multiple RGB. CMYK, and grayscale objects, each with
a dierent source profile.
To minimize file size, each object references its source profile instead
of having the profile embedded in the object-so six objects with Adobe
RGB (1998) as their source profile don't cause six copies of Adobe RGB
(19981to beembedded in the PDE but rather just one copy referred to six
times.These objects have acolor space called "KC-Based." which means
they're considered device-independent.
Any object that doesn't reference an ICC profile is device-dependent,
denoted as: IDeviceRGB, IDeviceCMYK, and IDeviceGray, as the case
may be.
Other noteworthy items. Rendering intent is set at the time the profile is
embedded and is nor related to the default rendering intent in the ICC
profile. Also, destination profiles are not embedded in PDF documents,
at leastnot yet.

PostStript Color Management

Postscript's color-management philosophy has affected PDF's color-management philosophy, and thus its behavior, so a little history is in order.
Postscript color management predates ICC-based color management,
and is built-into RIPSwith Postscript version 2017.m and greater. PostScript profiles are always unidirectional, so there are always at least two
profiles for an output device. One profile is used as a source profile(device
to PCS] called a Color Space h a y , or CSA for short. The other is used as
a destination profile (PCSto device) called a Color Rendering Dictionary.
or CRD for short. CRDs only contain a single rendering intent, so there's
a separate CRD for each rendering intent.

496

Real World Color Hanagrmmt, kcond Edition

Fundamentally, PostScript color management operates in the same

way KC-based color management does: CSAs and CRDs. like profiles.
establish a relationship between device values and PCS values, and the
color management system uses the information they contain to convert
the device values, using a specified or default rendering intent.
Where do CSAs and CRDr Ihc?The usual location is in the RIP itself, and
it's dl-too-common for built-in CSAs and CRDs to have no basis in reality either as assumed source profiles or as descriptions of the behavior of
the output device in question, so the resulting conversions are frequemlv
hideous. Recent products--in particular, newercolorlaserprinters--have
more reasonable RGB CSAs and CMYK CRDs, so the printers can receive
RGB graphics and convert them properly to printer CMYK. But that's still
the exception rather than the rule.
The other location is in the PostScript print stream itself.The RIP then
uses the CSAs and CRDs in the Postscript file instead of the built-in ones.
If you create a PostScript stream that contains CSAs but not CRDs, the
built-in CRDs are used.
In theory, each object gets its own CSA (sourceprofile) and is also tagged
with the CRD that should be used (because the CRD controls the rendering intent). Then the RIP uses each CSA and included CRD to perform
color space conversions in the RIP In practice, though, you quickly find
out that some RIPSignore CMYK CSAs altogether, others ignore CSAs in
the print stream and use only built-in CSAs, and still others ignore CRDs
as well. When it comes to implementing such a system, you usually end
up with a mess (the fact that the most common page-layout program.
QuarkXPress, can't produce either CSAs or CRDs doesn't help).
PoslScriptlo PDF, PDFlo PoslSerlpt. Enter PDF 1.3, which uses KC profiles,
not Postscript color management. Adobe created a mechanism whereby
with PDF 1.3 and higher, profiles and CSAs can be interchanged when the
document is converted from PostScript to PDF and vice versa. When you
print a PDF to a Postscript printer, it's turned backinto PostScript, and all
the ICC source profiles become PostScript CSAs. When you convert PostScript into PDF, all CSAs are convened into ICC profiles, and any CRDs
are simply tossed into the bit-bucket.

Chapter 16: Color Managcmenf and PDF

497

You may have experienced Acrobat Distiller 3 and 4 converting CMYK

images into LAB.This generally happens when those versions of Distiller
encounter a Postscript CSA embedded in a CMYK EPS (probably from
Photoshop with the "Postscript Color Management" checkbox selected).
Distiller 5 either converts the CSA into an KC profile or ignores it entirely,
depending on the Distiller color settings.
With the possible exception of thoroughly troubleshot proofing systems, we recommend avoiding Postscript color management unless
you're one of the handful of people who've already gotten it to work reliably. If you have, congratulationsmaybe you should write a book!

PDF/X
PDF supports many different types of digital content, but it doesn't
really help your PDF print workflowat all when thePDF contains the sixth
Brandenburgconcerto orthelatest Starwars movie. So the PDFlXfomat
and usage guidelines were created to ensure that PDFIX compliant PDFs
contained only the data required for print. PDFlX is rapidly emerging as
the standard for PDF-based print workflows.
Did we say "standard"? Of course, life is rarely that simple, and while
the PDFIX bandwagon continues to gather momentum. PDFIX isn't just
one standard. Instead, it has several different conformance levels, predicated primarily on the PDF 1.3 specification. that fall under the umbrella
of International Standard IS0 15930 as follows:
b

PDFIX-la:2001.Thisversion.defined in IS0 15930-1:2001, isdesigned

for "blind transfer" of print data in press-ready forn-the PDF itself
contains all the elements needed to print. Color must be CMYK (plus
optional spot) only. and all images and fonts must be embedded. It is
based on thePDF 1.4 spec, but explicitly prohibits the use of transparency. It's also a pure subset of PDFIX-3:2002.
PDFIX-2:2003. This version, defined in IS0 15930-2:2003. is hnctionally similar to PDFIX-3 but allows for externally referenced data, so it
isn't usable for blind transfers, and uses portions of PDF 1.4.

PDFIX-3:2002,Thisversion,
definedin IS0 15930-32002,allows deviceindependent color.Thisincludes IAB, as well as tagged RGB and tagged
CMYK data. Untagged device dependent data is also allowed. As with
PDFIX-la:2001, it's designed for blind transfers, and hence all images
and fonts must be embedded. It's a subset of PDFIX-2.

Real World Color Management, a n d Edition

AU versions of PDFlX were updated in 2003. The current versions all

have 2003 in the name, but the changes are more clarifications. They're
based on PDF 1.4, but transparency is expressly prohibited for example.
We know of no third-party products supporting the 2003 versions. Effectively the versions listed earlier are currently supported.
We expect thatPDFIX-lawill become thedominant standard for blind
transfers in the United States since this market is primarily CMYK-only
(plus optional spot) and rarely works with device-independent data requiring additional handling at output time. Both PDFIX-la and PDFIX-3
will become the dominant standards for blind transfers in the European
market PDFIX-2canbeused anywhere an OPIworMowis currently being
used, and as an internal format inside large enterprises. Since it supports
usage of externallyreferenced data, and supports both device-dependent
and -independent color, it can reference images, fonts, or PDFIX-la and
PDFIX-3 documents.
Various third-party tools make creating PDFIX straightforward-we
discuss some of them in "Third-Party Stuff," later in this chapter. While
you can make PDFIX directly usingAcrobat 6 Professional, and lnDesign
CS,third-party tools are still very useful, and sometimes necessary.
For more information on PDF and PDFIX, check out these resources:

Oulpullnt8nl.All PDFlXdocuments must havean Outputlntent specified.

This can be an ICC-registeredprintcondition name or an 1CCprofile.That
means there's an implicit and an explicit possibility The implicit one is
really vague, in our opinion-check out the ICC registry at m c o l o r .
orglregistry2.html. and you'll seewhat we mean. So we recommend setting the OutputIntent with an actual ICC prolile, at least until the registry
matures into something useful. If you produce a PDFIX-2 or PDFIX-3
document that contains device-independent data, theoutputintent profile must be fully embedded.

Chaplet 16: Color Management and PDI

499

PDF/X-la. By far. PDFIX-la interests us the most since it's relatively

straightforward and has strong vendor suppon Objects must be devicedependent (CMYK plus spot1 color-1CC-Based color isn't allowed. However, an Outputlntent is requited. It's important lo realize that while you
can use an KC profile for the Outputlntent, theobjects in the document
are still device-dependent because PDFIX-la prohibits per-object ICC
profiles. The primary intent of the Outputlntent in PDFIX- l a is for verification purposes--it's there to ensure that the document has been color
managed to a specific destination in mind, and if that doesn't match
what's in the job order or ticket, someone needs to pick up the phone
and get clarification.
The Outputlntent is the profile that was used for color separation of
the PDF's content, so it becomes the source profile for all CMYK content
(which is everything except for spot color).
PDW-3. PDFIX-3 can contain device-dependent data (either untagged

RGB or untagged CkDIK, but not both in the same document), as well
as device-independent data (tagged RGB, tagged CMYK, and LAB). Each
device-independent object, by definition, has an ICC profile associated
with it, and the actual color space is referred to as "ICC-Based."
For PDFIX-3, the Outputlntent is the source profile for the untagged
device-dependent data, and is the intended destination profile for deviceindependent data. So the device-dependent data must be separated for
the intended output already, as it won't be converted again. This is the
only PDFIX variant that

Allows a destination profile to be included in the PDF for the explicit

purpose of performing conversions
b

Supports device-independent data

Supports RGB output devices

PDFIX-3 is a superset of PDFIX-la. which means that anything that

claims to be aPDF1X-3compliant reader must also correctly handle PDF/
X-la. Likewise PDFIX-2 is a superset of PDFIX-3. The numbers are out of
sequence, but that's the order the standards were agreed upon.
If your goal is to convert everything to CMYK in advance (an early
binding workflow),and submit a CMYK-only (plusop~ionalspol) file, you

430

Real World Cdor Management, Second Edlion

should make PDFIX-la. The notion that one would create CMYK-only
PDFIX-3 is silly because that's not what it was designed for, whereas it's
inherent in PDFIX-la. 'There is no advantage to a CMYK-only PDFIX-3
document, even though it can be done, and the disadvantage is that it's
potentially a source of confusion.
PDFIX-la is not "for the U.S." and PDFIX-3 is not "for Europe." That
the two markets havedifferent ideas about color management and printing implies that PDFIX-3 will be more popular in Europe, but there are
markets for which PDFIX-3 would be well suited in the U.S.This includes
large format priming, which has essentiallyno standard print condition at
all, and digital presses. The latter have excellent register but with generally
poor adherence to SWOP even though they like to claim so, and hence are
good candidates for implementinga PDFIX-3 workflow.
In subsequent sections of this chapter, you'll find that specifying perobject ICC profilesisn'tforthe faint of heart, and requiresspecial workflow
and software considerations. Once you have a PDFIX-3document, printing or proofing it requires both faith and courage.You'll need to use products that specificallysupport PDFIX-3 documentsto effectively proof and
output them.

Adobe Distiller isn't the only product capable of convertingPostScriptinto

PDF, but it's one of the most common. and it serves well as an example
application-just look for similar features and settings in the application
you use.

Compatibility
We previously mentioned the various versions of PDE Distiller's lob Options dialog box is one of the places where they're directly relevant to
color management. The Compatibility pop-up menu in the General tab
of Job Options affects the options available in the Color tab. If you select
"Acrobat 3.0 (FDF 1.2)," your only colormanagement policy optionsare to
leave colors unchanged or to convert them. Since PDF 1.2 doesn't support
ICC profiles, there's no way to simply tag the documents. Unlessyou have
good reasons to do otherwise. we recommend choosing'Acrobat 4.0 (PDF
1.3),"because that's what PDF/X is based on (see Figure 16-1).

Chapter 16: Color Management and PDF

4N

Acrobat Distiller 5.0

Compatibility settings

Color Settings
The Adobe Color Settings for Distiller, which you'll find under the Color
tab ofJob Options, look like those for other Adobe applicationsbut the
similarity is misleading,because their behavior is unlike those other Adobe applications (see Figure 16-21,
Settings Rle. DistiUer uses .csf fdes created in other Adobe applications,
and displays them in the Settings File pop-up menu, but it can't create
them itself. When you choose a setting from the Settings File menu, the
Color Management Policies menu is grayed out-the only way to enable
Color Management Policies is to select the settings file called "None."
Possibly the best way to handle this is simply not toworry about creating
settings files for Distiller, since Distiller has such unusual color behavior
anyway. If you save a .joboptionsfile, any custom color settings are saved
as well.
Color Management Policlcr.The color management policies in Distiller are
quite different than in any other Adobe application, by necessity, since
Distiller's job is to convert PostScript into PDF. Other Adobe applications

432

Real World Color Management, Second Edition

iiii
-

Figure 16-2

Acrobat Distiller 5.0

Color Settings

open and display files, then let you edit them in some fashion. Distiller is
just a completely different beast, and has policies to match.
Leave Color Unchanged-this means preserve the numeric values in
the PostScript file when converting it into a PDF. Ignore any CSAs in
the PostScript file, and don't embed the Working Space profiles.
b

Tag Everything for Color Mgmt (no conversion)-alsomeans preserve

the numeric values in the Postscript file when converting it into a
PDE CSAs associated with objects are converted into per-object ICC
profiles. The selectedworkingSpace profiles are embedded, and apply
to all other objects. All objects are considered tagged.
Tag Only Images for Color Mgmt (no conversion)-means the same as
tagging everything, except that only images are tagged. So vector objects, including text, are untagged, and considered device-dependent.

Convert All Colors to sRGB-any CSAs associated with objects are

used as source profiles. Objects with no CSAs take the Working Space
profiles as assumed sources.The destination profile is sRGB,so everything is converted to sRGB.

Chaptw16: Color Managanml and PDF

433

A brief note on the Intent pop-up. Rendering intents are chosen at the
time of PDF creation, just as in PostScript, which seems backwardswe
generally find that we can't ten which rendering intent will work best for
a given object until we know the output destination. But that's the way it
currently works.
The Default option is unclear-it's either the default PDF rendering
intent (relative colorimetric), or it's the intent. set in the PostScript file.
that points to the CRD that would have been used had the PostScript
file been interpreted bya PostScript RIP rather than by Distiller.We suggestsettingit to relative colorimetric-then you at least know what you're
dealing with.

Worklng Spaccr.TheWorkingSpace profiles are theones that are used for

tagging, depending on the color management policy selected.
For example, ifyour document contains amix of RGB, CMYKand gray
scale images, and you print a PostScript file, then distill with one of the
"Tag for Color Management" color management policies. a single copy of
eachworkingspace is embedded in the PDF document.The RGBWorking
Space profile becomes the source profile for all RGB images, the CMM(
Working Space profile becomes the source profile for all CMYK images.
and the Gray Working Space profile becomes the source profile for all
gra)scale images. Whether these embedded profiles apply to non-image
objects depends on which color management policy you selected.
The key thing to note is that "Working Spaces" in Distillerreally means
the profiles that are embedded into the document. They start out as assumed profiles and become embedded profiles in the PDF.
Any objects in the PostScript stream that have CSAs associated with
them don't use theworking Space profiles, but have their CSAs converted
into ICC profiles which are embedded into the PDF along with theworking Space profiles. CSAs are the only mechanism that allows Distiller to
specify profiles on a per-object basis. Distiller also ignores embedded
ICC profiles in the PostScript print stream (which usually appear only
when you print a document containing an EPS file with an embedded
ICC profile).
Dewice-Dependent Data. As the name implies, these senings only affect
device-dependent data, and prior to PDF generation typically only apply
to EPS files, since the options are all based on PostScript functions. With

434

Real World Color Management, Second Ediffon

the possible exception of overprint settings, ICC-based objects aren't affected by these options.
It's generally a good idea to preserve overprint settings. If they exist in
the Postscript file, they have to be preserved in thePDFfileifyou wantthe
document to print the way the originating application intended.
"Preserve Under Color Removal and Black Generation" exists mainly
for legacy support. If your images have already been separated to CMYK.
then placed into a page-layout application, and output to PostScript, the
images are "press-ready" so Postscript UCR or GCR functions shouldn't
exist to begin with, and presenting them is moot.
Remove transfer functions unless you really know what you're doing,
and have a specific reason for using them. Since a PDF could end up being printed anywhere. it's almost always best to remove them-it's usually
counterproductive to use them on other people's equipment.
Unless youexplicitly want an element to override RIP settings because
you placed an EPS with specifichalftone screens in your page-layout software, you usually want to remove halftone information too.

Making PDFs
Three main factors affect your ability to make decent PDFs whether you
intend to use them for the Web, soft proofing, or final output (plain PDF,
PDFIX-la, or PDFIX-3):
b

Proper preparation of all support files in heir respectiveapplications,

especially the final application you use lo print the PostScript file that
Distillerwill process.This includespage settings, font embedding, irnage preparation and linking, and so forth.

Useofan appropriate PPD file for your Postscript printer driver (which
includes making the application you're printing from aware of the
PPD--something that may or may not happen automatically).

Use of appropriate settings in Distiller, paying special attention ro the

color settings.

Chaptu 16: W o r Manrgmnmt and PDF

US

Output-only PDF--press. As previouslv mentioned, there are many re-

sources on how to make good PDFs, so we'll stay focused on the color
aspects. We've already discussed the first of the three factors mentioned
above in the earlier application chapters.
The second and third factors mentioned above are worktlow-specific.
but veryimportant.We'vecome to hust andrecommendasastartingpoint
the DDAP PPD, and their accompanying DistillerSettings Files. which are
available for both platforms at www.ddap.orgltoolsl~~nivppdl.
Note that the Distiller Settings File turns color management off, so
you'll end up with a device-dependent PDE This works as long as vou
produce output-ready Postscript prior to building the PDE The resulting
PDF won't have automatic soft-proofing or hard-proofing capabilities.
since there's no source profile embedded in the document. (If you refer
back to Figure 16-2, you'll see that it's based on the DDAPOptimizedvS
settings file for Distiller 5. Colors are left unchanged and untagged)
From here, you can feed the PDF to an application such as Apago PDFl
X-1 Checkup, or Enfocus Certified PDF (part of PitStop Pro and Server,
and Instant PDF), which preflights the PDE Fixes problems it can fix,and
makes a note of those it can't Assuming everything can be fixed,you end
up with a PDFIX-la document.
Color-managed PDFdaklop. This is a PDF that has profiles associated
for allobjects.\Vhileit's possibleto produce n PDF that has pmfilesassociated with some objects and not others, we recommend you avoid doing
so unless you enjoy lessons in frustration.
Using Distiller, we recommend producing an "output-ready" PDF in
your page layout application, as previously described in "Output-only
PDF-press," only embedding the final output device's profile. Since all
the color is already converted for one kind of output device, you can describe it all with a single device profile. Embed that profile, and Acrobat 5
and Acmbat Readersuse it automatically for on-screen preview, while other
applications that can read color-managed PDFs may be able to repurpose
the content for some other device (such as for proofing--see Figure 16-31.

436

Real World Color Managmnent, Second Edition

nguni6-3
m
w&
~i t -bnr .

ColorMana~ementPolicies-

'

Tag Everything for Color Mgmt (no conversion)

Intent: [Relative Colorimetric

DeYItF-DependentData -0 P r e l ~ r v overprint
e
SeRinpr

0 Preserveunder Color Removal andMerkOcneratlon

0 Prer ewe Halftone Information

Be careful: as configured in Figure 16-3, Distiller won't convert anything in the document, so you must make sure that all the color is already
converted for output before you feed it to Distiller. Distiller is just being
used to tag objects with a source profile (and of course make a PDF Ne).
Also, while you can effectively use such a PDF for soft and hard proofing,
it isn't necessarily repurposable--since all objects, including black-only
text, become device-independent, they won't be black-only if repurposed.
The real problem is that Distiller's options don't really offer enough
flexibility. If you choose "Tag Only Images," non-image objects will only
print correctly to the intended destination and will look wrong everywhere else, whereas if you choose "Tag Everything For Color Mgmt (no
conversion)" and print to adestinationother than theintended one, black
text gets converted to rich black, which can create major problems.
PDFIX-3 addresses the problem by letting you define hlack text as
device-dependent, so that it will print as black-only on any destination
device, while tagging all objects that need color management to produce
the correct appearance.

Chapter 16: Color Managmen1and PDF

437

The suggestions in the previous two sections generally provide the best results. But if you want to create the
smallest possible files, and have the color display reasonably well on the
random collection of uncalibrated monitors that senreas the delivery system for the Internet, you may want to try these settings instead.
In Job Options, Color tab, change the Settings File to None-the Web
Graphics Default color settings file is useless with Adobe Distiller. Next,set
the Color Management Policies to "Convert All Colors to sRGB." Finally,
set whatever output profile you used to create the PostScript as the working space-if your document is entirely Adobe RGB (19981, for example,
set that as the RGB Working Space so that Distiller uses it as the source
profile when it converts everything to sRGB. Likewise if your document is
CMYK, set the working space to the CMYK profile you used as the destination when you created the Postscript file (see Figure 16-41,

Color-managed PDF-Internet.

Figure 164

J0DW3E%i%tr7eiqfiRX$

~
p
p
p

Adobe Distiller settings:

Convert to sRGB PDF
Settlngr nle: [None
CDlorManagementPollrler

[ C O ~ W ~ A I co~ors
I
SRW

Intent: Default

Devlre-~ependentData - Preserve Overprint Senlnllr

Reserve Under Color Removal rind BlsckOenerstlon

Reserve

Transfer Funrnonr

Preserve Halftone Informanon

--

~.

Real World Color Managemenl, Sccond Edilfon

Acrobat 5
Compared to Adobe's other applications, Acrobat 5's color management
capabilities are just as funky as Distiller's.While the Color Management
Preferences dialog box looks familiar, its scope is limited compared to the
other applications, and printing from Acrobat makes us wonder what in
the world Adobe was thinking.
The redeeming feature of Acrobat is that color management really just
happens without any configuration.Acrobat grabs the display profile automaticallyfromthe operating system, anduses embedded profiles automaticallyand seamlessly, regardless of the color management preferences.
Setting up remote soft proofing is as simple as calibrating and profiling a
monitor and installingAcrobat on the client side. (On the "server" side, of
course, someone needs to embed the correct profiles into the PDFs.)

Color Management Preferences

To open Acrobat's Color Management Preferences click on "Color Management" in the list on the left of the General Preferences dialog box (see
Figure 16-51, Like Distiller, Acrobat can use Color Settings files created
in other Adobe applications as well, but it can't create them itself. Since
Acrobat always uses either anembedded or assumed source profile, color
management is always on, even though there's a "Color Management O f f
settings file.

Acrobat Color
Management Preferences

Chrptr 16: Color Management and PDF

439

The Working Spaces are the assumed profilesthey apply to all untagged objects in PDF files. If you set the RGB Working Space to monitor
RGB (yourdisplay profile),the numbers in untagged RGB images get sent
straight to the display, so you're effectively getting no color management,
but everything else gets color managed and there isn't anything you can
do about it.
Embedded profilesare always honored, and thus override the Working
Spaces.Therefore, becautious about what profile you select for theworking Spaces, because they affect untagged objects, which in the PDF world
are fairly common. If you set your RGR Working Space to ProPhoto RGB,
for example, most RGB content in PDFs will he superbly oversaturated
compared to any intent the original producer had for the PDE
The Conversion Options only affect conversions for Proof Colors (described later) and for printing.

Printing
Source profiles are either embedded in the PDF document, or assumed
by Acrobat's Color Management Preferences. but you need to specify a
destination profile. Yet for non-Postscript printers, Adobe has provided
absolutely no mechanism for speciFyinga destination profile or a rendering intent. When you print to a non-Postscript printer, it converts from
the source profiles to the display profile.
So when you print to raster printers from Acrobat,you're best off selecting Colorsync or ICM in the printer driver. We've confirmed that Acrobat
handsoff the display profile to the operating system as the source profile
for the data stream at print time. From there, the destination profile is
dictated by the printer driver.
When you print to a PostScript printer, the Adobe section of the driver contains an Advanced button that doesn't appear in non-PostScript
printer drivers. Clicking it opens the h i n t Settings dialog box, with no
fewer than seven color-management-related options. not one of which
is a rendering intent setting-see Figure 16-6!
We generally ignore the first three checkboxes regarding halftones,
transfer functions and UCRIGCR. If you need to use these feamres, you
know it-ifyou aren't sure, only experiment when you have lots of spare
time. The same goes for 'Apply Overprint Preview" which is useful for
forcing Acrobat to render the overprint function if your prinrer doesn't
supporl it.

440

Real World Color Management, Second Witlon

ngure 16.6

Acrobat Print Settings

(PostScriptprinter only)

"Print ICC colors as Device colors" produces identical results whether

checked or unchecked, so it seems to do nothing. (We're really unclear on
what it would do if it didwork-if you don't want conversions, just select
Same As Source as you would in other Adobe applications, instead of a
destination profile.)
'IApply Working Color Spaces," fortunately, does work, but we fail to
understandwhy it's optional.Whenit's checked,Acrobat uses theworking
Spaces in Color Management Preferences as source profiles for untagged
objects in the conversion to print space. When it's unchecked, Acrobat
uses a hardwired conversion that produces results charitably described
as lurid and weird.
Last but not least is the Color Profile pop-up menu. This works like the
Print Space pop-up menu in other Adobe applications, but it lacks renderingintent control-maybe someone from the Photoshop team should
walkacross the hall and have achat with the color people on thePDF team
to try to standardize user interface features and terminology.

Chapter 16: Color Management and PDF

441

Distiller 6
Very little has changedin the way of colormanagement in Distiller 6 compared to version 5.0. Therefore we recommend startingwith that section,
and then reading this one, which discusses additional features found only
in Distiller 6.0
The major change in Distiller 6.0 is that it can create PDFIX-la and
PDFIX-3 documents directly rom Postscript. See the new presets in Figure 16-7. PDFIX-la will likely be the more viable in most workflows because of limitations with Distiller 6's PDFIX-3 generation. If you need to
produce PDFIX-3 documents, checkout the section on InDesign CS later
in this chapter.
ngurt 16-7

Acrobat Distiller 6 presets

When set to produce either of these PDFIXflavors, Distiller 6 internally

normalizes a Postscript file to PDF, then it essentially preflights it for PDFl
X-la or PDFIX-3 compliance, and if it passes a PDF is saved. The distinction is important because neither Distiller nor Acrobat has the ability to
fix minor problems that might prevent a file from being valid PDFIX-la
or PDFIX-3. For that you'll need one of the products mentioned later in
this chapter.

442

Real World Color Management, Second Edition

PDFDC-la, PDFDC-3
If you intend to make PDFIX-la from Distiller, hopefully the "PDFXla"
preset is already sufficiently configured for your needs. But chances are
many of you will need custom settings, if for no other reason than to set
an Outputhtent profile other than "U.S. Web Coated (SWOP) v2" which
is the default.
Thus, the main new area of interest in Distiller 6 is found in the PDFlX
section of the Adobe PDF Settings dialog, located from the Settings dropdown menu (see Figure 16-81,A major oversight in the user interface is the
fact you can apparently select both PDFIX-la and PDFIX-3 at the same
time. Don't try it, it doesn't work, and will likely cause aberrant behavior in
Distiller thatwill make you homicidal or suicidal, or both, when you attempt
to troubleshoot the ensuing problems. Please select one or the other.
ngure 16.8
Adobe PDF
Settings dialog

The remaining area of interest, From a color management perspective,

is the selection of the OutputIntent Profile Name. Make sure you have an
explicit understanding with the receiver of the file as to what the OutputIntent should be for the job. If they accept PDFIX-la, they should know
what this request is about, and if they don't it should raise a red flag until
you have the contract proof in your hot little hands.

Chrptw 16: Color ilbnagemcnland PDF

b43

The RegishyName(URL)field defaults to the ICCWeb site, since with

device-dependent PDFIX including PDFIX-la, it's not necessary to embed a full ICC profile as the Outputlntent. Instead you can reference the
colorimetric behavior. If the Outputlntent you select is not one of these
registered behaviors, you should either provide a valid URL or leave it
blank. Ifyou fail to do this, your job won't be hosed, but technically itwon't
be a valid PDFIX-la file.
It's technically possible to create PDFIX-3 using Distiller, but we don't
recommend it, for several reasons.
b

If thePostScript filecontainsonly CMYK, you'realmost certainly better

offcreatingPDFIX-lainstead.If it containsCMYKandLab. Distillercan
and will create a PDFIX-3.

If the PostScriptfile contains RGB, Distiller's PDFIX-3 preset expects all

RGB objects to be tagged withCSAs,and if they aren't, Distiller will fail to
produce a PDFIX-3. Very few applications insert CSAs in the PostScript
stream. Theonethat doesso reliably, InDesignCS, has a PDFIX-3 export
feature that we recommend using instead of Distiller.
b

Instead of relying on the existence of CSAs, you can change the Color
Management policy in Distiller to Tag Everything for Color Management. If you tag only images, any RGB objects that remain untagged
will then cause PDFIX-3 creation ro fail, so that's a limitation. A bigger
one is that everything that's CMYK in the document will also be tagged,
includingimages you may not want repurposed later. But if your document contains only RGBand LAB content, this wuld beaviable means
of creating PDFIX-3.

You can set theCMYKWororkingSpace profile to match theOutputIntent profile vou select in the PDFlX portion of the dialog. That way,
CMYK objects undergo a null transform at output time, and RGB and
LAB objects are dutifully converted to CIMYK. But that means you're
stuck with that OutputIntent in order to ensure CMYK objects null
transform. thereby negating the whole point of PDFIX-3, which is its
portability to unknown destinations.

The bottom tine is that Distiller's PDFIX-3 support seems fairly duhious for any serious usage of PDFIX-3. You'll almost certainly prefer using lnDesign CS. which exports PDFIX-3 documents directly awhole lot
more easilv, and with far fewer limitations.

444

Real World Color Managcrmnl, kcond Edition

Acrobat 6
There are a number of subtle but useful added features in Acrobat 6, but
otherwise it behaves very much like Acrobat 5, so we recommend reading
both sections to get a complete picture.

Color Management Preferences

The preferences dialog, shown in Figure 16-9, has one new and important addition-the "OutputIntent overrides working spaces" is checked
by default. It makes Acrobat use the OutputIntent profile as the assumed
source profile for untagged content, for viewing purposes only. It's the
source profile for viewing all content in PDFIX-la documents, and for
viewing untagged objects only in PDFIX-3 documents. It is not used as a
source profile when printing, so it is also non-binding.

Acmbat 6

Preferences

The one situation in which you might want to uncheck this option is
when you receive a PDF in which the specified OutputIntent doesn't describe the actual output device. You can uncheck the option, and then
select the correct profile as the CMYKWorkingSpace. When you click OK,
Acrobat updates the display ofthe PDF to show you howitwill print to this
different destination. It's very important to note that this does not remove
or change the OutputIntent in the PDE it simply ignores it in favor of the
selected CMYKWorking Space.

Chaptcr 16: C o b Managnmt and PDF

445

Proof Setup and Proof Colors

This feature works as in other Adobe applications (see Figure 16-10].For
moreinformationseechapter 12, TheAdobe Common ColorArchitecture.
Unfortunately it neither shows nor automatically selects the Outputlntent profile when you open aPDFIX-la or PDFIX-3document. Unless you
know what the Outputlntent profile is, and have it separately installed.
you cannot fully soft proof. See the discussion on Acrobat Preflight for a
workaround.
Overprint Preview simulates overprintson-screen, which is important if
your document calls for overprints because they'll look very different onscreen without this option selected compared to how they would be print-

ed on a press. Another useful feature is Separation Preview, which shows

CMYK+ spot numeric values, with checkboxesto enable or disableviewing
each channel. Maybe the next version will also show RGB values.

Preflight
Acrobat G mcorporates a Preflight tool, which includes a standard complement of configuration sets, to which you can also add your own as you
see fit. Preflight is accessed from the Document drop-down menu. See
Figure 16-11.The top three useful color management related features here
have to do with verification of PDFIX-la and PDFIX-3, report generation
which shows source profiles and Outputlntent, and a feature to extract
the Outputlntent profile.
A n a b a n d ReporLThe easiest way toverify a PDF for PDFIX-la or PDFI

X-3 compliance is to select the conformancelevel you want from the Preflight list, and click Analyze. Preflight won't repair problems in PDFs but
it will generate both passlfail and a detailed report on any problems it
finds. See Figure 16-12.

mgure 16-n
Acrobat 6
~ t p r o 6 l ~

~cumentfk&?iiPDF/X-la
compliance

Document passed PDFIX-3 compliance,

also shows OutputIntent info and ICC
profile embedded.

Chapter 16: Color Management and PDF

447

You'll note two different items for PDFIX-3, one which implies only
CMYK+spot, and the other that ICCILAB are acceptable. By definition
PDFIX-3 allows for CMYK, spot, ICC-Based, and LAB colors. We recommend using the "ICCILab O K preset for PDFIX-3 verification. Again, if
you want CMYK-onlyplus optional spot color, use PDFIX-la.
If you have a PDF file that is not PDFIX-la or PDFIX-3 but should be
capableofcomplyingwith one of those standards, click the PDFIX. ..button, which reveals a new dialog. Note the specification pop-up menu at
the top. The "Save as PDFIX-la" or "Save as PDFIX-3" option will first
verify, and then if the PDF passes it will convert the file to the selected
conformance level. See Figure 16-13.
Figure 16.13
Acrobat 6 Preflight:
PDFIX dialog

V+,...+I
hl,Ughl m r AK* n d k k
*hchernnot if IS r vskd POFIX- 3 hlc

"ammw POFIL." nmovar d AX/X .p&s

n h t o n , for both PDFIX-3 uld WFIX-I a hlu.

Vmh...

Lsnm w

/ x...

Extract 0utputlntcnt.The thirdoptionin the Preflight.PDF1Xdialogis"Extract KC profile..."This option will extract the OutputIntent profile from
aPDF1X-laor PDFIX-3 document if it was fully embedded in the PDF.You
can then install the profile in the correct location, and now you can select
it in Proof Setup: Custom for both soft and hard proofing purposes.

448

Real World Color Management, Second Witlon

Printing
There are ~ w obig changes in Acrobat 6h printing abilities. First, you can
now color manage prmts using ACE to non-Postscript printers. Second.
you can print proofs.
From the print dialog, find the Advanced but~onand clickit to reveal
the Advanced Print Setup dialog (see Figure 16-14). In the Advanced Print
Setup dialog, ensure that you select Output in the list on the far left.
Printer Protile. This functions like the Print Space pop-up menu in other
Adobe applications. Select a prolle for the printer you will currently be
printingto. If you don't want any color management to occur at print time,
select "Same As Source (No Color Management)".
Applv Worklng Color Spacer. Although it's as mysterious in Acrobat 6 as
in Acrobat 5. you probably want this selected so that Acrobat will use the
Working Spaces as source profiles for any untagged content in your PDFs.
If you find a reason to not check this option, we'd like to hear about it.
Apply Proof kltlngs. Checldng this option references back to the profile
and other settings selected in Proof Setup: Custom, and lets you do cross
rendering. Unfortunately that profile is not listed in the print dialog, as
in other Adobe applications.The rendering intent options are affected by
the simulation checkboxes in Proof Setup: Custom.
Relative Colorimetric, with Black Point Compensation-is the default
with neither Paper White nor Ink Black che&ed.
Relative Colorimetric, no Black Point Compensation-Ink Black is
checked.
Absolute Colorimetric-Paper White is checked (and Ink Black is
checked and grayed out).

InDesign CS Export PDF

InDesign CS canexport PDFs directly, without theneed forDistiller.When
it comes to color management. inDesign's Export PDF capability is pretty
good, with a couple of potential pitfalls. It produces fully color-managed

w-u
Acmbal6 Prinl dialog
box, and Advanced h i n t
Setup dialog box

PDFs, with or without embedded profilesfor every object,or mixed-mode

PDFs. It can also create PDFIX-la and PDFIX-3 documents directly, although there are limitations with its ability to create PDFIX-3,mentioned
later in this section.

450

Real World Color Management, Second Edltlon

PDF Export Presets

The PDF Export Presets dialog, which you open by choosing File>PDF
Export Presets>Define.lets you create and save presets for making PDFs.
and is a real time-saver (see Figure 16-15). we suggest you familiarize
yourself with the various PDF settings for two reasons-familiarity with
them seems to make all the difference in the world between exporting
problematic and problem-free PDFs, and we don't discuss most of the
settings because they don't pertain to color management.
Figure16-15

InDesign CS PDF Export

Presets dialog box

Scrwn and eBook.These presets set the Destination Profile to Document

RGB on some machines, and Monitor RGB on others.We don't know why
but the badnews is that ineither case"Include1CCProfiles" is not selected
so you're guaranteed to see the color differently on every machine that
views the resulting PDE Create your own presetwith arationaldestination
profile (sRGB?)and include the K C profile.
PDF/X-la. This preset works as advertised. Any RGB or LAB content is
converted to CMYK using the selected Destination Profile, which is then
embedded as the OutputIntent profile.
PDFIX-3. This preset works with the following limitations: your destina-

tion is CMYK, and you don't want any tagged CMYK objects. InDesign
produces compliant PDFIX-3 with a CMYK OutputIntent only, and all
CMYK objects in the layout are untagged IDeviceCMYK in the resulting
PDE If the document contains placed tagged CMYK images, the tags are
ignored and not embedded when making PDFIX-3.

Chapter 16: Color Management and PDF

451

Due to limitations intheExport PDF interface, werecommend youlive

within these limitations. If you have more sophisticated requirements,
use a third-party application such as PitStop Pro to re-tag CMYK objects
you want tagged, or to create PDFIX-3 for RGB destinations, and be very
brave and even more cautious.

Export PDF
The Export PDF dialog is accessed one of two ways: by choosing
FilesExporl and selectingAdobePDF as the format, or by choosinga preset from File>PDFExport Presets. We prefer the latter. In the left column
is an Advanced option, which is where you'll find the Color and PDFlX
sections (see Figure 16-16).

InDesign CS
Export PDF dialog box

Color. Color offers three options. Leave Colors Unchanged does exactly

that-if you create a document with mixed-mode images and elements,

you'll create a mixed-mode PDE Selecting either RGB or CMYK enables
the Destination Profile pop-up menu, discussed next, and converts all the
content in your document to the destination profile.

459

R d World Color Management, Second Edilion

Destination M e . Here you select a desrination profile for output. If the
destination is apress, make sure you select Document CMYKas thedestination. which is the default. Ifyou don't, InDesignwill converteverytliing,
tagged anduntagged, because it sees the CMYK source profile, Document
CMYK, as being different than the Destination Profile.
Include ICC Pmflla. This option embeds ICC profiles into the PDF. If you
selected Leave Colors Unchanged in the Color pop-upmenu, each profile
used in the document is embedded in the PDE If you choose to convert
thedocument to RGB or CMYK, only the destination profile is embedded.
since all document content is converted to that destination.
Simulate Overprinl.This option does two things. It makes InDesign do the
overprint itself before producingthePDE which isuseful for makingdualpurpose PDFs for screen and print because many consumerdevices--and
even somePostScript RIPS-haveno idea what overprinting is. But it also
convertsspot colors to ptocess, so it's not anoption you want to use when
preparing PDFs for p r e s s u s e the Ink Manager instead.
Ink Manager. The Ink Manager can perform a variety of functions, which
are described in the lnDesign User Guide. If you make changes here, they
affect the exported PDE not your InDesign document. Be careful with
these settings if you rely on InDesign to perform trapping, because some
of them are used by InDesign's trapping engine.
If you select the option to convert all spot colors to process, they're
converted using the CMYKequivalentssuppliedby Pantone in whichever
Pantonelibrary you used to specify thespots.They aren't converted using
KC-based color management, so your process Pantone colors probably
won't look right unless you're really lucky.
PDFIX Output Intent Proflie. Ensure you select a profile for the intended
output conditionswith this pop-up menu. W~ththe PDFIX-lapreset, this
will snap to match the Destination Profile, and viceversa. Unfortunately,
even though PDFIX-3 allows RGB Outputlntents. InDesign CS does not.

Chaplsr 16: Color Harugemmiand PDF

453

EPS, PDF, and Al Formats

InDesign CS will color-manage EPSs containingeither PostScript CSAs or
ICC profiles. It uses either for on-screen preview, as well as for outputting
Postscript andexponingPDF. InDesign assume Document profiles as the
source profile for untagged EPS, which means if the Destination Profile
is something other than Document CMYK,untagged CMYK EPS will be
convened.
Placing PDF or A1 formatted files into an InDesign document follows
similar rules. Keep in mind PDF is a container format, so it can contain
multiple objects with different source profiles, as well asuntagged objects.
The untagged objects inherit the Document profiles. The only ways to
prevent objects from being converted during PDF export are
Select Leave Colon Unchanged.
Make sure that the Document Profile (the assumed source for untagged objects) and the Destination Profile you select in Export PDI:
are thesame.
Because of the complexities of these formats, and the simplicity of the
InDesign Image Color Settings interface, it is not possible to change the
source profiles, either assumed or embedded. or per-object rendering intentsfrom within InDesign. Therefore, Image Color Settings is grayed out
when placing EPS, PDE and A1 (Adobe Illustrator1 documents.

Third-Party- Stuff
'I here are a number of really good products on the market. but the one
we've worked with the most, and hence can discuss with some authority,
is Enfocus PitStop Pro 5 for Acrobat 5, and version 6 for Acrobat 6.

Enfocus PitStop Pro

This plug-in fort\crobat lets you do all kinds of manipulation and editing
on PDFs, but we'll only look at the ones that are color-management related. Consider ir oneof the necessarytools in the arsenal ifyou're a heaw
PDF user. (Or for high-volume workflows, consider theindustrial-strength
PitStop Server-for more information, go to www.enfocus.com.)

454

Real World Color Management, Second Edition

PltStop Pro strengths. PitStop Pm offers the following useful capabilities
for creating and massaging colormanaged PDFs:
distinguishes between device-dependent (RGB. CMYK. Gray) and
LAB and ICC-Based color spaces on a per-object basis.

b It

It lists the embedded profile associated with each object.

It lets you untag objects, reassign profiles to objects, and embed

profiles on saving the PDE
b

It lets you specify a destination profile to which any object can be

converted. It supports RGB-to-CMYK. CMYK-lo-RGB, and RGB- or
CMYK-to-grayscale conversions using ICC profiles.

It lets you do any of the above object by object, or globally.

r It pteflights and produces certified PDFIX-la, and PDPlX-3 docu-

ments with a preflight report, and tracks changes to the PDF as it

moves through production.
b

It 0!3ersdownloadable"PDFProfiles" that extend its ability to preflight

PDFs for othertarget viewers. such as the Internet.

PMlop Ro wcakncssct. However, it isn't perfect, and has the following

weaknesses:
b

It offers no rendering intent control for conversions. which use the

destination profile's default rendering intent.

It doesn't directly support RGB-to-RGB or CMYK-to-CMYK conversions, though you can work around the limitation by first converting
an object into LAB as an intermediary space.

It doesnStsupport D e v i c e L i profiles.

Chapter 16: Color Msnagcmcnt and PDF

455

Other Products
We haven't worked with these products as much as we have with PitStop
Pro, but they come recommended bypeople whose judgment we trust:
b

Apago PDFIXCheckup, www.apago.com

Callas PDFlX Inspektor [free),and PDP Inspektor, www.callas.de1

DalimTWiST and SWNG,www.dalirn.com

Onevision (various products), wwrv.onevision.coin

The Wave of the Future

If. as we suspect, print publishing will be forced by brute economics to
make the transition from arcane craft to modern manufacturing process,
it seems likely that PDF worktlows, whether based on the current PDFlX
standards or on ones that don't yet exist, will become prevalent. But while
unambiguous communication of documents between creator and producerwillultimatelp benefit both sides, note that the shift from delivering
application files, fonts, and linked graphics to delivering a print-ready
PDFIX-la involves a shift of responsibility back to the content creator.
In some ways. PDFIX is a 21%-centuryequivalent of the Postscript
dumps Bruce used to send to senrice bureaus in the mid-1980s-when
things went wrong, the blame was always his. PDF workflows hold huge
promise, but they're still experimental, and you can always spot the pioneers by the arrows protruding from their backs--so if you're considering
PDF workflows, take it one stepat atime, test thoroughly, and don't attempt
untried procedures when you're facing critical deadlines.

This page lefi blank. ..

and Scripting
The Smart Way to Be lazy

Unfortunately, our desktop applications and our equipment-scanners,

RIPS, and printers--don't always give us the exact color-management
options we need. Fortunately, automation and scripting provide us not
only with solutions to these limitations, but also bring a lot of power and
flexibility to workflows that we may not otherwise have considered.
The purpose of this chapter isn't to cover every conceivable automation
or scripting option, or even to serveas a primer. Instead, it's meant to give
you a taste of the kinds of things automation and scripting can do, and
teach you the right questions to ask when you look for products to solve
your particular needs.

Why Automate?
We use automation for the following reasons:
b

To free humans from repetitive, brainless grunt work that doesn't require intelligent human decision making or intervention

To make efficient use of our available processing power

To workaround color-management limitations in our applications or

equipment

458

Resl World Color Manrgrmant, Second Edition

For example, we may have scanners or digital cameras that we've

profiled, but the software that drives them doesn't embed profiles. We
know which profile we need to assign to the images they produce, so
rather than doing so manually in Photoshop, we can use automation (or
scripting) to batch embed the appropriate profile in a single image, in
multiple images, or even in folders of images. Since we don't need to open
or display the images, the performance advantages can be significant.
especially with RAM-gobbling high-resolution images, and we free ourselves from mind-numbing, repetitive button-pushing.
A good example of using automation to work around an application's
limitations is using a color server to address desktop applications' nearuniversal lack of color-management support for vector EPS-many automation products can process EPS files,and sometimeseven embedded EPSs
or TIFFSwithin EPS files. On the device end, it's not uncommon for an
imagesetter, platesetter RIR or proofing device to have no idea what ICC
profiles are. Color servers placed in front of such hardware can perform
color management tasks before Postscript or PDF get to the RIP
Color management offers one other very specialized type of automation-so specialized. in fact, that many people don't even recognize it
as an automation feature. DeviceLink profiles offer an elegant way to
store specific, hardwired device-to-device conversionssee "DeviceLink
Profiles." later in this chapter.

Color Servers
Color servers are sometimes called batch processors, or generically automation products. Color servers placed at the front-end of another devicelRlP do their work by parsing Postscript or PDF files, changing only
the numeric value+they don't damage the integrity of the Postscript or
PDF l3es.When they process bitmap information (such as TIFE JPEG, or
bitmap EPS),they effectively open the file, perform theconversions, then
rewrite the file.
Common color server features include
b

Processingmultiple files or folders of files.

Chapter iT: Autonutkn and Snlptlng

459

Performing color conversions on TIFF and PEG. Some also support

EPS (both bitmap and vector).

Supporting embedded profiles, with a mechanism for assuming

profiles if no embedded profile is present.

Using a LAB-based look-up table for converting Pantone (sometimes

also Focoltone] colors to custom RGB or CMYK builds.

Batch embedding profiles, or batch removing embedded profiles.

Most support hot folders. Some support print queues that appear on
the network as though they were Postscript printers.

Color Server Pros

Like any other technology, color servers have their pros and conswe'll
look at these in a general way before getting into specific examples.
b

Fast processing-processingis much faster than doing an equivalent

procedure in anapplication suchas Photoshop, because the files don't
need to be opened or displayed.
Coniigurabiliy-you can set up color servers to do tasks ranging from
simply embedding profiles into images to simultaneously producing
multiple converted files for scatter proofs and monitor soft-proof versions. You can set it to color manage TIFE but not EPS. You can set it
to color manage RGB images, but not CMYK.

t Front end to almost any hardware--sticking a color server into a

workflow provides a single front end for colormanaging a variety of

devices. When you add a new proofer, you can simply add another
queue to support that device without having to worry if it supports
ICC profiles, and your future purchasing decisions don't depend on
built-in color management support.
b

Hot folders and print queues available over network-you can set
up hot folders to accommodate remote clients over a network and
publish print queues on thenetwork, too. Users don't have todealwith
configuration issues or decide which profiles to use for source and
destination-instead they simply place the tiles into the appropriate
hot folder. Faster, and less prone to error.

460

Real World Color Management, kcond Edition

t Named color system support-lets you deal effectivelywith named

color systems like Pantone and Focoltone by generatingcustom CMYK

equivalents for vour specific devices. Equally important, it uses the
some CMYK equivalents for all instances of a given Pantone or Focoltone color, effectively solving the seemingly unending problem of
applications that use libraries containing different CMYK equivalents
for the same solid colors.
Colorserversare very wellsuited to repurposingalreadyseparated jobs
(coated to uncoated, or between inksets) or to turning any device with a
large enough gamut into a proofer, even if it doesn't have explicit color
management support.

Color Server Cons

Lest the picture we've painted above sounds overly rosy, we feel bound
to point out the foUowing downsides to most sewer-based automation
products:
b

Costly--especiallyifyouereusedto the pricesfor desktop applications.

Prices range born just under $1,000 to about 53.000.

Limited rendering intent control-if you plan on using a server in

late-bind'ingworkflows-such as using an-RGB page layout, and printing RGB PostScript to a color server that converts to the appropriate
CMYK space--you generally can't specify per-object rendering intent
control. (Wedon't know of any page-layout applications that can generate a PostScript stream that contains rendering intents.) You can.
however. specify rendering intents for classes of objects, such as "images" and "vector graphics," to use something else.

t Limited source profile control-when a page-layout application gen-

erates a PostScript stream, it doesn't embed ICC profiles for formats

other than EPS, and it only does so for EPSs that already contain embedded profiles. If you use TIFF or PEG file formats with a variety
of source profiles, the color server can only assume source profiles.
Again, you can specify source profiles for different classes of objects,
but thatk it.

Chapter D: Autmtlon and kripting

461

Most of these cons have workarounds. The subtle differences between

relativecolorimetric and perceptual rendering may be within acceptable
tolerances for many workflows, so the limited choice of rendering intent
may only be a problem in workflows aimed at very high-quality output.
And if you can standardize on a single source space in the page layout,
the limited source profile control isn't a prohlem because the assumed
profile is always correct.
Late-binding workflows that use color servers for final conversion
simply don't provide per-image control over conversions--but even in
high-end workflows that demand per-image control, servers can still do
conversions to use devices that don't have explicit color-management
support as proofers, and to assist with solid-color matching.

Examples
This is by no means an exhaustive review of every server-based automation product available.We've chosen the followingexamples both because
we're familiar with them and because they provide good illustrations of
the capabilities of server-based automation in general.

Praxlsofl ICC Autoflow. ICC Autoflow, available for Macintosh and

Wndows, processes PostScript, EPS, and-on Windows only-PDE A
reduced-cost desktop version doesn't support hot folders-you point it
manually to a folder, then the files it contains are processed according
to theconfigured settings. Both versions let you create and save multiple
configurationsets (see Figure 17-1).Note that this product doesn't process
grayscale content41 simply passes through unchanged.
Asdiscussed inchapter 15. ColorManagemenr in QunrkXPress,VectorPro creates custom palettes that are also supported in ICCAutoflow.

GrctagMacbeth IQueue. iQueue comes in several flavors: 110.120 and

140. The primary differences are that 110 doesn't support Postscript (or
EPS), while the other two do. In addition to 120's features. 140 adds PDF
and print queuesupport instead of just hot folders (see Figure 17-21,

Main window

Color Queue Setlings

Color Sets

Chaptef 1Z Automation and krlptlng

Main window, Queue Setup window, and ICC Options

A few other features:
b

Explicit support for PDFIX-la and PDFIX-3 (140 version)

DeviceLinkand abstract profile support

Additional He formats: BMP,Scitex CT,and PNG

b A built-in mini-Web server that lets you obtain job status from any

workstation

Color Server Questions

If you're contemplating a server-based automation product, here are
three important questions you need to ask:

464

Real W d d Color Management, Second M H o n

b

Which Ble formats are supported? Most though not all worldlows require PostScript (and hence EPS) support. Increasing numbers want
PDF support.

What black-preservation options are offered? Does black-only text

remain black-only when converted? What about black-only drop
shadows? Most products have some means of ensuring that blackonly text is ignored, and thus remains device black instead of being
converted to a rich black.

What is my intended use for the color server? As we've previously

discussed, color servers are well suited to repurposing and proofing.
They can also make separations in high-volume workOows in which
the separation is subsequently checked by a skilled operator. The lack
of per-image rendering intent or source profile control may be acceptable in some workflows and not in others.

DeviceLink profiles are one of the seven profile classes defined by the
ICC specification. A DeviceLink doesn't describe a device--instead, it's a
concatenation of twoormore profiles, in asingleNe. It actsas both source
and destination. They're an automation feature because the alternatives
involve manually manipulatingfile content, which is why we cover them
in this chapter.

More about Links

Three things make them different from the device profiles we've discussed
throughout this book: They're an apparent exception to the rule that you
need at least two profiles to make a conversion happen. Even though a
DeviceLink is a single file, it always contains at least two device profiles.
The DeviceLink itself is not a device profile.They rarely contain a PCS:
therefore they cannot describe device behavior in relation to a deviceindependent color space.This also means you can't embed them into images.They don't describe asingle colorspace, but a hardwired conversion
from one set of devicevalues to another.

The source and destination tables both contain device values, so the
conversions are directly from RGB to C M K , or CMYK to CMYK, depending on the device profiles used to build the DeviceLink. (The PCS
is involved in the initial joining of the rwo profiles, but once that initial
concatenation occurs, the PCS has played its role.)
DeviceLinks are always unidirection-conversions are only from the
profile that was originally defined as the source to the profile that was originally defined as the destination at the time the device link was created.
Some caution is required when building workflows around products
and scripts that use DeviceLink profiles. Most implementations assume
that if you are sending documents to be processed by a DeviceLink that
youare doing so with foreknowledge that any embedded ICC profileswill
be ignored in favor ofthe DeviceLinki source profile. This implies content
is normalized to favor the DeviceLink's source profile, and if that's not the
case you will get less than ideal results.Thus. DeviceLinks are well suited
for proofing and repurposing scenarios, but will require more caution if
you use them for making separations. They're just a look-up table, afterall, but due to the lack of a PCS in the middle, their unique channel data
is not lost in the course of a PCS becoming involved in the conversion.
These advantages seeminglydefy traditional color management convention, and are at least as much about automation as color management, if
not more so about automating.

Why use DeviceLinks?

Normal profile-to-profile conversions sometimes do things we don't
want. For example, we usually want 100% black (CO MOYO K100) to stay
100% black, but CMYKconversions through the PCS always turn the four
original channels into three before turning them back into four, so there's
just no way they can preserve black-only text or drop shadows.They turn
into a rich black, with the attendant registration headaches, instead. A
DeviceLink lets you force the conversion to 100% black
Likewise, if you have a yellow background, or yellow text, normal device profile conversions will add scum dots as it attempts to compensate
for hue differences between source and destination. The most common
instanceof this iswhen proofing, but it could just as well be a job originally separated for a low-qualitystock and the job now needs to be printed
on a higher-quality stock as well. Traditional color managemen! will try

466

Real World Cdor Msnagmml, kcond WMon

to match the dingier grayyellowon the lower-qualitystock,and a DeviceLink can be set to preserve channel purity.
Traditional worMows will use simple curves to adjust for tone reproduction differences, but they cannot compensate for differences in ink
hues where you should compensate for them, such as in photographicimages, nor do curves compensate for differences in wet trap. A DeviceLink
can, while also preserving channel purity unlike output device profiles.

Making DeviceLinks
For generating DeviceLink profiles, our favorite tool is Link-o-lator 2.0, by
Left Dakota Inc. (www.leftdakota.com),becauseit offers conmls and options
we haven't found anywhere else. You can download a trial version, which
comes with documentationinPDFformat describingthenitty-gritcy of various features (see Figure 17-3).Two features are particularly valuable:

You can build in black point compensation, so you get a DeviceLink

profile that simulates a conversion using Adobe's black point compensation (see"Black Point Compensation" in Chapter 12, The Adobe
Common Color Architecture.) For non-Adobe products, it's the only
way to get conversions using black point compensation.

You can specify pure channel presenmtion with dot gnin wnzpensation,
which can beused to preserve black, or anyof the other channels in the
destination profile. Most products offering black channel preservation
do only that, preserving the black channel exactly so the pre-conversion and post-conversion black is identical. This is rarely useful, because the source and destination devices often have different black dot
gain. Black channel dot gain compensation preserves the blackchannel
purity while compensating for the dot gain difference, so you still get
a black-only drop shadow when you repurpose C,MYK images, but the
tonal characteristicsof the black channel are preserved.

Another application for making DeviceLink profiles is UltraLinks 1.0.

also from Left Dakota (We are convinced they are totally obsessed with
DeviceLinkprofiles.)This application is uniquely suited for building RGB
to CMYK DeviceLinks. You can ensure 255 R, 0 G,OB always converts to
specific ChfYK values that, for example, contain no cyan or black. Or 255
R, 255 G. DB always converts to 100% yellow!
There are controls to afFect the blending between this "color-by-thenumbers" approach, and the colorimetric result mandated by theoriginal
source and destination profiles. While we have not hlly put this product through the wringer, we think it's interesting and solves a number
of problems some people consistently complain about with respect to
conventional RGB to CMYK conversions.

How to Use These Dam Things?

None of the major color-managed applications discussed in this book
support DeviceLink profiles. While long in the tooth, Apple's free ColorSync Plug-ins for Photoshop provide support for DeviceLinks on import
and export, and but most of the time DeviceLink profiles belong in RIPS,
in color servers, andin prepress equipment that explicitlysupports them.
Somescriptingmethods,such as Applescript,discussed later in thischapter.
will also support DeviceLinks.

468

Real World Color Management, kcond Edition

Prepress Solutions
Prepress solutions are proprietary modules that fit into an existing prepress worldlow, such as Creo Prinergy or the Rampage System made by
Rampage, Inc.Their featuresvarywidely--some only offer color-management capability as an extra-cost add-on; others have v e y limited colormanagement functionality. While they vary in their capabilities, these
types of solutions generally share two common characteristics:
b

The worldlows into which they integrate are inherently late-binding.

Conversions happen at the prepress stage, immediately prior to printing, rather than at the design or customer approval stages. You can't
get much later than that.

The solutions are proprietary. In conmst to color servers. which are

designed to work with practically anything that has a profile, integrated prepress solutions work only in the specific preptess workflow
system for which they were designed.

Who Should Consider Prepress Solutions?

These products plug into complex and expensive prepress workflow systems that cost tens or even hundreds of thousands of dollars, and critiquing such systems is well bevond the scope of this book. If you're an
individual or a small shop, and such workflow systems are something
you neitherwant nor need, the only reason to read this section is to gain
some insight into the kinds of things that happen to your job once you
submit it.
This discussion deals solely with the pros and cons of the color-management aspectsof theproducts anditsonlypurposeis to help you decide
whether you'd be better served by a proprietary color-management module for your existing [or intended) workflow system or an open color-management solution.Thetrade-off is typically one between convenienceand
cost. A proprietary color-management module offers the convenience
of seamless integration into your workflow system and one-stop shop-

Chapkr 11: Automation and Scrlpling

469

ping for service and support, but you can expect to pay a hefty price for
something that more or less replicates the standard color-management
features found indesktop applications. That said, prepress solutions can
help in several situations.

Lftcctive lak-binding workflow solutlonr For workflows where the intended output is unknown during the design process, products in this
category offer per-object source profiles and per-object rendering intent
control, so that you can prepare the content for separation without actually knowing the destination. Once the destination becomes known, everything is ready to create final separations. In lower-volume workflows,
youcancolor manage the content in Quar!dPress at printtime, but that's
less practical in high-volume workflows that rely on high-end prepress
solutions.

Hard proollng. Some of the products in this category offer explicit inkjet
printer support. Enabling color management can turn an inkjet with a
sufficiently large gamut into a proofingdevice.

PosHIIght normalization. If preflight rejects a submitted job solely because it contains a few RGB images, the workflow system can assign appropriate source and destination profiles. and convert those silly RGB
images into CMYK. This is much more efficient than the traditional alternatives, which are either to reject the job, or to have someone open
the images in Photoshop, convert them. update the placed images in the
page-layout application, then output another Postscript file

....

Solid color subst~lutlon.Repress systems can often parse the PostScript

to find instances of solid-to-processPantone or Focoltone color conversions.
When it finds them, it can substitute your custom hand-tuned CMYK
equivalents, so you get the best possible simulation of solid colors, no
matter which application originally specified them.

570

Real World C o b Managemenl, kcond Edition

The Downsides
You need to judge the downsides to thesesolutions in thc larger context of
the entire workllow system, of which the color-management component
may be a relatively tiny part.

Price. None of the vendors we've spoken to are willing to quote exact
prices because "they vary" depending on the particularworktlow bundle.
That said, while pricingcovers afairly wide range, the approximate numout there" is$5,OOO, which sounds like a hefty
ber we keep seeingWthrown
chunk of change to replicate what boils down to the basic desktop colnrmanagement features.
With grnwingdemand, we predictthat color-management capabilities
will becomestandard in futureprepress solutions, but for now, they often
cost extra. The best piece of advice we can offer you is to be a s a y consumer-ask lots of questions, and see demos (includingones with typical.
and maybe a few atypical, jobs) to see just what they'll buy you.
Ralning.The challenge with closed ptoprietarysystems is finding people
who can train you how to use them and integrate them into your specific
workflow.Thefewpeople who are really knowledgeable about integrating
color management into proprietary workflow products tend to command
top dollar.

Examples
The two products presented here typiFy the potential differences in this
class of product and illustrate the need for careful research. Color-managernent support is available in Xinet. Rampage, and Apogee, among
others. But the term "color-management support" means quite ditrerent
things to different vendors. You need to dive in and find out what each
one means.
Helior lmrgekmr and PDF Handshake. These are separate products.
but are complementaryin many worktlows. Imageserverused to be called
Helios EtherShare OPI, and is an image handler more than a prepress
workflow solution. It combines OPI and color-management features, so it
can change image size, resolution, and compression, in addition to doing
color management. Unique features we like about this product include:

Chapm n:Automallon and Scripling

S7l

r Print queue support. In Helios Imageserver,configured print queues

appear as virtual printers, so the user simply prints to the appropriate print queue--such as "Coated StockW or, for a proofing queue.
"EpslOKMgl CS#2" (to denotean Epson 1OOOusingaMitsubishi glossy
coated paper, simulating the press using coated stock #2). Naming
systems are workflow-specific-the point is that you can name the
queue as whatever makes sense for your workflow and your people.
b

Remote proofing. You can upload PDFs to PDF Handshake, which

rasterizes and color manages it, then soft-proof the result in a Web
browser on a suitably calibrated and profiled display. Note, however,
that PDF Handshake honors embedded profiles in all content, both
bitmap and vector. For bitmaps. PDF Handshake lets you remove or
reassign profiles in bitmaps when you don't want to use the embedded
one, but for text andvector content, you'd need to use something like
PitStop Pro or Server.

(no Prinegy. Prinergy is a full-blown prepress workflow system that's

similar to Scitex Brisque, except that it's PDF-based, so everything that

goes into Prinergy is in PDF format. Prinergy offers two levels of colormanagement support:
Standard.The standard level sees and uses the embedded profiles for
each object in the PDF as source profiles, and converts them to one
hardwired SWOP-based destination profile. If you're satisfied with a
worMlow that converts from the embedded source profiles to a SWOP
destination, the standard level of color management is all you need.
b

Advanced.The advanced color-management license lets you override

or assign sourceprofiles for each objectin thePDE andselect acustom
profile for the destination. It supports not only CMYK, but up to eightchannel ICC output device profiles, so you can use it to make Hi-Fi
separations. It also supports DeviceLink profiles, black preservation.
and proofing.

Prinergy can perform color-management tasks (embed, change embedded, or convert) when the job is first brought into the workflow, or
at any time thereafter, but proofing conversions happen only when you

474

Rwl World Color M.nagrmsnt, kcond Edition

generate the output stream to the proofer, so i t doesn't create a separate proofing Me. You select the job, the proofing device, and the proofing
profile, then the job is convertedso that the pmofersimulates theselected
destination [the pressor acontract pmofingsystem). Ifyour proofingsystem does its owncolormanagement, and you preferit,youcan submit the
job for proofing as final press CMYKwithout specifyinga proofer profile,
so that you don't end up with double color management.

Important Questions
If you're contemplating introducing such a product into your workflow,
here are some of the questions we think are critical:
t Does the product support black preservation? It's almost always a

problem when 100%black-only text and drop shadows get converted

to rich black, or when your carefully crafted black channel vanishes
because it gets converted to LAB, then reseparated. This always happens when you use color management for repurposing (CMYK-toC W conversions.
b

Does the product support black scaling? A black channel for a lowquality uncoated stock won't give good results when the job is repurposed for a higher-quality coated stock if the black channel is simply
preserved. What's really needed is to preserve the tone reproduction
curve of the black channel. not the original numeric values. For this,
you need a scaling function for the black channel to compensate for
the differences in dot gain between the source and destination devices. If the product you're loolang at doesn't have this feature, and
you think you need it, see if it supports DeviceLink profiles instead
(see the section, "DeviceLink Profiles," earlier in this chapter).

Can automatic RGB-to-CMYK conversions give me the quality1 need?

This is something you can test without buying or upgrading existing
equipment. Profile conversionsare made with no knowledge of the image content-they'reone-size-fits-all,treatingapictureofa black cat in
a coal cellar exactly the same way as apicture of apolar bear in the snow.
With good profiles, you may he able to create acceptable separations
without any intelligent human intervention,butpremium-quality work

Chapter Il:Automation and Scripting

473

usually needsimage optimization either before or after separation.You

need to look at the work you do, and test the quality of your automatic
profile-to-profileconversions, before you can answer this question.
t Are DeviceLinks supported? Many of the above issues can not only

be resolved, but resolved more easily and with greater flexibilityifthe

product supports DeviceLinkprofiles,and you use one oftheproducts
mentioned in the DeviceLinks section, earlier in this chapter.

In-RIP Color Management

Ln-RI I' color management is most commonly used for proofing, because
the total lack of control inherent in in-RIP separations usually doesn't
work in other workflows It occurs in three different ways:
Front-end. The conversions are done before the PostScript is interpreted. This is like bavinganintegrated color server in your RIP product, and is most common in proofing products such as the BestColor
RE! You can treatbitmap andvector objects separately, and automatically replace solid Pantone colors with process CMYK equivalents.
r Back-end. The PostScript is interpreted, creating a rasterized page

image which is then color-managed and sent to the output device.

Since there's only one raster object, there can be only one source and
one destination. This approach is usually adequate for basic proofing
n e e d s a n example is the noworphaned Adobe PressReady.
t In-RIP. This is literally in-RIP conversion--color management is per-

formed bythePostScript interpreteraspart ofthe PostScript interpretation

process. The two flavors are PostScript color management and ICCbased in-RIP color management. In theory, any PostScript level 2 RIP
at version 2 0 1 7 m or higher supports PostScript color management.
but in practice, workable PostScript color management is rarer than
hen's teeth. Global Graphics' Harlequin RIP has an ICC option that
literally builds an KC-based color-management engine into the RIP:
so it uses regular KC profiles to perform the conversions.

474

Real Wotld Color Managemnl, Second Edlllon

The Good
We're generally not fond of solutions that rely on sending our jobs into a
mysterious black box, but proofing is a simple enough and unambiguous
enough conversion-hm output CMYKto proofer CMYKor RGB via absolute colorimetricrendering-that it's an exception to our general rule. The
main advantage it confers is simplicity-you simply send the final output
fdes to the proofer, and the RIP takes care of the necessary conversion.

The Bad (and The Ugly)

In-RIP separations do invollre sending your data into a mysterious black
box, and hopingthatthe mysterious blackboxdoes theright thing. In-RIP
separations tend to rely on assumed source profiles and rendering intentsthere's no consistent or reliable mechanism forspecifyingcustom,
per-object source profiles and rendering intents.
InDesign 2 does provide a consistent mechanism for creating PostScript source profiles for each object (Color Space Arrays, also called
CSAs), and Postscript destination profiles with the proper rendering intent for each object (ColorRendering Dictionaries, also called CRDsl. But
you need a RIP that consistently uses them. Some RIPSignore CSAs and
CRDs entirely, others honor RGB CSAs and ignore CMYKCSAs,and some
use CSAs but ignore downloaded CRDs and use only the RIP's built-in
CRD. Ifyou're tenacious enough orluckyenough to sort through the mess
and create a functioning system for anything more complex than basic
proofing conversions, we take our hats off to you-and as we mentioned
elsewhere, maybe you should write a book!

We sometimes find itstrange that smart people who use computers every
day to carry out tasks of great complexity simply glaze over as soon as the
word "scripting" is mentioned. If pressed, we'd have to confess that we
often number ourselves among those smart people-we're far from being
scripting mavens, but we often use simple scripts, and we've seen what
the more complex ones can do.
Scripting can save enormous amounts of time and eliminate equally
enormous amounts of mind-numbing drudgery. But making it do so demands a hefty initial Investment of time and energy.

Chaplr n:Automation and Scripting

575

What Is Scripting?
Scripting is the act of witinga script in a language like AppleScript,lavaScript, orVisual Basic, that makes the computer do useful things. As such.
it's a bit like programming in the way parking a car is a bit like landing the
Space Shuttle-both requireattention to detail and skills that you have to
learn, but one requires attention to a lot more details, and learning a lot
more skills, than the other.
Writingscriptsisagrratdealsimplerthanwritingcode, but it doesrequire
you to learn a scripting language. and while scripting language resembles
natural human language much more than programming languages do, it's
not by any stretch of the imagination natural human language.

What Can kripting Do?

Scripting can do almost anything you can imagine. It lets you leverage
existing applications--for example. Photoshop 7 offers extensive lavaScript support, and almost anything you can do in Photoshop can be
IavaScripted.You can also write scripts that automate the communication
betweenapplications-forexample,an AppleScript could pass offimages
to Photoshop, have it run a JavaScriptto resize. rotate, and color manage
lhem, and then pass them onto QuarkXPress for automated layout. The
possibilities are nearly endless.

Scripting Advantages
One of the biggest advantages that scripting has ovm other automation
tools is that it%essentially fre+the tools for scripting are already built
intoyow operatingsystem. (Of course, your time is probably worth something too.) But scripting offers other advantageshere are the ones we
find most compelling.

Customlubili(.l. Perhaps the biggest benefit of scripting is that it can

automate almost any custom task as long as you can break it down into
manageable chunks. Somescriptinglanguages and tools--suchasAppleScript with AppleScript Studio, and Visual Basic-let you create actual
standalone applications, includinga user interfaceto prompt the user for
information or to provide status report feedback.
Lmryllng existing applicatlom. With AppleScript, you can leverage
the unique capabilities of existing applications as well as existing scripts.

476

Real World Color Managmwnl, Sccond Edltion

For example, you can create a script that launches Photoshop, asks it to
open a folder fun of images, embed a profile, then resave the image in a
different format.
You can do many of the same ldnds of automation using Photoshop's
Actions, but Actionsdon't allow conditionals, while scripts do. If you simply use an Action to embed a profile in a folder full of images, it will fail
if the folder contains images in different modes: with a script, you can
make Photoshop run a dserent Action to embed the appropriate profile
in RGB and CMYK images.

Sprcd. By eliminating unnecessary human interaction, you can drastically reduce the total time from beginning to completion. We've seen
advanced, moderately complex scripts cut task time from 3045 minutes
down to 1-2 minutes.

Scripting Disadvantages
The only real disadvantage to scripting is that you have lo learn to do
it, and complex scripting seems to require a certain mind-set that some
people have and others simply have not. You can't buy scripts off-theshelfthe way you can other automation products. though the Macintosh
platform supplies some handy scripts that automate simple color-management tasks. On Mac OS 9, they're in the Apple Extras:ColorSync FxtrasAppleScripts foldec on Mac OS X, they're in the ILibrarylColorSync
/Scripts folder.
Learning the scripting languageis only one part ofthe challenge.While
you can eventually save time using custom scripts to automate common tasks, the development, testing, and debugging time is often quite
lengthy.

Scripting Examples
The three scripting methods thar we've selected-PhotoshopActions, JavaScript, and AppleScript-certainly aren't the only ones, but they're the
most widely supported, and probably the most widely used. Photoshop
Actions are mostly cross-platform, JavaScripts are almost completely
cross-platform, while AppleScript is only available on Mac 0s.

Chapter lk Antonutlon and kriptlng

477

Our example scripts show what it takes to make all three scripting
methods assign aprofile to an image (ColorMatch RGB),convert it to U S .
Sheetfed Coatedv2, and then save itwith U.S. Sheetfed Coated v2 embedded. These scripts were provided courtesy of Nathan Wade, who can be
reached at [email protected] [in case you'd like to hire him instead of
learning scripting yourself).
Photoshop Actions are by far the easiest of the three-if you can do it
with your mouse in Photoshop, you can probably record it as an Action.
Then you can either run it manually, make it into a droplet, or use it with
the Automate:Batch command. Photoshop Actions are generally crossplatform-the main problem area is file and folder navigation, which uses
somewhat different syntax on each platform (see Figure 17-4).
Figure 174
photoshopAction
example

.,
.

c o
I,

7 1r

2-* s g n-G-~ n ~ " f m b ! d S e l ~ ~ ~ ~

~

j l -

*iqn.Cmnrl.Lu

""hhiOI*

~ r s n ~ ~ i u ndruman
nt

Photoshop also supports JavaScript,which has more cross-platform

capability than Actions, but demands that you know something about
lavascript. Ifyou can do it with the mouse, you can do it with JavaScript,
and even you can't do it with the mouse. you might still be able to do it
with JavaScript.See Figure 17-5 for the JavaScript version of our scripting
example--it does the same thing as the Photoshop Action.
Last, but not least, is AppleScript. AppleScript is unique in that it can
not only access ColorSync directly and do transformations on supported
image types, but it can also drive scriptable applications. A single AppleScript can ask Photoshop to do a series of tasks--even ask it to run a prewritten JavaScript-and then forward the processed files to QuarkXPress
or InDesign for automatic formatting and layout.

478

Real World Color Iknagcmenl, Second Edition

JavaScriptexample

try
var docRef = activellocument:
var cmykPmfileName = "U.S. Sheetfed Coated v2";

docRef.colorProfileName = "color~atchRGB";
docRef.comertProfile( cmykProfileName, 1ntent.RELATIVECOLORIMETRIC,
true, false );
alert( "Your document has been converted
from ColorMatch RGB to U.S.Sheetfed Coated v2,
using Relative Colorimetric Intent, Black Point
Compensation, and No Dither.
The file will now
be saved with embedded profile.* )
docRef.save();

k
catch( e

/ / An error occurred.
throw e:

1
/ / Everything went Ok.

>
else

t
alert( "You mtmt have a document open to
add the filename 1 " ) ;

>

AppleScript Studio, new in Mac OS X. is an AppleScript development

application that lets suitably motivated individuals create full-blown applicationswith an actual userioterface.AppleScriptsin MacOSXcan also
access UNIX command-line saipts. and makeXML-RPC caUs and SOAP
requests for remoteproceduresover the 1nternet.Ver)lcoolstuff for those
who understand it.
In the AppleScript example we've included, you'll first notice it's a lot
longer, because AppleScript assumes nothing. You have to teach it pretty
much everything. including the fact you want to open an image. While
you need to be a lot more specific and detailed when using AppleScript,
it lets you do things that would otherwise only be possible if you became
a programmer (see Figure 17-61.

flgun 17-6 on run

Applescript example

display dialog "~ssigntbe default RGB

Profile to an image, Convert to the default CMYK
Profile, and Save with Profile Embedded. Modify
the script to embed another profile instead."
set somefile to choose file with prompt "Choose an
image''
open somefile
end run
on open draggeditems
tell application "ColorSyncScripting"to launch
display dialog "~ssignthe default RGB
Profile, Convert to default CMYK Profile, and
Save with Embedded Profile?"
set sourcefilesto
filelistFromSelection(draggeditems)

repeat with thispile in sourcefiles

tell application "Finder" to reveal item
thisFile
set thisFile to thisFile as alias
tell application "ColorSyncScriptingset sourceProf to default RGB profile
set matchProf to default CMYK profile
trlmatch thispile from source
sourceprof to destination
matchprof matching with
relative colorimetric
intent using quality best
end try
end tell
end repeat
tell application 'colorSyncScripting' to set quit
delay to 5
end open
--returns dragged files OR files at Erst level of one
dragged folder
on filelistFromSelection(theselection)
set hasfolder to false
tell application "FinderY
repeat with thisItem in (theselecrion as lisrl
if (class of item thisItem is folder)
or (class of item thisItem is disk) then

480

Real World Color Llanagement, Second Wlllon

set h a s f o l d e r t o t r u e

Figure 0-6

Applescript example,
continlled

end if
end repeat
end t e l l

if ( ( c o u n t it& o f ( t h e s e l e c t i o n a s list))> 1 ) and

h a s f o l d e r then
d i s p l a y d i a l o g r rag m u l t i p l e f i l e s o r a s i n g l e
f o l d e r . " b u t t o n s {"OK"} d e f a u l t b u t t o n 1
return " "
end i f
t e l l a p p l i c a t i o n "Finder"
if h a s f o l d e r then
s e t f i l e l i s t t o (every i t e m of f o l d e r
( i t e m 1 of (theselection a s
l i s t ) ) ) as a l i a s list
else
set f i l e l i s t to ( t h e s e l e c t i o n a s l i s t )
end if
end t e l l
return f i l e l i s t
end f i l e l i s t F r o m S e l e c t i o n

Hybrid Scripting and Remote Profile Access

Our friend Nathan Wade (www.scriptingpro.biz)has provided a simple
script that further demonstrates the power of combining technologies.
This script is an AppleScript that accesses a destinaTion profile remotely
stored on a server, installs the profile, then uses lavascript to instruct
Photoshop CS to convert a document. and save it with the profile embedded. Because it uses JavaScript, this capability is not Limited to just
Photoshop CS.
On Windows machines, you can useVisual Basic to access the URL
location for the remote profile, download and install it, and then have
Photoshop CS execute the TavaSctipt.
Platform-specific scripting is used to enable the Web access portion of
this automation sequence,with abulkofthe code in platform-independent
JavaScriptto perform image handling and conversion.

Chapter

Autmtlon and Smipthg

4M

The AppleScriptis shown in Figute 17-7-make sure you name it "Download Profile - Use with Photoshop CS JavaScript.scpt".The JavaScriptis
shown in Figure 17-&make sure you name ir'rwcm.js".Thesearetheonly
files you need,and you can download them from ~ c o l o r r e m e d i e s . c o m l
realworldcolor.

AppleScript to download
remote profile

set the target-VRL to "http: //www.colorremedies.com/

downloads/rwcmpro~le.icc"
set the destination-file to ("/Library/ColorSync/Profiles/

rwcmprofile.icc")
t e l l application "URL Access Scripting''

download t a r g e t to file destination-file

replacing yes w i t h progress
end tell
tell application "Adobe Photoshop CS" to do javascript file
((path to desktop as string) h "rwcm-js")

try (:

JavaScriptto convert
image using remote
pmAie. and save with
profile embedded

oar docRef = new File(app.path.toString0 +

"/Samples/~une.tif");
open (dome);
var downloadedCMYKProfile = 'rwcmproii1e.i~~';

activeD~~~ment.convertProfile(downloadedCMYKProfile,
Intent.RELATIVECOLORIMETRIC, true, false);

alert('The file has been converted from its working

space to the profile located at http://www.colorrernedies.
com/downloads/rwcmprofile.~ccusing ~elativeColorimetric
Intent, alack Point Compensation, and No Dither.');

I
catch(e)
throw e;

489

Real World Color Hanagcmenf, SIcond Edition

The Automation Trade-off

The trade-off inherent in automation is that you gain efficiency at the
expense of control over individual elements. It's up to you to decide where
that trade-off needs to lie in your workflow, but a good rule of thumb is
that if you find yourself doing exactly the same task over and over again,
it's a good candidate for automation.The kind of automation you choose
will depend on your workflow,your budget, and your skills, but it's always
agoodidea to keep your eyes open for aspects ofyour work that cry out for
automation.When you think about it, you probably have mote computing
power on your desktop than NASA had in its entire organization when it
launched thevoyager program. If that power can free you fromrepetitive
drudgev, why not let it?

Workflows
Bringing It All Together

And so we come to the end of our color management odyssey. You've

learnedall about the many and varied ways our eyes interact with photons to produce the sensation we call color; howwe use colorants such as
inks, filters, and phosphors to make our color-reproduction devices emit
or reflect photons that we interpret as color; and how we use profiles and
CMMs to control thesecolorants so that we have a better chance of seeing
the color we want. We've also told you repeatedly that color management
does only two things-if you can't remember what they are, you aren't
ready for this chapter!
More importantly, we hope you now realize that color management is
partofabiggerpicture. It's more than just making orusing ICC profiles-it
requires calibration, process control, collecting good measurement data,
understanding application and driver settings, and above all, bringing all
these together into a coherent workflow.
We've tried to give you the insights and the vocabulary you need to
think criticallyabout color management, to understand the hidden meaningbehind the often-abstruse menu commands and dialog boxes that you
encounter in your applications and device drivers, and to break down the
color management operations those menu commands and dialog boxes
control into their component parts.

483

485

RraI World Color Management, Second Edition

Why? Because in this chapter, we'll show you how to bring all this
knowledge together to build a color-management workflow that suits
your unique requirements.
We can't possiblyspell out every possible color-managementworkflow,
because there are probably as many worMows as there are practitioners.
We don't want to simply tell you which buttons to push-because for one
thing, youwon't really learn anything except for a rote set of steps, and
for another, as soon as you've learned the steps, one or another vendor
will rename, move, or change the functionalityof the buttons, leaving you
back at square one.
Instead, we'll show youhow to breakaworMow down into its essential
components, step you through the questions you need to ask, and show
you the possible strategies to use in building your worldlow. We will, of
course, include some examples, but you should take (hem as illustrations
of the possibilities rather than hard-and-fast rules.

The Four Stages of Color Management

All color-management workflows can be broken down into four basic
steps, and understanding these steps is the key to analvzing, and then
meeting, your needs. The basic steps are:
b

Defining color meaning-specifying just what colors your RGB and

CMYK numbers represent, or collecting LAB measurements of colors
that you'll eventually translate into RGB or CMYK.

Normalizing color-bringing all your color elements into a common

color space.This step is optional, but most workflows benefitfrom the
simplicity it brings.

Converting for output-producing the set of numbers that will make

your final output device(s)produce the desired color appearance.

Proofing-making sure that your final output will, in fact, have the
desired color appearance.
Let's examine these steps in detail.

Chsplcr 18: Building Color.knal.9.d Workflow

585

Defining Color Meaning

You can't match a color until you know what that color is, so the first
step in a color-management workflow is always to attach a specific color
meaning to the document or object. Just how you accomplish that, and
which color meaning you attach, depends on the nature of the document
and its source.

Profiled mptura. Profiled captures are pretty straightforward. If the

capture device driver can embed the capture profile, allow it to do so. If it
can't, you must assign the profile manually, either in an editing application (such as Photoshop) or using automation (such as Applescript).
Unpmfikd aptum. Unprofiledcaptures are tricker.You have to decide
on a color appearance. perhaps arbitrarily. 7Lpically, you'll either assign
a profile that produces something close to the desired appearance, or just
assign an editing space. In the first case, you should convert to an intermediate editing space, and in either case, you'll probably color-correct
the document-see "Workflow Into and Out of Color Management" in
Chapter 10, Color Mnnngemmt Workflow.
Unproflled legacy imagcr. Unprofiled legacy images present similar
challenges to unprofiled captures.With RGB images, there's a good chance
that a monitor-based profile such as Colormatch RGB (gamma 1.8) or
sRGB (gamma 2.2) will produce close to the desired appearance, since
they're almost certainly based on monitor appearance (see "When Color
Management Starts" in Chapter 10, Color Management WorL~orul.
With CMYKimages,our philosophy is that when the CMYK destination
is known, you should simply assign that CMYK profile and edit as necessary-if theCMYKdestination isunknown. you have no business working
in CMYK in the first place. In desperate situations, though, we may try
assigning different CMYK profiles, then convert to ourCMYK space.

Vector art. Some vector art applications insist that all the content in a
document must be either RGB or CMYK, though you can force them to
accept imported graphin in the other mode. In the case of legacy vector
art files, you have two options:

486

Real World Color Management, k o n d Edltlon

If you h o w the new destination, make itthesource profile, andrework
the file.

if you know the original destination, make it the source profile, then
repurpose the file [either by converting at print time in a desktop application, or using automation).
In the case of new vector art files, you have three options:
Workin RGB using the same space you use in your other applications.
but realize many of the colors in this space won't exist in the final o~rtput space. Take advantage of soft-proofing to a generic CMYK space
(such as SWOP) to get a general idea of such limitations.
b

Work in the widest-gamut device space in which the artwork could

possibly be used, and the onewith which you're most familiar.

If the destination is known, assign that profile,and work in that space.

Spot colon. Spot colors probablycausemore aoublethan any other type

of color element. Pirst, unless the color will actually be printed as a spot
color. it makes little or no sense to choose a spot color from one of the
many Pantone or Focoltonelibraries,because many of these colors simply
aren't reproducible by CMYK printing-yet a depressingly large number
of people continue to do so, with accompanying disappointment when
the job hits production. Second, the published CMYK values in the various solid-to-process color guides are only valid for the printing process
that was used to print the guides-if you use these values in your job, the
resultingcolor may bear little resemblance to the intended one.
Currently, only the non-process Pantone libraries in Photoshop 7 and
automatically produce custom RGB and CMYK values
CS use -they
based on the active document's profile. AU other applications use libraries containing hard-coded CMYK values for print (see Chapter 15. Color
Mnnngement in QuarkDmss, and Chapter 17. Automation mrdScripting,
for suggestions on how to deal with solid-to-process).

Normaliring Color
Once you've defined all your color. youmay well find that you have a raft
of profiles-several different capture devices, a default RGB and a default CMYKprofle in each application. and possiblysome legacy CMYK

Chap* 18: Bulldlng Cob-Managad Workllown

487

profiles. It's conceptually possible. and even, with some difficulty, practically possible to carry these profiles through the workflow to final output-but yourlife will be madea great dealeasier ifyou normnlizeall your
to asingle RGR
color by conveningit to asingle space, or, morepracti~allv~
and a single CMYK space.
Normalizingyour color simplifies your life in two ways:
b

It frees you from having to deal with profile mismatch warnings in all
stages of production except the initial ones.
Itletsvou rely on assumed profiles. Even if you prefer lo embed profiles,
mostvector applicationsdon't embed profilesin EPS files,so they force
you to rely on assumptions. iYormalizing your color makes it much more
likely that the default assumed profiles will be the correct ones.

In addition, most RGB capture spaces don't work particularly well as

editing spaces since they're typically neither perceptually uniform nor
gray-balanced. So for all these reasons, we recommend that you normalize
your color as soon as is practical.

Standardize defaults. All color-managed applications let you set default

profiles for RGB and CMYK. Your sanity will be much less endangered if
you make sure that all your color-managed applications use the same
default profile for RGB and CMYK. Even ifyou're fanatical about embedding profiles, the occasional untagged file may slip through, and as noted
above, vector EPS files often don't contain embedded profiles. If you normalize allyour color by converting to a single RGB and a single CMYK.
and then set your defaults to these same profiles. your workflow will be
much more robust than if you fail to do either.

Whm to nonnaliu. In thevast majority ofcases. we recommend normalizing RGB color as soonasyour applicationswill allowyou to do so.Adand
marketing agencies can normalizestock photography to the desired RGB
intermediate space early in thedesign process, and ask photographers to
supply files in the same space.
Repress and printing shops already have dedicated preflighting, and
can include normalization as n part of it. Typically, prepress shops will
normalize everything to CMYK, converting RGB content to CMYK and
repurposing CMYKcontent when thesource is different from the intended

488

Real World Color Management, Second Wition

destination. Converting RGB content to CMYK before the CMYK destination is known isn't normalizing-it's premature binding (see the sidebar,
"Premature Binding," in Chapter 10, ColorManngementWo&/1otu)-and
is an idiocy we'd like to see stamped out!
The one exception is black-only text. For example, neither InDesign
nor QuarkXPress exclude black-only text from color management-in
LnDesign, it's in Document CMYK, and in QuarkXPress it uses the CMYK
Default Source Profile for Solid Colors. If these sources aren't identical to
the destination profile used on output, you'll get rich black text, and registration headaches on press. We'd like to see vendors treat black text by
default as device-dependent, and untouchable by color management.

Converting for Output

At some point, you need to convert your color to the final output space.
The question is, when? Obviously. you can't convert to the output space
until you knowwhat that output space is. Once you know, two issues bear
on the decision of when to convert.

O m output or many?If you're going to a singIe output process, the question of when to convert boils down to the trade-off between seeing and
editingvourcolorinfinaloutputspace, and workingwith smaller, more agile
RGB files. If you find that automated conversions in a color server, or even
in the RIP itself, give you acceptableresults, you can achievesignificant produ&vity gains by working in RGB and adopting a late-binding worktlow.
The downside is that you don't see the separations until they're on press.
If you have multiple output destinations, the question then becomes,
do you prepare separate files for each output scenario, or do you rely on
late-binding separations?The issuesare confidence and quality. If you're
confident that you can obtainusable resultsfrom automatic separations.
alate-binding workflow where final separations happen in a color server
or RIP mav make sense.

Witingin wtpvt Space. Our experience has taughtus that for theultimate
in quality, you need to make final edits in the output space. Automated
conversions can get you in the ballpark, but they only h o w about color
spaces, not about the colors in the actual document. Hence they treat all
content identically, ignoringthe kinds of perceptual issues we raised all the
way back in Chapter 1 (see "Where the Models Fail" in Chapter 1, What Is

Chapter IS: Building ColobManagd Workflow

489

Color?),and so they sometimes produce results that fall short of ideal. If

you need to edit color in the final output space. you need to make separate
files for each output scenario.
You can, however, edit in an RGB working space while looking at a
simulation of the final output, which in many cases is almost as good as
editing in final output space (see"Soft-proofing," below). Rut if you need
to make different edits for different outputs, you'll need to makeseparate
fdes for each output.

Proofing
You can't, ohviously.pmfyourcoloruntilyou knowthe final destination.But
you can proof your color before converting it to the final output space.
Soft-proding. Most color-managed applications let you view a simulation of final output while you're working on files in someother space--for
example, previewing CMYK output while working in RGB (see the sidebar, "Soft-Proofing Basics," in Chapter 10.Color Mnnngement Workflow).
While some types of edits, such as tweaks to the black plate, can only be
made in CMYK. you can easily make many other kinds of optimization
for tone and color on an RGB file prior to conversion, using an accurate
display simulation as a guide.

Hard proofing.The majority of color-managed applications let you print

asimulation ofyour final output toacompositeprinter, even ifyou haven't
yet converted to final output. We don't recommend using prints made in
this fashion as contract proofs,even if you can get theother party to agree
to them, but they can be helpful in cutting the number of contract proofs
you eventually need down to a single set. In a multiple-output scenario,
this feature is particularly useful, becauseyou can pmof the different outputs beforeconvertingtofinal output from thesame master file by simply
changing the final output profile.
More-sophisticated proofing systems let you send the final output data
to the proofer, and either perform an internal conversion in the RIP or
color server. or configure the hardware to emulate the final output. In
either case. you can make your color-managed application perform the
conversion to final output space as part of the print process-so again.
when dealing with multiple output scenarios, you can proof each one
from the same master He by simply changing the final output profile.

490

Real World Color Msnagemenl, kcond Edition

Step-by-step Color Management

The breakdown of color management that we've presented above should
help you conceptualize the necessary steps in the workflow, from color
capture or specification, through editing, to proofing and final output.
The next stage in building a workflow is figuring out just what you need
it to do.

Determining
- Your Needs

At the hegmning o l this chapter, we pointed out that there are probably
as many color-management workflows as there arepractitioners. We can't
possibly spell out all possible workflows in detail. Instead, we'll step you
through the questions you need to answer to determine your specific
workflow needs.

Software, File Formats, and Devices

The first set of questions deal with the software and hardware you use to
do your work. Ask yourself the following:
Which applications will I use?
b

What file formats will I need to support?

What devices will I use to capture and reproduce color?

Applications.Your workflow options depend on the capabilities of the

applications you use. Some applications let you attach a different profile
to each element in a job, some are limited to a single RCB and a single
CMYKprofile,andsome force you to normalize all your color into a single
profile's space.
If you're working with a single application, the choices are relatively
straightforward, but if you're working with multiple applications, you
need to make sure that your workflow accommodates them all-and that
almost invariably means making sure that the workflow can handle the
most limited application of the set.

Chapter 18: Bulldlng Cdor-Managed Worknovr

491

File bwtltltr. TIFF, IPEG,and PDF have robust support for embedded
profiles, though you need to be vigilant for applications that either fail
to detect embedded profiles or use them incorrectly. EPS and DCS EPS,
however, are much less reliablein terms of profile embedding.
Youneed to figure outwhich partsofyourworMowcan rely onembedded profiles, and which parts must rely on assumed profiles. Then you
need to ensure that you have appropriate safeguards in place to make
certain that the correctprofilesare assumed.

Device. Capture devicesvarywidelyin their ability lo handle and embed

profiles.Can your input devices,or their drivers, embed profiles correctly.
or do you need to assign them manually? Can they convert the captures
into an intermediate space and produce normalized captures, or do you
need to hendle nonnalition elsewhere?
Proofing devices also vary in their capabilities. In theory, you can use
any device whose gamut wholly contains the gamut of the final outpu~as
a proofer. The question is whether you can configure the proofer to accept final output data (as you can with a proofing RIP or color server). or
whether you need to rely on application-levelcolor management to send
a simulation of final output to the proofer.

What Outputs Do I Need?

Your output needs may vary from a single output that's known before
you start the job, to multiple outputs that may be unknown until the last
minute. or may even be unknownuntil afteryou'vesubmitted the job.The
issue is when to convert to final output space. You can't do that until you
know what the final output is.

Knom output. If you know the output right from the start, and you're
dealing with a single output process, an early-binding worknow, where
everything is converted to output space early in the process, makes a great
deal of sense, particularly if you have to interact with outside providers
who aren't color-management savvy
If you have to deal with multiple known outputs, things become a little
morecomplex Early bindingkeeps thingsimple. but increasestheworkload
because you have to prepare separate documents for each output process.

Real World Color Management, Second Edition

One posslble solution is to keep imagesand criticalvectorcolor in RGB
until it's time to generate final output, while defining black text and noncritical nativecolor elements in CMYK, then assign the final output profile
when yougenerate final output You may have to tweaksomeCMYKcolor
builds, but your black-only elements such as text stay black-only, and
your imported RGB elements get separated to final C W when you create [he final output file.

Unknown output. Ifyou have to prepare files for unknown outputs, you'll
almost certainly find that the politicalissues outweigh the technical ones.
From a technical standpoint, the best solution is to submit the job in a
device-independent form such as profiled RGB color, or even LAB. In the
real world, though, the chances of such a job getting processed correctly
alter it's left your hands are uncertain at best. If you can obtain a signed
contract stating unambiguously that the downstream operation will accept Nes in a specified, named RGB space (or in LAB),and will take full responsibility for the color from then on, by all means submit profiled RGB
or LAB. Absent such a contract, you're almost certainly inviting trouble.
Often, you'U be forced into a premature-binding workflow instead
(see the sidebar. "Premature Binding," in Chapter 10, Color Management
IVorkfloru).From a color management standpoint, such workflows make
no sense, but we do have to livein the real world, which of course doesn't
always make sense. If you're stuckin a situation like this, we offer the following suggestions:
b

Specify a contract proofing system, such as Kodak Approval or Fuji

ColorArt, by name, then separate for the proofing system and submit proofer CMYK along with the had-copy proofs created from the
CMYK data.

Obtain as much information as possible about the output-type of

press, coated or uncoated stock, anticipated dot gain-and choose a
generic profile that approximates those conditions. The press profiles
that accompany Adobe Photoshop. Illustrator, and InDesign, for example, are fairly"safe." but the more you know about theprintingprocess, the better your chances of picking an appropriate profile.

Chapter 18: Bulldlng Color-Managed Workb

493

Absent any better information, use Adobe's SWOP Coated v2 profile

(or, if you know that the job will be printed on uncoated stock, the
SWOP Uncoated v2 profile). The results may not be ideal, but they
should be usable.

If you have to deal with multiple unknown outputs, the first and last of
the three suggestions above are equally applicable. The unknown is the
unknown, whether it's one or m a n y t h e best vou can do is to provide a
file that's approximatelysuitable for printing, preferably accompanied by
hard-copy proofs that indicate your intentions clearly.

Automation and Quality

Aurornation can save a great deal of time and effort. But automated
processes are dumber than the dumbest person you've ever had to deal
with-we hold firm to the view that the acronym 'XI" stands for Applied
Ignorance rather than Artificial tntelligence!
Automated color management knows nothing about images, nothing
about color context, nothing about memory colors. It simply performs
one-size-fits-all operations, one pixel at a time, and it has no means of
evaluating its own results. Automation does the same thing every time.

What you can automate. Good candidates for automation include things
like normalizing profiled RGB captures to a single RGB editing space. For
example, if all your imagesources embed profiles, you can safely set Photoshop's Color Management Policy for RGB color to 'Convert to Working
RGB." Then, whenever Photoshop encounters an image whose embedded
profile is different from the working RGB space, it converts it from the
embedded profile space to working RGB.
The downside, of course, is that Photoshop has no way of knowing
whether or not the embedded profile is the correct one--it just blithely
goes ahead and makes theconversion. So ifyou also receive imagery from
sources that don't embed profiles correctly, the automation may produce
a bad result, creatingworkinstead of eliminating it.
Automated output conversions can often work well, but again, they
treat all color identically So you'll get good quality, but not as good as
you'd get with a skilled operator optimizingeach image. You can, however,
create optinkzed solid-to-process conversions for solid colors, then apply
them automatically using a color server or Praxisoft'sVectorPro.

494

Real World Color Management, Iscond Witlon

What ~u can't automate.You can't really automate anything requiring
intelligent human intervention-optimizing images, choosing different
outcomes on acase-by-case basis. That said, it's often worthwhile, albeit
humbling, to compare totally automated results with those produced by
your careful handwork. The questions you need to ask in making such a
comparison are
Is there a discernible quality difference between the two?
b

Can you make a business case for preserving that quality difference?

Once you've defined your needs, the last step in buildingyour workflow
is to look at the workflow tools that your applications and device drivers
offer, and decide how best to employ them in the four stages of color
management.

Workflow Tools
You understand the stages of color management, and you've determined
what you need to produce.Thelast step is to lookat the various tools your
software and hardware offer, and decide how to apply them.

Embedded or Assumed Profiles

Color management workflows really only offer two ways to define color
meaning, which is always accomplished by associating a profile with the
document or object:
b

Embedding a profile
Assuming a profile

Each approach has its pros and cons. Our very basic rule of thumb is that
we generally embed profiles in RGB, and we generally assume profiles for
CMYK. Wealways tryto use CMYKas afinaldestination-CMYK-to-CMYK
conversions do work, but they rarely give as good results as going back to
the original RGB and reseparating--so we seldom need to deal with more
than one flavor of CMYKat a time. That makes CMYK an ideal candidate for
an assumed profile, since we know we have only one flavor of CMYK.

C h a p l o 18: Building Cob-&nag4 W o r m

495

CMYK. We often make families of profiles ibr a given CMYK process, with
different black generation characteristics. For example, we used three
different profiles to create the CMYK in this book. with different black
shapes, bur we only used the different profiles to convert from RGB to
CMYK. Once the content is converted to CMYK, the blackgenerationused
in the profile is no longer relevant.
in this land of workflow, all the assumed CMYK profile does is attach
a color meaning to the CMYK values by acting as a source profile. Any of
the three profiles we created can do that equally well-its sole purpose is
to provide asource profile for display simulations and bard-copy proofing.
CMYKpmfiles are also quite large--the ones we used for this book are
2.4 MB each-so embedding would mean we'd have to sling a lot more
data around, and sinceall our CMMiis thesame CMYK, much of that data
would be redundant.

ffiB. RGB matrix profiles, on the other hand, are tiny. We normaliue our
RGB color into an editing space as soon after capture as is practical, and
editing space profiles are tiny, so we always embed them.
One key differencebetween embedded-profileworktlows and assumedprofileworkflowsare thatin the former, thecolormeaning is automatically
attached to each element, while in the latter. the color meaning is applied
manually by you, the user.
Theotherkey differenceis that, in anembedded-profile workflow, each
element can have a different profile embedded, while in an assumedprofile workflow, you can generally have only one RGB and one CMlX
profile.Therefote, the trick to making assumed profiles work is normalization-making sure that allyour content is converted to just one flavor of
RGB and just oneflavor of CMYK before it enters the assumed-profile part
of your workflow. Do that, and you can safely rely on assumed profiles.
Fail to do so, and all bets are off!

Normalization and Editing Spaces

The case for normalizing CMYK in an individual job should be obviousultimately. it's all going to print using the same four CMYK inks on the
same paper. The case for normalizing RGB is a little less so. Here are the
main reasons we tend to do so:

596

R u l World Color Management, Second Wltlon

We prefer not to edit images in capture spaces because they're usually

far born gray-balanced or perceptually uniform, so we convert to an
intermediate editing space. That being the case, it usually makes sense
lo settle on a single intermediate editing space.
Applications vary in their ability to handle embedded profiles cotrectly, QttarkXPress being anexample of one that often does not. If we
know that all our RGB is, for example, Adobe RGB (19981,we can set
it as the default RGB. Then, even if an application fails to honor the
embedded profile, we know that it will use the correct interpretation
of RGB because it's the only interpretation available.

It simplifies the workflow.

Somevery quality-conscious~~or~ows
may need multiple RGB spaces:
all editing spaces represent a series of trade-offs, and no single one can
produce ideal results for all image sources and destinationssee "One
later in this chapter, for an example. In most cases,
Pixel's Journey
though, the incremental gain from using multiple RGB spaces is outweighed by the increase in complexity that doing so brings.

...."

Choosing an RGB space. Our main criterion when choosing an RGB

space is that its gamul provides a reasonable match to the gamut of our
Intended output without wasting a lot of bits on color definitions thar
don't correspond to anythingwe can capture, reproduce, or possiblyeven
see. This is always a trade-off.
Spaces defined by RGB primaries, white point, and onecurve all have
a distinctive 3-D shape. An RGB space that completely encompasses the
gamut of most CMYKoutputs has to bevery large indeed-so large that a
lot of the color definitions it contains don't correspond to real colors. For
example, in ProPhoto RGB (whichis certainlylarge enough to encompass
any output process we've experienced), R 0 , G 0. B 255 corresponds in
LAB to a blue that's fully saturated but has a luminance (L*) of zero! That
doesn't correspond to anything we can see--all colors with a luminance
of zero look the same to us, and they're black.
We use a variety of specialized editing spaces for different purposes,
but our general recommendation is to start outwith a single mainstream
editing space such as Adobe RGB (1998) or Colormatch RGB, and stick
with it unless and until you run into a specific limitation that another
space can better address.

Some workflows avoid RGB altogether, and useL4B as the normalized

space. Hardly anyone in the United States seems to use such a workflow.
but they're quite popular in Europe, where Heidelberg has evangelized
Great Normalizer").
them for years (see the sidebar, "1-The

The Cop at the Door-Warnings

Someapplications. notably the BigThree from Adobe--Photoshop, InDesign, and Illustrator-can be configured to warn you when they encounter color that hasn't been normalized-that is, it contains a n embedded
profile different from the o n e specified as the default, or in Adobe terminology, the "Working Space."You set this option by turning o n the "Profile
Mismatch: Ask When Opening" checkbox in the respective applications'
Color Settings (see "Warnings-Manual Overrides" in Chapter 12, The
Adobe Common ColorArchitecrure). Unfortunately, no applications warn
you when you sow a file in a space other than the default.
Most other applicationslacksuch features, so you have t o develop your
own proceduresfor checkingand enforcingnormalization,perhaps using
scripting (see "Scripting" in Chapter 17, Autontotion a n d Scriptin@.

LAB-he Great Normalizer

LAB-hased workflows may seem
like a great idea-LAB is relatively
unan~biguous(you do need to
specify white point and standard
observer, hut unless otherwise
indicated, it's safe to assume D50
2-degree LAB), and by definition.
it contains all colon. So why not
simply use LAB until it's time to
convert to output?
Well, if the output is known.
and you deal with all your image
optimizarion issues before you
convert to IAB, a LAB workflow
is jmst fine. Likewise, if you'rewilling lo accept the gamut mappings
you get h m automated conversions, a LAB workflow is equally
fine. But if you plan to optimize
your color for diicrent outputs.

a LAB workflow has the foiiowing

disadvantages:

possible eight-bit-per-channel

Eight-bit-per-channel LAB files

are fairly fragile things. Since
IAB encoding has to cover the
entire range of vision, the data
points in LAB channels are
stretched pretty far apart. So
it's hard to make subtle edits
on eight-bit-per-channel LAB
files.

Since LAB encoding has to

contain all the colors we can
see, it's also forced to contain
many values that don't correspond to colors we can see.The
estimates we've heard indicate
that somewhere between 5 and
7 million of the 16.8 million

LAB values actually correspond

to real visible colon, so encoded LAB contains a lot of values for which there are no real
visible color equivalents.

LAB isn't the most intultwe

space for editing or specifying
colors.

Many key applications and

some key file formarsdon'tsupport LAB.
If you understand the llrnita-

tiotls and the requirements of

LAB-based workflows, you can
makethem work, but they'requite
specialized.

498

Real World Color Manogetnenf, Second Edition

Black Preservation and Device-Dependent Data

One of our biggest reasons for avoiding CMk'K-to-CMYK conversions is
that K-only elements such as black text or black-only drop shadows almost always wind up being converted to a rich black that contains other
inks besides black. This is almost never a desirable outcome since it introduces all kinds of registration problems on press.
The simplest solution for page-layout applications is to define all native elements in RGB so they can be repurposed, and to define black-only
text and anything else for which you want specific CMYKvalues preserved
in CMYK. Then ensure that their CMYK source profile is the same as the
destinalion profile to prevent conversion.
Early-binding workflows don't usually have this problem, but latebinding worldlows often do. There's no single ideal solution. For QuarkXPress users, Praxisoft's Compass Pro XT XTension lets you exempt 100%
black-only objects from color conversions (see "Compass Pro XT" in
Chapter 15, ColorManagement in QwrkXPressl. For other users, serverbased conversion products or devicelink profiles, both of which we discuss in Chapter 17,AutornnTionandscripring, provide possible solutions.
Othenvise, you simply have to watch out, and assign the output profile
manually to prevent any undesired conversion.

Named-Color Systems
Named-color systems such as Pantone and Focoltone present some special
color-management problems. Some are technical-Pantone is noted for
revising its libraries,so that different applications often wind up with different definitions of the same color. Some are human--users specify spot
colors forprocess jobs, or specifythe CMYKvalues from a spot-to-process
swatchbook, failing to realize that the CMYK d u e s provided mav be totally inappropriate for the job at hand. Both cause trouble.

Libraries. Wherever possible. standardize your named-color libraries so

that all your applications use the same definitions. Sometimes this isn't
possible-applications tend to use proprietary formats for their color libraries--but it's always agood idea to check for updates, thenapply those
updates in a controlled fashion so that all your applications match.

Chmpla18: kllding Color-Munagcd Workflow

499

The solid-to-process values for Pantone colors changed around May

2000 as a result of a change in the p r e s behavior Pantone selected. The
applications containing the post-Map 2000 guidevaluesare Photoshop CS,
InDesign CS. Illustrator CS, and QuarkXPress 5. FreeHand 10 and CoreDRAW 10 both use pre-May 2000 equivalents.
However, we've found that we can almost always impmve on profiledriven solid-color simulations by creating our own hand-tuned color
builds. If accurate simulations of spot colors are very important in your
workflow, consider one of the automated solutions for substituting handruned color values for spot colors that we discussed in Chapter 17. Automation and Scripting.

Appllwtlan. Applications exchange named color through EPS or DCS.

\men you create artwork in an application, the CMYK equivalentsdefined
in the libraries areincluded in the EPS file so that it can print on composite four-color devices. So artwork prepared with Illustrator 10 and FreeHand 10 produce EPSs that contain different CMYK equivalents for the
same Pantone color, and those are the values that get honored. Neither
QuarkXPress nor InDesign modifies the CMYK equivalents for Pantone
colon in EPS files.
However, if you open a QuarkXPress 4 document into QuarkXPress 5,
or vice versa. native colon get updated-the CMYK equivalents used at
print time for native Pantone elements are determined by the libraries
belonging to the application from which you print
People. It's not easy. and it's sometimes impossible, but we always 1i-y
to train our users to refrain from specifying spot colors for process jobs.
Many spot colors in the commonly used systems are outside the gamut
of CMYK prlnting, and specifying one of them for a CMYK job inevitably
leads to disappointment (orworse) later in the production process.

Simulations and Soft Proofs

Most color-managed applicationslet you force the monitortosimulatethe
final output. We find this vital in image-editing applications such as Photoshop, but we also tind it surprisingly useful in page-layout applications.

500

Real World Color Managemmt, W n d Ldnion

Our eyes always force us to judge color in context, and elements that look
great in isolationsometimes lookquite differentwhen they're placed in the
final layout. When we're working on a job with a single known output, we
always view the simulation. If we're working on a job with multiple known
outputs, we look at the worst one!
Even when the final output is unknown,we often use a CMYK siniulation
(assuming the job will eventually wind up being printed with some kind
of four-color-process printing). Doing so helps us anticipate the excesses
of over-optimistic designers who make use of the full RGB palette. If you
aneone of theseover-optimisticdesigners, try setting your monitor simulation to Adobe's U.S. Sheetfed Coated v 2 profile-CMYK may get a little
betterthan that onveryhigh-qualityjobs, but ifyou'redesigningforprint,
it will give you a reasonable estimate of the color palette you can use for
the job.

Hard Proofs
We don't typically bother with hard-copy proofs before we know the output processthey basically won't tell us anything that our monitor can't
do less expensively-but we ofren make use of our applications' capability to produce a hard-copy simulation of final output before we actually
convert to final CMYK.
Printing hard-copysimulationsof yourfinaloutput can alert you to potential problems with the final conversion, so that you can take any necessary remedial action before the potential problem turns into an actual
one. It's also a good idea to print hard-copy proofs after you've converted
to final output-you may find very slight differences behveen the proof
of the simulation and the proof of the final converted result.
It's unrealistic at this stage in the acceptance of color management
to try to use cross-rendered proofs as contract proofs (though many periodicals do so, having achieved a confidence level through repetition).
But they are useful for guidance. Of coune, they don't show problems
with screening or defects on the tilm or plates, but neither do most other
digitalproofingsystemsshortof aKodakApprovalor Creo Spectrum.That
said, inkjet printers are beginning to approach the resolution needed to
produce actual dot-based proofs--we expect to see a lot of action in this
area in the next few years.

Chapter 10: Building Cokr-hnapal WoMiow~

501

One of the consequences of the adoption of direct-to-plate printing

seems to be that nobody knows how to proof anymore. We've had the
experience where a printer provided film-based proofs--Matchprints,
in fact-for a job being printed direct-to-plate. More disturbingly, the
Matchprints were all around 20 percent heavy on the cyan:a call to the
printer provided us with the assurance that "the red always heavies up on
press"-whichleft uswonderingwhythey'd gone to the timeand expense
of running film in the first place!
In the end, the profiles we'd built for the press proved accurate, and
at the press check. the press matched our Epson inkjet guidance ptints
closely as soon as it was brought up to density The moral of this little tale
is that contract proofing is as much about responsibilityand confidence
as it is about accurate color matching.

Output Conversions
Convertingall pour color accurately to the final output is the ultimate goal
of color management. The keys to achieving that goal are first to make
sure that you've selected the correct profile for output-that's the trivial
part-and second, to make sure that each element in the job has the correct source profile--that's the nontrivial pan.
Earlybinding. Inan earlpbinding workflow,each element is either created
in final output space or is converted to final output space early in the process, so early-binding workflowsare relatively straightforward. It's always a
good idea to preflight the job, either using the built-in tools offered by the
applicationsoradedicated preflighting tool, to make sure that no stray RGB
images haw slipped through--but in general, early-bindingworkflows usually mean that someone, somewhere in the production process has seen
and checked the final color.

Late blndlng. In a late-binding workflow, things are a lot trickier-in extreme cases, the final color may not exist until you burn plates. So you
need to be very sure that every element in the job is handled correctly.
Normalization is a good safeguard for late-binding workflows, because
it reduces the final output conversion to one conversion from a single
source to final output-you still need to make sure that each element has
the correct rendering intent applied, but you don't have to w o w about
multiple source profiles.

509

Real World Colw hnagsmrnl, Second Edition

Final Analysis
-

If you'veplanned your workflowcarefully,youshould knowexactly what's

going to happen to every element in the job, including the tricky ones
like black-only text, and spot colors destined for conversion to process.
Remember the stages of color management, and ask yourself the following questions:

Has each element's color been correctly defined?

Has each element been normalized as needed to either default HGB

or output CMYK?

Does each element have the correct rendering intent applied?

Have special cases such as black-only text, black-only drop shadows.

and spot-to-process conversions been adequately addressed?

Are all the color-reproduction devices in the chain properly calibrated

to ensure that they behave the way the profiles predict they will?

When you can answer each question with a yes, you can be pretty certain
that you have a relatively bulletproof color-management worktlow.

One Pixel's Journey..

To illustrate at least part of theworkflow we used to create this book, we

thought it might prove instructive to trace the journey of just one of the
many pixels that went into its making. This particular pixel started life
one atypically sunny San Francisco afternoon in August 2001, as photons
reflected from one of the many fine specimens in the Dahlia Garden in
Golden Gate Park.

Capture a Pixel
Bruce captured these photons in his Kodak DCS 460 digital camera, took
them home, and found. upon opening the image in the Kodak DCS Acquire plug-in, that they produced an RGB value of R 248, G 13. B 0 (see
Figure 16-11. He then acquired the image into Adobe Photoshop.

ngurc 18-1

The capture

Define the Color

Like many Photoshop Import plug-ins, the Kodak DCS Acquire plug-in
doesn't embed profiles. In fact, it has no facility for using profiles at all. In
the ProPhoto RGB working space that Bruce usually uses for Kodak digital
captures, the pixel's values translated to a screaming fluorescent orange
with LAB values of L* 60, a* 128, b*103, quite diierent from the dahlia red
that first attracted Bruce's attention.
So the fist step on the pixel's color-management journey was to obtain
the right color meaning by assigning a profile that correctly described its
.appearancein this case, a custom profile he built for the camera using
.MonacoPROFILER 4.0. Assigning this profile in Photoshop changed the
translationfrom RGB to much more reasonable LAB values of L* 50, a*90,
b* 72 (see Figure 18-2). Thus we gave the pixel (along with all its siblings
that made up the rest of the image) not only a specific color meaning, but
the desired color meaning.

504

Real World Color Management, Second Editlon

Defining the color

rn

Normalize the Color

Unfortunately, La 50, a* 90, b* 72 simply isn't a color that our four-color
press can reproduce, so we knew that the image would need significant
editing to preserve the spirit, if not the literal color values, of the original when rendered in print. The DCS 460 profile space doesn't provide a
good editing environment-it's neither gray-balanced nor perceptually
uniformso we decided to convert the image to the ProPhoto RGB working space for editing, using Photoshop's Convert to Profile command (see
Figure 18-31,
Converting to ProPhoto RGB with relative colorimetric rendering preserved the LAB L* 50, a* 90, b*72 values, while changing the RGB values to
R 174, G 53, B 19. (We used ProPhoto RGB because converting the same
pixel to Adobe RGB produced R 215, G 0, B 0-clipping both the green and
blue channels-and LAB L* 53, a* 80, b* 69. It's not that Adobe RGB is a
bad space; it just doesn't workwell in the extreme reds with this particular
camera and profile.)

Simulate the Output

In this workflow, our final output was known from the start-we'd printed
other books on this press, and we'd already profiled it-so we went for
relatively early binding. Before converting to CMYK, though, we opted to

Chaptr 18: Building Cdw-Managed Wwktlowr

-

Figure 18-3

505

Normalizing the color

view a simulation of the CMYKconversion using Photoshop's Proof Setup

feature (see "Simulations and Soft-Proofing" in Chapter 12, The Adobe
Common Color Architecture).

Optimize tor Output

The soft proof immediately confirmed our fears--the predicted CMYK
values of C 0,M 88,Y 89, K 0 produced corresponding LAB values of
L* 52,a*68,b*48. We edited the image, still in RGB mode, whiie viewing
the CMYK simulation, and by increasing saturation and reducing lightness we were able to change the predicted CMYK to C 0,M 91,Y 93,K 1,
which we felt was as far as we could go in RGB mode.We decided to defer
further editing until we'd converted to output CMYK.

Convert for Output

We converted the image to our output CMYK, using the settings that we'd
determined worked best by looking at the soft proof-our heavy GCR
press profile gave us the best saturation while still holding detail. One
more edit produced a final output value of C 0,M 93,Y95,K 2, LAB L*50,
a* 69,b* 48a far cry from the original, but the best we could do in the
circumstances (see Figure 18-41,

506

Real World Color Management, Second Edtlon

Figure184

Converting for output

Proof the Output

Thus far, we've described the journey of our sample pixel from capture to
output. But it also took some side-jaunts in the various proofing scenarios. Photoshop always does a conversion on the data sent to the display:
to make the monitor reproduce our ProPhoto R 174, G 53, B 19, it had to
send R 221, G 0, B 0 to the monitor. (You can't get this information directly
because the conversion happens under the hood, and it isn't anything
that normal people need to track anyway-we just mention it to illustrate
that the pixel's journey isn't quite as straightforward as might seem at first
glance.)
And it's also mildly interesting that when we proofed the image on our
Epson2200usingPremiumLuster paper, Photoshopsent R245,G41, B102 to
the printer to make it produce the same color as press C 0, M 93, Y 95, K2.

Hand OH the Color

We saved the converted image as a CMYK TIFF with no profileembedded,
and placed it in PageMaker,where the default CMYK source profile and the
Separations profile were both set to our final output press profile. This
ensured that our pixel had the final output profile assumed as its source,
so no conversion would take place on output, but we could still enjoy accurate display and accurate inkjet proofing. The final image is shown in
Figure 18-bur pixel is in one of the petals on the left edge of the dahlia.

Chapter 18: Building Color-Managed WorMowi

507

The final image

The Devil Is in the Details

Throughout this book, we've pointed out that color management is in essence quite simple,since it does only two thimgs: define a color meaning and
preserve that colormeaning. But in a complex production workflow, you'll
encounter a host of details, each one of which must be addressed.
In Part I, we laid out the foundations for color management-the way
we see color, the ways we make our devices produce color, the models we
use to relate what we tell our devices to do with the results we see, and
the basic operation of color management systems. We also pointed out
the limitations of color management, and showed a few of the perceptual
effects that the models simply don't take into account.
In Part 11,we told you not only how to build and tune profiles for your
various devices, but also the things you must do before and after profiling
to make sure that your devices behave-and continue to behave-the way
their profiles predict. No color management system can compensate for
uncontrolled device variation. We also pointed out the importance of the
environment in which you make your color judgments--correctlighting
is critical.

508

Real World Color Management, Second Edillon

In Pan 111,we laid out the ground rules for parsing color-management
worMows. We examined the color management features of some key applications,andwalked you through theirusesandtheirpotentialpitfalls,but if
weGe done our jobs correctly,we also gave you the vocabularyand insights
necessary to analyze new applications, and to build a color management
worktlow that suits your unique needs.
The rest is up to you!

Appendices

This page lefi blank. ..

Profile Anatomy

The ICC specification prescribes a format for various classes of what are
generically referredto asICCproHes.Theintent of the format is to be both
platform-independent and application-independent. Before the work of
the ICC, profiles were either platform-specific or application-specific.
This appendix provides color geeks with more technical information
about whatvarious ICC profiles contain. Naturally it's not a substitute for
the ICC spec itself, which is available at www.color.org, theweb site of the
International Color Consortium.
The specification is set to circulate as a Draft International Standard,
with possible approval by the end of 2004 as IS0 15076. After approval it's
expected that the complete IS0 version will be released as version 4.2 of
the ICC specification to bring the two into parity.

5l9

Real World Color Management, Second Edition

Profile Structure
All ICC profiles have the following three segments:a profile header, a tag
table, and tagged element data (see Figure A-1).

- 4 bytes
- 12 bytes for
each tag

Tagged
Element

- various sizes

Data

Profile Header
The profileheader contains information that allows searching and sorting
ICC profiles-the header is always exactly 128 bytes. The profile header
contains the following encoded parameters, in order:

P d l e rhc.The tolal profile size in bytes.

CMM Type r i g ~ t ~ rThis
0 . specifies the preferred CMM-effectively the
default CMM. Ilk possible to define no preferred CMM. Signatures must
be registered to avoid conflicts.

Profile rmion. The version of the ICC specification to which the profile
conforms. There are placeholders for major revisions (defined as newly
addedlchanged required tags, necessitating an updated CMM to use
the profile); minor revisions (defined as newly addedlchanged optional
tags that don't require an updated CMM to use the profile): and bug fix
revisions.

Proflk Clam. This defines the class of profile

Device Class
Display
Input
Output

Signature
rnntr
scnr

Ftr

Additional profile classes are:

Pmme Class
DeviceLink
ColorSpace Conversion
Abshact
Named Color

Signature
link
spar
nhst
nrncl

Color Spate slgnatme. There are 25 possible signatures for both device
and non-devjce colorspaces.This includesXYZ,CIELAB, CIELUVYCbCr,
CIEYq RGB, gray, HSV, HLS. CMYK and CMY explicitly, and any custom
color space containing between 2 and 15 channels.

ProRk Connection Space signaturc.There are only two options for the
PCS: C I W or CIELAB.

5l4

Real World Color Management, Second Wltlon

Primary platform signature. Defines the platform on which the profile
was created. The six possibilities are: Apple Computer, Inc.: Microsoft
Corporation;Silicon Graphics, Inc.; SunMicrosystems. lnc.;Taligent. Inc.;
and no primary platform.
Profile flags. There are two flags. One specifies whether the profile is
currently a standalone profile (not embedded) or is embedded in a document. The other specifies whether or not the profile, when embedded, can
be extracted from the document in which it's embedded, and made into
a standalone profile.

Device manufacturer and model signatures. Signatures for devices must

be registered with the KC. Too many devices are registered to list here.
Deviceattrlbutes.This describesthe media associated with the device the
profileapplies to. Options are: reflective ortransparency glossy or matte:
positive or negative: color or black and white. This frequently contains
incorrect data, which doesn't affect the performance of the profile [like
all of the data in h i s segment), but forces us to differentiate profiles using
the profile name rather than the device attributes signature.
Rendering intent.This specifies the default rendering intent tahlein the
profile. Options are percephd, media-relative colorimetric. saturation.
and ICC-absolute colorimetric. See Figure A-2 for the "mother" crossreference of rendering intent tags.
ProHle Creator signature. This identifies the creating manufacturer
of the profile from the device manufacturer signatures list (mentioned
previously).
Profile ID. This is generated with the MDS fingerprinting method-a
value of zero indicates the ID hasn't been calculated. This signature is
currently optional.

Appendix A: Profils Anatomy

n5

Pmfileclass to rendering
intent cross-reference

Tag Table
The tag table is the second and smallest segment of a profile. but is as
vital to a profile as the index file is to a database, or the table of contents
is to a book It contains a tag count listing the total number of tags in the
profile, followed by a sequential list of each tag contained in the profile.
The list refers to each tag using a 4-byte tag signature registered with the
ICC, a 4-byte offset to denote where the data for that tag stafls, and a
4-byte size value to denote how long that tag is.

Tagged Element Data

This third segment of a profile contains the meal and potatoes--white
point information,the profile description that appears in pop-up menus.
rendering intent tables, tone response curves, etc.
One way to envision this is that the tag table is like thecard catalog in a
librry, while the tag elementdata portion ofthe profileis the books-the
analogy works better if you imagine books with no covers, just page after
page with no obvious beginning or end to each book. The tag table tells
the CMS what each tag is, and where it's located in the profile. It's important because the tags can come in any order, and some tags can be any
length.

a6

Real World Color Many)cmcnt, Second Wition

Required Tags
~ i ICC
e profile specification includes both required and optional tags.
The following required tags must be in every ICC profile.

profileDescriptionTag
This contains the profile name that appears onmenus.Tl~efilenameand
profile name are two different things.The profileDescriptionTag is the real
profile name. This is required by the additional profde classes as well.

mediaWhitePolntTag
This is the measurement of media white, in CIEXYZ, which is used in
the calculation of the absolute colorimetric rendering intent. Absolute
colorimetric rendering is computed from the AtoBlTag and BtoAlTag in
conjunction with a mediaWhitePointTag.

chromaticAdaptationTag
Ifthe actual illuminationsourceisnotD50, this tagisrequired.andisused
to convert the actual illumination source to the PCS illuminant (which is
D50).

copyrightTag
The profile copyright is stored as a 7-bit ASCll string, also required by rhe
additional profile classes.

Appendh A: Profile Analomr

n7

lnput Profiles
lnput profiles support grayscale, RGB, and CMYK input devices. In theonc,
they could describe multichannelinput devices as well, although we don't
knowofapackagethat makesCMYKinputprofiles,let alone multi-channel
ones. (Wearen't aware of a packagethat makes monochrome input profiles
either.) So the options are. RGB matrix. RGB table-based, and RGR matrixand-table-based input profiles.

Matrix-Based Proflles
For RGB matrix-based input profiles, only two additional tag types are
required, for a total of 10 tags.The datacontained in these hvo tags is very
small, which is why RGB matrix profiles are typicallyonly a few kilobytes
in size.The additional required tags are:

MatrIxCoIumnTag. The three required MatrixColumn tags are redMahiColumnTag, greenMatrixColumnTag. and blueMatrixColumnTag.They
contain the XYZ tristimulus value of the primary (red, green, or blue,
respectively).CIELAB is not supported in matrix profiles.
TRCTag.There are also threeTone Reproduction Curve (TRC]tags: greenTRCTag, redTRCTag, and blueTRCTag.

Table-Based Profiles
For RGB table-based input profiles, there is onlv one additional tag
required, but it can contain a substantial amount of information when
compared to matrix-based profiles.

AtoBOTag. This table contains device-to-PCS data for the perceptual rendering intent. Only the perceptual rendering intent is required
for table-based input profiles, although other rendering intents are
supported as well. Both &bit and 16-bit precision are supported.The PCS
data may be represented as either ClEXYZ or CIELAB.

Hybrid Profiles
The ICC specification, version 4.0, supports input profiles that are both
matrix- and table-based.

5l8

Real World Color Managomcnt, Smond Edltion

Display Profiles
Monochrome display profiles are supported by the ICC spec with just a
single grayTRC'Ihg, but the much more common color display profiles are,
like input profiles. RGB matrix-based. RGB table-based, and RGB matrixand-table-based profiles.
The required tags for RGB matrix-based display profiles are identical
to those for RGB matrix-based inputprofiles.Table-basedand hybrid display profiles require one additional tag, described next.

BtoAOTag
This contains the PCS-to-device perceptual table. This tag is required to
ensure that display profiles are reversible. Though the perceptual table
is required, this does not mean that perceptual rendering is used: this
tag almosr always contains colorimetric data, so renderings are always
colorimenic--either relative or absolute.

Output Profiles
The ICC specification allows forTRC-only monochrome output profiles,
although we're again hard-pressed to think of a package that makes them.
For all practical purposes, you'll find that output profiles are generally
RGB, CMYK, and (much more rarely) grayscale table-based profiles. The
ICC spec currently supports up to 15-channel profiles (in version 4.0.0).
but the common CMMs support only up to 8-channel output profiles.
Here are the required tags.

AtoBTag and BtoATag

There are sixpossible rags representing both the rendering intent and the
direction (to thePCS or from the PCS). It's easyto get &em confused-one
way to remember is to think'% to P" and "P to A" instead of 'ato B" and
"B to A," where "P" is the PCS and "A" is the device.

Appendix k Proflle Anatomy

b

The AtoBOTag is device-to-PCS perceptual rendering

The AtoBlTkg is device-to-PCS colorimetric rendering

a9

r The AmBXag is device-to-PCS saruration rendering

t The BtoAOTag is PCS lo device perceptual rendering
t The BtoAlTag is PCS to device colorimetric rendering
b

The BtoA2Tag is PCS to device saturation rendering

Notice that there's only a single colorimetric table. See "mediaWhitePoint'kg" formore information. For a description ofhowrendering intents
are actually applied in conversions, see the sidebar "Rendering Intents and
Conversions."

gamutTag
This able contains PCS values o n the input side, a n d on the output side
a single value, either 1 or 0. A value of 1 means the PCS value is out-ofgamut. a n d a value of 0 means the PCS color is in-gamut.

Rendering Intents and Conversions

When w r perform conversions.
most applications only let us
choose a single rendering intent,
yet both source and destination
profiles contain rendering intents
that can apply to the conversion.
Here's how itworks.
If both source and destination
profiles contain the rendering
intent you specified. it's used for
both the source-to-PCSand PCSto-destination conversions.
However, matrix-based profiles
only contain a single rendering
intent. usually Ins we noted earlier in this appendix]relative colorimetric, even if it's often labeled
as perceptual. Conversion between matrixprofilesand the PCS

can only use the renderlng intent

in the matrix profile, so if either
source or destination profile is a
matrix, theconversion between it
and the PCS uses the matrix rendering intent. If the other protile
contains the specfied rendering
intent, that intent is used in the
conversion between the PCS and
that profile.
If no rendering intent is specified, either because the uspr interface doesn't allow it nr because
you chose "Automatic." the destination profile's detault rendering
intent is used for both source-toPCS and PCS-to-destination, if
the source profile supports that
rendering intent.

For rxample, it you request a

conversion horn a matrix-based
editing space profile such as
Adobe RGB I19981 to a CMYK
output profile using Perceptual rendering, the source-toPCS conversion will use relatlve
colorimetric rendering because
that's the only intent Adobe RGB
11998) contains, and the PCS-todestination conversion will use
percepmal rendering.
This is why rendering intents
sometimes appear to have no
effect-if you convert from one
matrix profile to another, the
conversion can only use the
rendering intents contained in
the profiles.

540

Real World Color Management, Second Edition

Additional Profile Classes

Besides device profiles (including"virtual device" profiles such as Adobe
KGB (1998).Colormatch KGB and sRGB),the ICC specification allows for
four additional classes of profileThe additional profile classes are: DeviceLink, Color Space Conversion, Abstract, and Named Color.

DeviceLink
DeviceLink profiles allow for direct device-to-device conversionsthey're essentially profiles that contain a conversion from one profile to
another. While they typically contain only a single source and destination profile. any number of device and nondevice spaces in series can
be combined in a DeviceLink profile, though the first and last profiles
in the chain must represent device spaces. Four tags are required for
Device-Link profiles: profileDescriptionTag,AtoBOTag. profileSequenceDescTag, and copyrightTag.
DeviceLinks may appear to break the rule that you always need two
profiles to make a conversion, but under the hood. a devicelink contains
at least two profiles even though it's a single file.

ProflkSequenceDescTag.This describes the sequence of the profiles

contained in the link.

ColorSpace Conversion
ColorSpace conversion profiles are used bv CMMs to convert between
different device-independent color spaces, such as between ClELAB and
CIELUV. These profiles can be embedded in image+if you have a LAB
image that isn't based on D50, you need a suitable ColorSpace Conversion
profile embedded in it.

Abstract
Abstract profiles are intended to perform image editing by transforming
color data within the PCS. In practice they are rarely, If ever, used. We
know of only two packages that create them: Kodak Custom Color KC.
and ITEC ColorBlind Edit.

Appendix A: Protlle Anatomy

591

Named Color
Named color profiles (often referred to as NCPs) are used to support
named color systems such as Pantone, Focoltone, or vendor-specific
custom colors. The requirement is for the named colors to be associated with a device-independent (typically LAB1 value. The optional, but
most practical aspect of NCPs is to reference each named color to device
values, thereby ensuring the best possible reproduction of a named color
on a specific device.

This page lefi blank. ..

Appendix B
Workflow Templates

Some time after the topic of workflow templates came up. Chris stumbled
upon a Mac OS X-only application called OmniGraffle. Developed by
the OmniGroup (www.omnigroup.com).OmniGraffle is used primarily
for diagramming and charting (see Figure 8-11,It's an amazingly deep
application with too many features to list here, but-in addition to
supporting layers, transparency, and tons of other stuff-the one that really rings our chimes is its unique auto-layout feature.
The more complex the workflow, the more useful this auto-layout
feature canbe. It usesthe established relationships between yourworkflow
components to alter the layout, using settings in Layout Info. The autolayout results may not be the prettiest thing you've seen, but they let you
see your workflow in new and often surprising ways.
One of Chris' more endearing qualities is his boundless energy and enthusiasm, and he immediately hatched grandiose schemes for exquisitely diagramming all kinds of color management workflows. When Bruce
pointed out that fully diagramming even the relatively simple workflow
weused to produce this bookwould demand an eight-page foldout, Chris
finally relented, realizing just what a rabbit hole it could turn into. So
what we're providingin this appendixis simply a taste that diagrams some
specific workflow situations.

594

Real World Color Managemml, k o n d Wilion

The more complex your workflow, the more important it is to plan,

and to figure out the interactions between departments and customers,
whether or not they are using color management. We find OmniGraMe
incredibly useful in assisting the diagramming of color-managed
workllows, and recommend it to anyone who needs to sit down and
figure out complex production scenarios.

A screen capture ofan early attempt at diagrammingAdobds Missingprofile

and Profile Mismatch warnings.

We've used the same conventions in a l l the templates containedin this

appendix--they're shown in Figure B-2-to distinguish between user actions, user decisions, and application options.

Appendix B: Workflow Tcmpkta

Flwn E P

action

Legend

needs to

user must

dialogbox option

based soli
olor

The four stages of

color management

rn

&fine cok
mpnnine

convert color
~rkinglediti,
space

Lonuem to
proofer space

5%

536

Real World Color Management, kcond Wition

Flgurr 0-4

A rational workflow for

content creators

(define sources,
mnal

"-.rr., r .

.oferprofil
from printer c
service burea~

with t a ~ e files*
d

'and an explicit nofationof the profiles to be

assumedforfiles that don't support
embedded profiles

A sadly typical wor161ow

for content creators

guess for outp

rim proof

,101

correct

The itemtiveproof-then-color-correctcycle ends when:

1. The proofing budget dries up.

2. Yougetfed up (orsomeonegets injured or killed).

3. The proofs come back looking the way you want.

548

Real World Color Management, Second Edltion

ngum
hditional and ideal
workflows for print
providers

_r

..

....

proofs based on

'embedded or assumed profile

well documented

Using Adobe's
Missing Profile and
Profile Mismatch
warnings

530

Real World Color Ma~geIIIent,kcond Edltlon

Figure B 4

Dealing with CMYK Mystery Meat

The image is effectively

color managed because
you're evaluating it in
a calibrated environment. The numbers may
not produce the desired
color appearance,but
they produce a known,
rather than a random,
color appearance.

Open m
'hotoshoo

The image is "Oficially"

color managed because
it's tagged with a
profile.
The gray areas represent
bad options. You have
no idea what this image
is, so you aren't in a
position to assign a
profile without looking
at the image. (And ifyou
want tradition, well, we
can't help youJ

Missing
Profile

I
I

I
I

wrr E9

-OaUng with RGB Mystery Meat

+I

Color M a ~ g e

h k a t the image

Go to Image > Mode >

Assign h f k and select

another pmfk

The image is effectively

color managed because
you're eunluating it in
a calibrated enuimnment. The numbers may
not produce the desired
color appearance, but
they produce a known,
rather than a random,
color appearnnce.
The image is "oofficially"

color managed because

it's tagged with a
profile.
Thegray areas represent
bad options. YOU h u e
no idea what this image
is, so you aren't in a
position to assign a
profile without looking
at the image. (Andifyou
want tradition, well, we
can't help you.)

Earlybinding

Separate1
conuertfor
oufput

1
you renlly like it, still using
monitor as guide.

Real World Color M a n a m , kcond Edition

mure B-10

Mac OS 10.3 raster printing

AE

absolute colorimetric

(Pronounced "delta-E.") Generically, the computed degree to

which two colors appear to match. Usually calculated in CIELAB,
but other color-differencing systems exist.
See colorimehic intent.

achromatic The attribute of light from asurface or light source that relates to the
amount of lightspecifically,its brigh mess. Contrast with chromatic.
additive primaries Three light sourceschosensothatthey can reproduce other colors by
addition of wavelengths. One primary contributes long wavelengths
(red), another the medium wavelengths (green),and the other the
short wavelengths (blue). See also subfractiveprimaries.

AM screening Also known as a halftone In printing, a method of doing screening

where the frequency of the dots (the screenfrequency, or the number of dots per inch) does notvarythroughout the screen. Instead,
the effect of tones is created by varying the size of the dots [hence
"amplitude modulation"). Contrast with FMscreening.
anisotropy The phenomenon whereby a printer produces slightly different color depending on the orientation of the paper movement
through the printer.

5%

Rwl World Color kbnagcmrnf, Second Edlllon

A small opening. Color-matchingexperimentsusually involve the

test subject viewing colors through a small aperture, and measurement instruments also view samples through a small aperture.

artifact Something added artificially. In general, any undesirable effect

that is visible in a displaved or printed image, but was artificially
introduced bythe equipment or software. Examples are banding,
compression artifacts, moire patterns, etc.
assigning a profile The term used in Adobe applications for the act of tagghga doctimenrwith a profile. See the sidebar on page 282.

anumed profile The profile that acts as the sorcrceprofileforuntagged documents

(orones that theactiveapplication sees asuntagged).See thesidebar on page 282.

banding

An artifactwhere there are noticeable jumps from one tone level

to the next.

black generation In four-color printing, the method used to generate the black,
or K, channel from the color channels. The two main kinds of
black generation are UCRand GCR.
black level On a monitor, the setttng that controls the base black of the
display. On a CRT; this is just the brightness (or offset) control on
the monitor.
black point

The density, and occasionally the color, of the darkest black

reproducible by a device. For a printer, this is the density and
neutral color balance of the darkest black achievable using the
printing inks; for a monitorthis is thedensity ofthemonitorwhen
displaying black, and can be adjusted by setting the hlnch-Inlel.

black point compensation

A setting in Adobe Photoshop that makessure that the blackpoint

in thesourceprofileismapped to the black point of the destincmtion
profile,and the rest of the tonal range is scaled accordingly.

blackbody radiator

A light source whose photonsare entirely due to heat energy given

off by the source. According to the laws of physics, there's a very
predictablecorrelation between the temperature of the blackbody
and the colorofthelight-lowertemperaturesare red, higher temperatures are white, and the highest temperatures are blue.

brightnen The perceived response to light intensity.Thisresponse (in a human

observer) is rton-linear.
calibration

Modifying or adjusting the behavior of a device (such as a color

reproduction device or a measurement instrument) to a desired
state (oftena factory specification, or some state that helps simulate
some other device).
The type of metamerism where two color samples match to a humanobserver,but not to acertaincamera, orviceversa (the human
does not see a match, but the camera does).

candela The unit of lrrminorrs intensit).

cathode ray lube See CRT.

CCD Charge-Coupled Device. Alight-sensitive microchip, usedin scanners and digital cameras as the basis for capturing a digital image.
charattcrization See profiling.
chroma

Technical word for saturation

chromatic The attributes of light from a surface or light source related to the
wavelength composition of the light-specifically, its hzre and
saturation. Contrast with achromatic. The chromatic attributes
describe the properties of the photons in the light. while the
achromaticcomponent describes the quantity of photons in the light.
chromatic adaptation See colorconstanc~?

CIE Commission Internationale d'Eclaimge (International Commission

on Illumination). An international association of color scientists
that has assembled many of the standards used as rhe basis for
colorimeny.

CIE LAB (Also known as L*a*b*.] One of the two main color spaces
proposed by the CIEto attempt a perceptually uniformcolorspace.
L* is the lightness value, a* is the red-green opponency, and b* is
the blue-)'ellowopponency. CIE LAB is one of the two color spaces
used as a PCSin ICC-based color management.

536

R u l World Color Management, Second Edition

ClE LCh

(Also known as LCH, for Lightness, Chroma, Hue). A derivative

of CIE LAB that uses polar coordinates rather than rectangular
ones. often employed as auser interface for manipulating deviceindependent data. The LCh space is cylindrical: lightness runs
along the venical axis from white (100)to black (O), chroma (saturation) is represented by the distance from the central axis, and
hue is represented by the angle around the cylinder, with reds
around 0 degrees, yellows around 90 degrees, greens around 180
degrees. and blues around 270 degrees.

CIE LUV (Alsoknown as L*uav'.) One of the two main color spaces proposed
by the CIEto attempt a percepnrallv rmifonn color space.

CIE rry diagram Wso knouu a5 the CTE xy chromaticity chan.1 The horseshoeshaped
diagram representing the relationships of the colors in the CIE
system.

CIE XYZ Shorthand forthe CIEXM.colorspnce.Thisdefines colors in terms

of three theoretical prinzaries. XY, and Z. that are based on thc CIE
research into human color response (the CIE Standard Obsener).
XYZ is not perceptua1l.v uniform and therefore can't be used for
computingcolor distance. XYZis oneof the two color spaces used
as a possible P a i n ICC-based color management.
tllek.balancing

Cry-balnncingan image by clicking on an areain the image that

you know should be neutral.

CMM Color Management Module. Some people know it as an abbreviation for Color Matching Method, or Color Manipulation Model.
In any case, a CMM is a drop-in component that provides the
"engine"for profile-to-profile conversions. It defines how colors are
computed using the sample points in the profilesas guidelines.

CMS ColorManagementSystem. Sottwarededicatedto handlingdeviceto-device conversion of colors. The ICC-based model for a CMS
consists of four components: a PCS. device profiles, a CMM, and a
set of rendering intents.

tolorant Something. such as a dye or pigment, or phospl~orsona monitor,

that produces color on some other medium.

<dot bar A strip of colorpatches, usually printed in the margin of a printed

page, and used for process control. A color bar is by definition
Mice-dependent and should be created in the color model of the
printing device. Color bars contain at least the maximum (100%)
of each primary and complenzenrnry color. Many also contain 50%
coverage m a s of primriaand their overprints, as well as neutrals.

color constancy The tendency of the visual system to consider the color of an
object to be constant despite changing viewing conditions. If all
the colors in a scene change in the same way. the eye tends ro attribute this to a change in illuminar~r,and discounts the change.

color gamut See gamrrr.

color mnagemmt system See CMS.
color-mixing fundion In avisualsystem or color-measurementdevice (suchas a camera. or
colonmeter),the unique mixture of red, green, blue responses to
different color samples. Visual systems and measurement devices
don't really have a gnmztc they have a color-mixing function.

d o r model A general system for assigning numbers to colors. An example

would be the RGB color model, where each color is defined In
terms of three numbers: the first the amount of some (as yet
undefined) red primary, the second a green primary, and the third
a blue primary. Contrast with mlorspnce.

color rpacc An Instance of a color modelin which every color is represented

by a specific point in space, and thus has a specific set of three
ormore numbers that describe it. An example would be the RGB
space of a particular monitor, where a certain color has a specific
set of three numbers representing the amounts of the red, p e n .
and blue phosphors needed to produce that color. Note that
there can be many color spaces that use the same color model
(for example, every monitor essentially has its own specific color
space, but all use the RGB color model).

538

Real World Color Management, Second Edition

color temperatura A description of the color of light in terms of the temperature of

the light source, in kelvins. Lowertemperatures are redder, higher
temperatures are white, and the highest temperatures are bluer.
Technically we should only refer to the exact color temperature
of a blackbody radiator. If the light source is not a true blackbody
radiator,then it's more correct tospeakof the correlated color remperflhtw of the fight source.
colorant

Somethingused to produce color,such as the (yesin inksor toner,

or the phosphorsin a monitor.

colorlmeter A device for computing colorimetry (color matches) from

measurements of a surface.
~ l ~ i m o t rintent
ic

The two rendering intents that try to preserve colorimehy of colors--colors are mapped to an exact match wherever possible, and
where not possible (because the color is out of the target ganr1r0.
the color is mapped to its nearest equivalent. In most cases, rhis
conversion shodd include a remapping of the white point so that
this colorimetricmatch is dative to the target white point-this is
known as re[atiw colorimetric. In some cases (in certain stages of
proofing), the colorimetric match should be absolut+the colors
should be converted as if the match were being done relative to the
source device's white point-this is h o w n as nlxosolritecolorin1eh.i~.

colorimetry The science ofpredicting color matches based on ever-growing research into typical human colorvision.Much colorimetry is based
on the work of the CIE.
ColorSync Apple's implementation of ICC-based color management. On
Macintosh computers, the components in the Mac OS that
handle such things as making profiles and CMMs available to
programs that need to convert colors.
complementary colon Two colors that make gray (orwhite) when com bined. For example,
red and cyanare complementary colors.
cone$ See photowceprors.

continuous tone A device that can represent many tonal values for each unit pixel.
A laser printer is not continuous tone as it can only lay down a
printer dot or leave it blank-to produce tones, it needs to use
the trick of screening. A monitor is continuous tone as every
displayed pixel can represent tone levels by varying the intensities
of the red, green, and blue light.

contone Shorthand for continuotrs tone.

contrast The difference between the lightest and darkest areas of an image.

concbtcd cdor temperature See color temperonire.

CRD Color Rendering Dictionary In PostScript color management, the
CRD is the equivalent ofthe destination profile, and either resides
in the RIPor is sent to it at print time.

CRI

Color Rendering Index. A measure of how well colors are rendered

by different lighting conditions as compared to a standard light
source. CRI values range from 1 to 100. whereby 100 means that
all colors that match under the standard would also match under
the measured lighting conditions.
Cathode Ray Tube. Themost common type of computer monitor,
consisting of tube with a source of electrons (a cathode) at one
end, and a flattened end coated with phosphors that glow when
excited by the electrons. (See also LCD.)
Color Space Amy. In Postscript color management, the CSA is
the equivalent of an ICC source proflle that is downloaded to a
PostScript tUP together with the documents it describes.

curing t h e T i e needed after printing by some colorants-for example, the

inks in inkjet printers-for the printed image to arrive at a stable
appearance.
Oneofthe CIE standard illuminntlts. D50 is a specification of daylight with a correlated color tetnpernnireof 5000 K.
Oneofthe CIE standard illrtminnnrs.D65 is aspecificationof daylight with a correlated color rernperntureof 6504 K.This is thestandard white poinrthat we recommend for monitor cnlibrntio~~.

540

Rwl World Color Management, Second Edlllon

delta-e See AE.

densitometa A device for computing densifyfrommeasurements of a slwface.
The degree to which a surface absorbs light. Density is a logurithmic
value--a densityof3.OD is ten times darker than a density of2.0D.
and a hundred times darker than a densityof 1.OD.

destination profne In a color conversion, the profile that defines how to convert colors
from the profile connection space (PCS)to the target color space.
See also soitrceprofile.

device profile See proj?le

devlcc-dependent The properly of a color model whereby the exact meaning of a
set of numbers depends on the specific device. For example, RGB
is a device-dependent color model because a specific set of RGB
numbers (say. 10.20.301 will "mean" a different thing-i.e.. it will
produce a differentcolor--depending on what RGB device you are
using. See also deirice-independent.
The property of a color model where the exaa meaning of a set of
numbers is unambiguous and does not depend on any specific device. For example, LAB is device-independent because a specific
set of LAB numbers "means" the same thing-i.e., it represents the
same color-regardless of any device. See also device-dependent.

display profile Also known as a monitorprofile. A profile for adisplay device such
as a computer monitor.
dithering (11 screening
(2) Any technique that simulates colors or tones by intentionally
printing or displaying dots of various primary colors in various
patterns to give the illusion of a larger set of colors. A halftone is
a type of dither (called an ordered dither) where the dot pattern
is uniform and uses the primaq ink colors together with paper
white to give the illusion of many colors. Another exampleof dithering can be seen when viewing a full-color image on a monitor
in 256-color mode-the illusion of additional colors is created by
displaying the limited palette of 256 primary colors in various dot
patterns.

dot galn The growth of halfronedots during print reproduction. The main
cause of dot gain is thespreading of ink as it hits paper, bulslight
dot gain can also be introduced during imaging to film.

dpi Dob Per Inch. Usually, but not always, used when referring to the
resolution of printers. Often incorrectly used as svnonynous with
ppi, or sometimes spi.
dyc

A soluble colorant. Dye-based inks exhibit color drift over time,

especially when exposed to light and other elements. Often used
in inkjet printers. Contrast with pigment.

dynamic range A range from brightest white to darkest darkas measured in rlensiiy.
The dynamic range of measurement devices (such as scanners.
cameras, or densitometers) describes the distance between the
darkest black the device can measure before it is unable to detect
differences in brightness, and the brightest white it can measure
without overloading. Also applied to media (for example, prints or
bansparencies) and images, to describe the range from the darkest black to the brightest white.

early-binding A workflowstrategythat convensall colors to the final output space

as soon as possible; for example, converting all scans immediately
to printer CMYK.

editing space A color space intended specifically for editing of color values. An

RGB editingspace should ideally (1) be gra~~-b~fl(~ncerl:

12) be percepnaally uniform; and (31 have a garnut large enough to contain
the values being edited.

elrtrophotostatic The class of printing devices that use a laser to set a static charge
on a point-by-point basis, to control where toner particles stick

embedding r profile Saving a profile in a document file. The profile defines the source
profileused when convertingthecolorvaluesin the file toany other
color space. The embedded profile effectively provides the meanings of the color numbers in the Me. See the sidebar on page 282.

emissive Having the property of emitting (givingom light. A ligh tbulb and
a computer monitor are emissive.

544

Real World Color Managemml, Second Willon

encoding

(Ij in a color space, a specific set of numerical values that can

represent a specific color. For example, in a monitor's RGB space,

each encoding is a specific triplet of eight-bit numerical values,
and approximately16.7 million encodings are available.
(21The act of assigningspecific numerical values to colors.

crmr diffusion

A form of FMscreening

and re-emission
fluorescence Theabsorption of ligh~energyat one rrnr~elengrl,
at a different wavelength.
In printing,a method of screening in which the dots are all the same
size, but the effect of tones is created by varying the frequencyof the
dots in agiven area (hence"frequency modulation"). Contrast with
AA4screening.
The area at the center of the retirm that is populated predominantly by cones,This is where colorvision occurs and where visual
acuity (sharpness) is highest.

foveal vision The viewing condition where the image is focused on the fowaof
the viewer, and thus produces the best acuity and color vision.

gamma

(1)Thedegree towhichadeviceor colorspnceis non-lirrenrin tonal

behavior, represented as the exponent of a power function.

(2) In CRTdisplaysystems. the relationship between input voltage
and output lrrrninance.

(3)In colorspaces, the mapping of tonal values to perceived brightness.A gammavalue of around22 is generally considered percepnrally uniform.

gamma c u m A simple example of a tone reprodnoion crrn~typicalof most devices. This is a simple non-linearcurve.
gamut

The range of colors aod densiy values reproducible on some

output device such as a printer or monitor. This is sometimes split
into the colorgarnu-the range of colors limited by the primaries
used-and the rl)~n(4n1icmnge--therangeof brighrnesslevels from
the darkest black to the brightest white of the device.

The remapping of color and tone values from a large gnmrrr to a

smaller one.

gamut mapping The remapping of color and tone values from one color sprrce to
another. If the destination color space has a smaller ganzrrt, this
remapping will require gamut compression.

GCR Graycomponent Rep1acement.A methodof blackgenemrion that

replaces an amount of CMY with the equivalent amount of K.

GDI

Graphics Device Interface. Microsoft's name for the display engine

in Windows.
See meflsurement geomehy.
A colorspacein which any neutral pixel has equal R, G, and B values is said to be grq-balnnced,

gray-balancing

(1) Correcting RGB values in an image so that neutral grays have

balanced R, G, and B values:
(2) Adjusting the behavior of a device to achieve good reproduction
of neutrals.

halftone In printing, the most commonly used method for screening. The
effect oftones is created bvdividingthe image into equally spaced
halftone cells, each of which is filled with a dot of a known size-larger dots create darker tones: smaller dots create lighter tones.
high-bit

A general term for anything higher than the minimum 8 bits per
channel (256 tone levels).A high-bit file is a 16-bit file. A high-bit
scannermay be a lo-, 12.. 14-. or 16-bit scanner.

HLS An adaptation of the RGB color model stands for hrre, lightness,
sohrration.\'Vhen lightness is at maximum, the color is white.
HSB An adaptation of the RGB color model, stands for hrre. sandration,
brighmess. Common in many color-managed applications.Similar
to HLSexcept that when brightness is at maximum, the color isat
its brightest.

H N An adaptation of the RGB color model, stands for hue, scrt~rration,

value, and is synonymous with HSB.

544

Real World Color Management, k w n d Edition

hubris Overbearing pride or presumption,arrogance. Acondition that leads

thoseaMicted to attempt towrite books oncolor management.
hue The property of the light from a surface or light source by which
we perceive its dominant ruauelength.

ICC InternationalColor Consortium.Aconsorfium of color-related companies that have cooperated to standardize profileformatsand procedures so that programs and operating systems can work together.

ICM Image Color Management. The implementation of the ICC profile

specification in Microsoft Windows.
iflumlnant A light source defined spectrall+n other words, by the relative
amount of energy at each point in the uisible specrnlm.
Infrared

(IR)The non-visible region of the electromagnetic specrrrtrrr just

below the low-energy, or red end. of the uisiblespectrrmr.

Input profile A profile for an input device such as a camera or digital camera.
Not to be confused with a solme prople.
intensiw The sheer amount oflight from a surface or light source, ~4thout
regard to how the observer perceives it; the number of plloto~~sin
the light. Compare to briglirness. which is the perception of the
light's intensity.
Interpolation The computation of unknown values that fall between known
(usually measured or sampled) values.

IR See infrared.
IS0 The International Organization for Standardizati0n.h organization that coordinates networks of standards bodies around the
world. (IS0 is not an acronym, but rather derived from the Greek
"isos," meaning "equal.")

IT8

One ofa family of targetsused for cnlihrntionand profiling of scanners and printers.

kelvins

(K)The unit physicistsuse to describe temperature, with the scale

starting at absolute zerc-the temperature at which all atomic
activity stops.

LAB See CIE LAB.

A workflow strategy that delaysconversions to final output as long
as possible, perhaps even doing this conversion in the RIP.

Liquid Crystal Display. The second most common type of computer monitor, consisting of two layers of polarized plexiglass between which are liquid crystals that change shape in response to
electrical currents. (See also CRT.)
Relative briglrtness.The brightness of a surface or light source relative to some absolute white reference.

Asimplerelationsl~ipbetween
stimulusandresponse, whereby (for
example) doubling the stimulus produces double the response.
The human sensory system is predominantly non-linear.
(1) cnlibration

(2) The act of making a device linear (which is a specific form of

calibration).
liquid eryrtal display

See LCD.

logarithmic A method of arranging numbers on ascale that compresses as the

numbers get larger. This is convenient when the same scale has to
represent bothvety small and very large numbers. (Seethe sidebar
on page 33.1
lookup table

(LUn A table that allows input values to be looked up and replaced by corresponding output values. For example, an RGB-toXYZ lookup table would let you (or a color management system1
look up an RGB triplet and find the equivalent XIZtriplet.
Lines Per Inch. This always refers to the resolution, or screen
fkqrrency, of a halfone. Not to be confused with clpi, which refers
ro theresolution of the microdots used by a digital printer, not the
resolution of halftone dom.

luminance The amount of light energy given off by a light source, independent ofthe response characteristics of the viewer. More precisely,
luminance is the lrrminorcs intensify per unit area of the lightemitting surface. (For a more detailed description of luminance,
luminous~ux,
and illuminance, see sidebar on page 213.)

546

Real World Color Manag.rmnt, Second Edition

luminous intensiQ

LUT

The amount of light energy given off over time from a point light
source. This is measured in cnndelas. (See also lriminnnce, and
sidebar on page 213.)
Pronounced "luht"1 See lookup fable.

LUV See CIE L W.

matrix

An ordered set of numbers listed as a rectangular grid. A 3x3 (or

"3-by-3") matrixcan beused to convert from one 3-channel color
space to another-forexample, itcanconvert any triplet ofnumben
representing the RCB values of a color, to another triplet of numhers representing the CIEXYZvalues for the color. All profiles use
either a 3x3 matrix, ora lookup table ILUn, to convert values.

measurement geometry

The physical geomehy of a measurement device that defines the

relative angles of the light source, measured surface, and detector.
These are important when dealing with such effects as glare and
surface texture on the measurements.

memory color

A color-such as a skin tone, grass green, or sky blue--that has a

special association for the viewer, and is therefore more important
to get right than other colors for which the viewer has no memory
reference.

metameric inkleiye A slight, but common misuse of the term mefamerisni.When

someone says that a printer has "metameric inks." he generally
mean that the inks are vulnerable to wide shifts in apparent color
depending on lighting conditions.
The condition whete two color samples, with different spectral
properties, produce the same color sensation under certain viewing conditions, and a different color sensation under different
conditions.
lhro color samples that exhibit n~etamerism.
In other words, they
match under some, but not all. viewing conditions.
monitor profile d s o known as a di~plfljrprofik.
A profilefor a computer monitor.
monitor RGB

The color spme represented by the display profile.

monochromatic light Light consisting ol' photons all of the same energy level or umwlengtl~An example would be a laser.
nanomder

(nm)One biiionth of a meter.

non-linear

A complex relationship between stimulus and response, where

(for example) increasing the stimulus produces less and less
response. The human sensory system is predominantly non-line=.

null transform When the sorrrcepmfileanddestination profilearethe same, CMMs

ensure no conversion occurs.
observer metamerlsm The type of metamerism where two color samples match to one
observer, hut not to another observer. Examples ofthisare scanner
meramerism and camera metamerism.
The theory of color vision (now well verified by experiments)
that we experience color in terms of various opponent pairs: redgreen, blue-red, and light-dark.This has now been reconciled with
rrichromacywith research that shows that one laver of the retina
has the three phororeceptors, and the next layer seems to son this
information into the opponent pairs.

ordered dither
output proflk

A halftone.
A profilefor an output device such as a printer or proofingdevice.
Not to be confused with a destination profile.

Profile Connection Space. The color spaceused as the intermediate form for conversions from one profile to another. In the ICC
specification, the PCS is either CIEXYZor LAB.

perceptual intent

The renderingintmf that tries to preserve the perceptual relationships in an image, even if this means remapping all colors both
in-gamut and out-of-gamut. This is usually, but not always, the
preferred rendering intent for images that contain many out-ofgamut colors, but if all colors are in-gamut for the target color
space, relati~wcolorimerric may be preferred.
The property of a mlorspace whereby distances between points
in the space correspond well to perceived distances between the
colors they represent. Close colorsare represented by close points;
different colors are represented by distant points.

548

Real World Cdor Managmnenl, kcond Wllion

A substance that absorbs energy and gives off photons, usually of

The photons are not the result of heat

a very specific ~uavelengrl~.
energy (in otherwords, this is not bhckbodyrcliation), but rather
ofspecific properfiesof the atoms in the substance. Primeexamples
are the red, green, and blue phosphors in a CRTmonitor.

photon A hmdamentalpacket of electromagnetic energy traveling through

space. In some ways photons behave like particles and in other
ways photons behave like waves.
Light-sensitive nerve cellsin the retina. (Sometimescalled simply receptors.) The two main types of photoreceptorsare rodsand cones,
so called because of their shape. Rods are responsible for low-light
vision, and cones for daythevision.There are three types ofcones.
each sensitive to a different part of the visible spectrum.
An insoluble coloranl. Pigment-based inks greater stability over
time, even when exposed to light (including UV) and other
elements, than dye-based inks. Used in press inks, toners, and
outdoor printing. Contrast with dye.

polarization

Filtering light in such a way that only light waves of a particular

orientation pass. Polarization is a key part of how LCDmonitors
work Many measurement instruments also incorporate polarizing filters to reduce glare issues with glossv targets.
See banding.
PixelsPer 1nch.Thisisusuallyused when referring to theresolution
of a digital image. Contrast with dpi.

primaries Shorthand for color primaries. A set of colors that, used together
in controlled amounts, can reproduce all other colors. See also
additive primaries and subtractir~primaries.
pIMeN color

The combination of three or more primaries, in various amounts,

to simulate the reproduction of full color. In printing, often used
as a shorthand for four-color process-printing using cyan.
magenta, yellow and black (CMYKl inks.

pmflk A file that containsenough informationto let a (1MSconvert colors

into or out of a specific color space. This may be a device's color
space-in which we would call it a device profile, with subcategories inpzrtprofile, outprrtprofile. and displayproJle(for input, output, and display devices respectively):or an abstract color space
such as a ruorkingspacelike Adobe RGB (1998).

pmflk mlsmatch The condition that arises when you open a file that contains an
embedded profile other than the opening application's assumed
profileor umrkingspnce.
The act of creating a profile by measuring the current state of the
device. Sometimes also known as cl~aracterimtion.
Apple's name for the display engine in Mac OS X.
Apple's name for the display engine in Mac OS 9 and earlier. See
also Quartz
( R ) The ratio of light shone onto a surface to the light reflected

back to a detector. This is the measurement used by a reflection

densitometerto compute density
Having the property of reflecting light off a surface. Asheet of paper and ink have reflective components. See also tmnsmissiwe.
See colorimetric inrmt.
The setting that tells the color management system how to handle
the issue of convertingcolor between color spaces when going from
a larger gamut to a smaller one.The ICCspecifies four rendering intents: percephml. sakiration, and two types of colorimehic intents.

retina The layer of nerve cells lining the back of the eye and receptive to
light.
RIP Raster Image Processor. The RIP is either part of a digital printer
or a separate computer attached to the printer. The job of the RIP
is to convert the page image from vector form (usually expressed
in Postscript) to the raster form needed by the marking engine of
the printer. Some color-management processingand conversions
can also happen in the RIP.

550

Real World Color Management, Second Edition

See photoreceptors.

saturation The property of the light from a surface or Light source by which
we perceive the purity of the light-how much does the light
contain photons of only a certain wnuelengtl~(highly saturated)
or a mixture of many wavelengths [less saturated).
saturation intent

The rendering intent that tries to preserve the saturation properties of colors as much as possible, even at the expense of lzue
accuracy or perceptual relationships. This is usually, but not
always, the preferred rendering intent for information graphics
such as graphs or maps, where it's desirable to maintain saturated vivid colors, or where the saturation of different regions is
designed to convey information.

scanner metamerism The type of metamerism where two color samples match to a human obsewer, but not to a certain scanner, or vice versa (the human does not see a match, but the scanner does).
screen angle

In a halflone, the angle of the lines that f o m the halftone screen.

relative to horizontal.

screen frequency In a hnlfrone, the number of halftone cells, or "lines" per inch or
centimeter.
screening In printing,the method used to simulate different tones and tints of
ink bv breaking the inkinto dots of controlled size and frequency.
simultaneous contrast The effect where the perception of a color is affected by other colors seen simultaneouslv in the same field of view. [Compare to
successive corrtrnsl.)
smart monitor

A monitor with a direct connection to the host computer [rather

than just an analog connection via the video card) that lets the
computer control the monitor's analog controls directly.

soft-proofing Using your monitor as a proofing device--displayinga simulation

of how a document will appear when printed.
source profile Inacolor conversion, the profilethat defines how to convert colors
from the first colorspncetothe profileconnection space (PCS).See
also destination profile.

Space

(1) The final frontier

(2) In color and color management, shorthand fora colorspace.We

often refer to the "RGB space'' of a monitor, or the "CMYKspace"

ofaprinter, when referring to the specific definitions of the colors
reproducible on that device in terms of its pritnnries.

spectral data A sampling of spectral powerdistriburion, limited to the number

ofbands supported by the measurement device. Colorimem'cand
densirydata can be derived from spectral data.
spectral power distribution The amount of each runvelength contained in a sample of emitted
or reflected light.
sp.mophotometer

A device for measuring the spectral properties of a surface-the

degree to which the surfacereflects light in differentregions across
the specnum.

The fullrangeofpossibleenergylev& (tmvelengthslof plzotons.The

risible spectrum refers to the range of energy levels (wavelengths)
visible to the eye.

mi

SamplesPerInch.This is usually used when referring to the resolution of a digital camera, scanner. or a scan. See also dpi.

rpol color A non-process color, usually based on a named color system,

printed on a separate plate on a press.

BGB A "standard default" RGB color spaceintended for images on the

Internet.
Sfandard llluminant One of the standardized list of illrrminnnts defined by the CIE as
representative of typical light sources. The best known of these
are the A illuminant (incandescent bulb), and the 050 and D65
daylight illuminants.
Standard Obsamr The definition formalized by the CEof the visual response of a
typical human observer.
stochastk rnocnlng A form of FM screenitrg
strip reader

An instrument designed to read a row or several rows ofcolor patches at a time. You feed the target, in strips, into the instnunent.

552

Real World C o b Managernenl, kcond Edition

Three pigmenrs chosen that can reproduce other colors by the

subhactionofrua~relengthshmwhite.Onepnmary(cyan) subtracts
longwavelengths,another (magenta)subtracts themedium wavelengths, and the other [yellow) subtracts the short wavelengths.
See also arlrliti~wprinraries.

~cccuivc
contrast Theeffectwherebythe perception ofacolor isaffected by other colors seen immediately before. (Compare to simulmnemw:conwa~t.)
tagging a document

The act of associating a source profile with an object. You can tag
an object either by assigninga profile inside an applicatinn, or
embeddinga profile in the object as you save it to a file. (See the
sidebar on page 282.)

target descriptor file

( T D n A file, used for scanner and digital camera profiling,

that contains the premeasured values for the color patches on a
profiling target.

TDF See target descriptor@le.

tonal compronlon The remapping of tonal values from a wide &mamic range to a
narrower one. Often pan of gamut compression.
tone reproduction curve

(TRC) The graphed curve that describes the tone reproduction

properties of a device.The graph shows the relationship between
input values to the device and the resulting tone.

total ink limit

The limit on the maximum amount of ink allowed at a time during four-color printing If you allowed 100% of all four inks to
get printed at a time, this would be a 400% coverage. As paper
cannot hold this much ink, the limit is usually set well below 400%.

transmLsive Having the property of transmitting light through a surface.

Backlit media and the RGB color filter layer of an LCD display are
transmissive. See also reflective.
transmittance

(n The ratio of light shone onto a surface to the light transmitted

rhrough to adetector.This is the measurement used by a transrnissinn densitometertn compute densify

The theory of color vision [now weU verified by neurophysiology)

that we have three types of photoreceprors,each responsive to different regions of the visible specmm~.
Consisting of three stimuli. Used to describe measurements and
experiments that ask a test subject to match a target stimulus by
adjusting the intensities of three pre-chosen stimuli.

UCR Under Color Removal. A method of black generation that uses

black only in the neutral and gray areas.

(UV)
The non-visible region of the electromagnetic specmrrrn just
above the high-energy, or violet, end of the r!isiDlespec?crrurn.

UV See ~tltmviolet.
UV brighteners Additives used in papers, inks. and even detergents. to make
whiteslook brighter by absorbing non-visible rtlnat~ioletlightand
re-emittingvisible light.
The lookup table ( L U n located in the mernorpof avideo card.The
videoLUTis accessible by software on the computer, which can use
it to convert all RGB values as they are sent to the monitor. This
provides an easy way for software to control the overall gamma
and white point characteristics of the video system.

visurl calibrator

Software used for monitor cnli6rnrion that requires you to make

visual judgments based on targets displayed on the screen, rather
than exact measurements using a device.

visible lpecmml That part of the specmum containing the range of ruorwlengths
visible to the eye, approximately 380-720 nanornetem
wavelength Ina periodic wave, such asalight wave propagating through space.
the distance from one wave crest to the next. Light wavelengths
are measured in nanometers Ibillionths of a meter). The wavelength of a photon is related to its energy-the higher the energy,
the shorter the wavelength.
white luminance The ltirninanceof the white poinrof a monitor.

5%

Real World Color Manrgcmml, kcond Edition

whtte point

(1) The color (often described in terms of color temperntrtre)and

or dendty) of the
intensity [often measured as either l~~rninance
brightest vrhite reproducible by a device. For a printer, this is the
color and brightness of the paper. For a monitor this is the color temperature and luminance of the monitor when displaying
white, and can be modified.
(2)The color (usuallydescribed in terms of color temperature) of
a light source.

white palnl adaptation The abllity of the eye to adapt to a change in uhirepoint. This is
related to the perceptual task of color constancy.
worktng Space The color space chosen as the default space for documents of a
certain mode. For example, in Adobe Photoshop an RGB working
space is used as the default color space for new RGB documents.
and a CMYK working space is used as the default for new CMYK
documents. In most cases the best choice of working space is an
editingspace, but any bidirectionalspace can be used.

xy chromaticity chart See ClExydingrcrm.

XY plotter A rig used with a measurement inshument for measuring a page
full ofcolor patches.The instrument is mounted in theXY plottet,
which can move the instrument precisely in horizontal or vertical
position on the page. See also snip reader.

Xn

5eeCIExuZ.

A-E (delta-E) values, 43,186,267,533

3-channel encoding. 62

Proofsetup feahlre. 445

third-party tools, 428,453455

3-color dye-sub printers, 192-193

Acrobat 3,424,430

4-channel encoding. 62.63

Acmbat 4,424,430

&hit color, 61,98

Acrobat 5,424,438-440

8-hit encoding, 61

Acrobat 6,424,428.444-448

16-hit color, 98

Actions, Photosbop. 476,477

16-bitencoding, 62

additive primaries, 19-20.21. 52,533

24-bit color, 61

Adobe

256 levels. encodingcolor for, 61-63

Acrobat. SeeAcrobat
blackpoint compensation feature, 355356

Camera Raw. Seecamera Raw

absolute colorimetricrendering, 89,91,291

color engine. See ACE

Abstract profiles 520

color-managed applications. 341. See also specific

ACE, 310,325,355

applications

achromatic components, 32-34,533

consistency of user interface. 341,379

Acrobat. See also specific versions

Distiller. See Distiller

Color Management preferences, 438-439,444

Illustrator. SeeIUustrator

and PDF versions. 424-425

InDesign. See InDesign

and PDF worMows. 423

Preflight tool, 445-447

PageMaker, 290
Photoshop. SeePhotoshop

printing in. 439-440

PressReady,473

555

556

Rwl World Color Managcmd, Second Edition

S\WP Sor SWOP

AdobeCalor Engine. 310.325. SeenlroACE

Adohe RCB space, 278
afterimages,22-23
A1 files. 453
algorithms,screening, 5 i
AM screening. 56.57.533
amblent light, 212-215
Amplitude Modulation screening. 56.57.533
analog controls, monitor, 138-143
anisatropy, 191.192.533
Apap. 455
aperture. 534
aperturesize. spectrophotorneter.165-166
APls, 307
Apple ColorSync. SeeColorSynr
Apple Computer. 65
Apple Preview. 323
AppleScript. 467.475.476.477.479-481
AppleScript Studio. 178
Application Programming Interfaces. 307
applications
color-managed,306307.324-326
CoreiDRAW 393-404
FreeHand, 381-392
Illusmtor, 341-375
InDeslgn. 341-375
Photoshop. 341-375
QuakWress, 405421
non-mllor-managed,306-307.326
wercominglimitationsot 457458
Applied Science Fiction. 153
artifacts, 60,534
anificial light sources. 15. See alsoliglrt sources
artificial trichromats. 21
Artisan Color RelerenceSystern. Sony. 130
assigning profiles
in Adobe appilcatlons. 356358
in color-managed application$.93-94
d e k e d , 282,534
in documents with multiple objects. 289
in FreeHand,381.387

scripts for. 477480

in tlntagged documents. 285-286
assumed-profileworMws. 283-284
assumed pmfiles, 534
assuming profiles, 282,285.295
AtoR tables. 102.264-265
AtnBTag, 517,518-519
autoexposurefeatures, scanner, 116,117,153
automation. 457482
with color serves, 456464
with DevimLink profiles. 464-4157
with m-RIP color management, 473474
with prepms solutions. 468-473
purpose of, 457-458
with scripting, 474-481
tmde-offs inherent in, 482
alltomation products, 458

B
hacking, measuring targets on black, 168
backlight. 127. 138
backwards transform,102
banding. 60.534
Barco Calibrator\! 130
BasICColor Display. 130
BaslCColorPrint3c. 174. 190, 191
batch processors,458
Baser-pattern color filter arm)! 37S376
BestColor proofing IILP: 171
biding
early, 275276,277,280, 501
late. 275,276277,280,501
pwnature. 281
biolog?i color eventsand, 6
Bitmap Export dialog box. CorelDnAW, 401
blts. 61
black
in CMYK model. 53
density of, 65
measuring targets against. 168
blackgeneration, 197,205-207,534
black level. 140441,534

I~lackligl~.llts.
15

defined, 114,121,535

black point

and dynamic range, 138

and Adobe applications. 355-356

and gamma settings. 133.135.22w222

defined. 534

goals of, 121-122

as device variable. 6.3

and monitor analogconuols. 138-143

measuring, 6546

andmonitor geometrj~settings.137

and monitor calibration,218-220

and monitor refresh rate, 137

testing. 218-220

and monitor resolution, 137

hlack point compensation, 355-356.534

and mouse point movements, 144

biack-point test, 218-220

preparing for. 136143

black preservation.472.498

printer, 194-195

hlack scaling. 472

recommended frequencyof, 136

hlack start settings. 195. 196.19n.199.200

mle of measurement in. 119-1~>

blackbody radiators. 10, 13.534

and screensaven, 136, 14.1

blind spot, in human q r , 17

tools available for, 128-132

bluingagents, 15. I67

using densitometers for, 38-39

brighteners

visual. 129. 145.553

fluorescent.15.167

LN,8-3.553
brightness
as nchrumatic attribute. 32-34

vs. profiling. 84. 115, 129

and white lnminance settings. 138-139
and white point. 65-66,133.13~&135
CalibratorV Rarco. 130

defined,32.535

calibrators,visual. 129. 145,553

detectingvariations in, 32

Callas, 455

of LCD displays, 138

camera metamerism, 29-30,157.161.244,535

non-linear response to, 31-32

Camera Raw, 375-379

representations of. 3VMi

built-inproMes. 161.376377

n li~htness,34

calibrating,2449-255

browsers. Web. 302,306

and color. 3763757

BtoA tables, 102,266

file formats 375-376

RtoATag. 518-519

inptlt. 376477

bytes, fil

output. 377-379
and Photoshop.375

purposeof, 375

Calibrate controls. Camera Raw, 249-"55

recommended book on. 379

calibratlon. 126-146
adlu~bnentsmade during, 126128
and black iewl settings. 140-141

RGB spaces supported by. 250

White Balance controls. 377

worldlow mntrols. 379

and black point. 218-220

cameras. Seealsodigital cameras

choosingsettings lor, 133-135

as artificial trichrnrnats. 21
and r o h r constnncy, 47
color filterarray, 157-158.375-376
candelas. 213,535

and color mnctancy, 48

and color temperature settings. 142-143
common problems wlh, 217

558

Real World W o r Management, Second Edition

candles. 13.213
carhode-ray rube monitors. SeeCRTmonitors
CCD arrays. 8
CCDs. 535
CFAcamems. 157-158,375-376
Champoilion, lean Franqois. 79
channels, color. 59,62.63
characterization. 114
charge-coupled device mays. 8
charge-coupled devices. 535
chemistry color evenrsand. 6
chroma, 535
chromatic components, 34-37.535
chromnticAdaptationTag,516
Chromiw. 78. Seenl~oColorThinklCulorThinkFm
CIE. 11.69.535
CIEcolnrimerricsystem,4143.45-46.71
WE LAB. 70-72. Sce uIso LAB model
and coior-management workflo$w.279
defined, 535
importanceof. 70
official name for, 72
and opponeneyltrichromacy,25
and PCS. 84.513
and pelcelved color. 84
primaries used by. 711
purpose of, 42
CIE L'a'b' (1976).72
CIE LCh. 69,72.536
CIE LUV. 536
CIE L'u-v' (197% 72
CIE models. Sw also specific models
caveats regarding. 45-49
officialnames for, 72
opponencyltrichmmacy in, 25
CIE Standard nluminanrs. 11-12.41
CTE uniform color spaces. 42
CIExydiagrarn.536
CIE xy\'diagrm. 41
CtEp,Ymodel. 69.72
ClEXYZ,536
CIEX1'L (19311.69.7?.79.84

CIE xiz Primary System, 41

CIELUV;69.72
classes, profile, 100-102,513,515,52&521
click-balancing, 536
CMMnipe signamre. 513
CMMs. 8688
choosing specific, 88.96
competition among, 87
defined, 536
ICC-complianr.06,88
meaning of acronw, 83,86
purpose of, 83.86.87
specifying preferred, 513
storagelocstionfor, 311.316
using multiple. 06
and white point adaptarion. 07
CMSs. 7S98
assigninglembeddingprofiles with. 93-94
components of, 83-92
conveying color meaning to. 280
critical tasks performed by, 80
defined, 536
limitations of.99
meaning of acronym. 25.79
and one-way us. two-way profiles, 100-I01
m d opponency, 25
performingconversions with. 94-97
role of PCS in. 82
and ~ c h m m a e y25
,
CMYK dye-sub printers, 1'59-200
Cm halftones.55. Seenlso halftones
CMYKinkjet printers. 196198
CMYKmodel
analog origins n t 52.53
device-specific nature of, 67
digital evolution of. 57-59
meaning of "K" in, 53
CMYK numbers
attachingcolor meaning to. 80.84.93-94
changing, for specific devices. 80
converting,with profiles. 94-97
CMYKpnnrers, 57,62,64, 193-200

Index

cMM(p

~ t i 53
n ~
CMYK separations. 53,179, 199
CMYK simuintions. 5 W
GMYK-lo-CMYKconversions
and black presen'ntion. 4i2,4Y8
in CorelDRAW.393
and embedded profiles. 494
in FreeHand, 385
in PitStop Pro. 453
in QunrkXPress.410.413.421
CMYK-lo-RCB conversions
397
In CUI~~DRAW.
in PreeHand,385
in PitStop Pro. 454
and Postscript RIPS,IRS
in QuarkXPress. 410,113
color-banding. 256
color bars. 5 3 i
color blindness. 22
Color Checker. ProPhoto RGB.250-251
color constancy 4618.537
color copiers. 194. 195
color definitions, 63
color-diflerencecalculations. 43
color events. 5-97
defined. 5
illl~suated.5
and metamerism, 27.29-30
pnicipants in, 5
role of light in, 6-12
mleofohfmin.12-15
rnlvof obsemr in, 1 6 3 7
color filler anay cameras, 157-156.375-376
color filters, i52
color gamut. M.72-78, 100
color laser printers. 119.194-196
color-managed applications. Ser nkospecificapplications
benefits of, 306307
andColorSync. 325,326
dibrences among, 325
distinguishingfmm non-color-managed, 326
and 1CC profiles,324

559

and IChi. 324.326

and source profiles. 326
fmm specific companies
Adobe. 341
Cnrel. 393
Macromedia, 381
Quark,405
color-managed worMows. Se~color-management
workflowvs
color management
in applications, 306307,321-326
Acrobat, 438
CorelDRAW. 393-404
D~stiller,441
FreeHand, 381-392
Illuswator,341-375
InDesign, 341-375
Photoshop. 341-375
QuarkXPress.405421
automating, 457-458.493-494. Swalsoautom~tion
and color constancy, 47-46
and color gamut, 72-73, i4.88
and coiorimeny,4 3 4 4
determining needs for. 490-494
disabling 292
and dynamic range, 72-73
and fluorescence. 15
genesis of, 81-82
in-RIP 473-474
key concepts regarding. 34.50.507-508
and L9B model, 71, i 9
in operating system. 305-307. Seealso 0s-level color
management
and PUP workflows. 423-424
and PostScript,425427
in printer drivers, 327-33i
and profile accuracy, 110.113
profile-based,85
~nscannerdrivers. 337-338
and screening algorithms. 57
sirnplicityof,98, 507
and specfrophotometry.45

560

Real World Color Management, Second WiHon

stagesof. 484490
systems. SeeCMSs
terminolow, 282.533-554
ultimate goalof, 46
worldlows. Seecolor-managementworkfimvs
andM9'SM9'G myih, 83
Color Management Modules. SeeCMMs
Color Management Policies settings, Adobe. 344-346.
431-433
color-management systems. See CMSs
color-management workflowvs.271-304
basic sreps in, 48+490
bringing color into. 29&297
as component of overall workflow, 296
conveying color meaning in. 280-285
defined. 272
determiningneeds for, 49W91
and display conversions. 274
eariy- 1,s. late-hinding, 275-277,280,501
and h a d proofing, 489
and non-color-managed environments. 2 s 3 0 3
hetween programs 272.295
within programs. 272,285295
questions to ask abouc 502
and wft-proofing. 28&295,489
timing of conversions in. 275-274,27&280.488
tools for, 494-501
variables d ~ adifferentiate.
t
273-274
Culor Manipulation Model. 86
color matching
and color gamut, 72-73
devicelimitations, 72-73,88
and d!.namic range. 72-73
and lighting, 212
and metamerism. 211
Color match in^ Method. 86
color meaning
assigning to CMYKIRGB numbers. 80,84.93-94
aod color-management workflows, 275-274.280
conve)ing. 28&285
defining. 484.485-486,503-504
color-mixingfunction. 73.537

coior models. See alsospecific color nxrdels

caveats regarding. 4 M 9
defined. 53i
device-independent. 69-72
device-speciBc. 6 7 4
officialnames for. 72
color monitors, 54. SeeaLromonilors
color names, 4 W 9
color-negative scans. 147. 153
Color ReferenceSystem, Sonv Artisan, 131)
Gdor Rendering Dictionaries. SeeCRDs
Color Rendering Index. 539
coior reproduction, 99
color samples, relationship hewecn 26
Color SawyColorMouseToo!, 169
color sewers, 458-464
common featuresof. 4-59
examples of, 461-463
pros and cons of, 459-461
questions to ask about. 463464
ColorSettinp dialog hox
lllusUBtor.342-356
1nDesign.342-356
PilOtOshop. 342-356
Color Solutions. 130
color space
converting. 94-97
delined. 20-21.537
gray-balanced.543
and trichromacy, 20-21
uniform. 42
vs. color gamut. 78
Color Space Arrays. SeeCSAs
color space profiles. 109
color spacesignntures. 513
color temperature
and CIEStandard illurninants. 11-12
and color constancy. 48
defined, 13.538
graph. 12
setting, for CRT monitors. 142-143
color vision

achmmntic w chmmatir components of.32-37

and briphmess. 32-34,3637
and color constaocy. 46
and hue. 34-35,3637
loss of. 22
and metamerism 25-30
and non-linearity.31-32
and opponency,22-25
and sahlration. 35-37
simulating,21
nnd structureofhumaneye. 1 6 1 8
theories of. IS. 2-24
and trichromacy,15-21
mlomnts
and Ctv1YKlRGR f i l e . 53.57
defined. 53.57. F36.538
and device profiles. 100
measlying. 63.64
tone-reproduction characteristics of, 100
ColorChecker, GretagMacbelh.151-152.223-224

calorimeters
and artificialtrichromats. 21
deflned, 530
ilmitations of, 43
and metamerism. 43
purpose of.38
vs. spectropholnmeten. 131. 165
colorimetric inlent. 538
colorimetric rendering
absolute 1,s. relative. 83.91.92.291
and color constancy. 4748
obj~~ctive
tests for, 210
and out-of-gamut colors. 89.91
and sofl-pmofing.291
vs. other tyQes, 50,92
colnrirnetricsvstem.CIE. 4143.45-46
colorimetricvalues. 38
colorimetry, 40-44
as basis tbr quantificationof colon. 50
and CIEcolorimetricsystem, 41-43
and color management, 4344
and colorimetric 1,s. perceptual renderings.50

defined. 40.538
purpnse of, 40
ColorLab. Lago. 225,22%232.261
ColorMouseToo!. Color Sany, 169
colorrsl
additive I-. subtractive, 1%20,52
calculating differences herween. 43
capturing with scannerldigiral camera, 5d55
comparing. 186
cnmplcmentary,21
coniplewityof,78
defining, 4-6,484,485486,503-504
displaying on monitors, 54
encoding.59-62
haw light affects. 5 4
matching. Seecolor matching
mcasurina, 37-45.51
memory. 49,546
metameric, 25. Seealso nietamcrisnl
mMng, 73,537
norniaiizing, 404,486488,504-505
out-oi-~amut.74,83,81(-92
plotting relationships henveen. 20-21
primary. Seeprimarycolors
process, 548
as pmperty oflipltl. 5
as property ofobjects. 4
psychological attributes ol; 48-13
representingon computers 51
reproducing, 99
spot. 486
CalolSpace Conversion profiles.5211
ColorSync. 307-322
contml panel, 30V309
and CorelDMW. 396
defined, 538
and FreeHand. 382.384.388
history of,85
and Mac OS 9,30&311
and Mac OSXPanther.311-322
preferences.316
Profiles folder, 145.311

562

Real World Color Management, Second Edltion

purpose oC 05.306
Colo~SyncConsoniun~.
85
ColorSync Enension, 308
ColorSyncUtility. 314,316322
ColofThinklColoi~inkPro. 78. 186.236-237.257.261-262
Colonron, 1139
ColorVision. 131. 138.139, 141. 143
Commission Internationale de I'Bclairage, I 1,69.See aLw
CIE
Common Unk Printlng Sptem. 325
Compass Pm XT,419-421
complementary colon. 21,538
composite CMYK printers, 193-200,374,385,393,416
composite printers
bullding profilesfor, 193-200
and CorelDRAW, 397,399,402404
and FreeHand.386.390
RnclQuarW(Press,409,412,414,416
compression, tonal, 552
computer monitors. Sermonitors
Computer-To-Plateworkflom. 205
computers, representing color on, 51
cones, in human eye. 17-18
continuowtone color, 59-60
contim~ous-tonedevices, 56.57,539
contone. 539
contrast
defined. 539
simultaneous. 22.46.550
successive, 23,46,552
Conversion Optionssenings
Illustrator, ,354-356
InDesign. 35b35fi
Photoshop. 354-35G
conve~sions
CMYK-to-CMYK.kCMYK.to-ChOK conversions
CM-to-RGB. SeeCMK-lo-RGBconversions
and data loss. 97-98
early. 1,s. late-binding 501
input-to-ourput, 81-81
PCS-ro.destinatinn, 519
profile. 288

and rendering intents, 519

RGB-to-CM'K SeeRGB-to-CMYK convemions
roleof destination profiles in, 85.94-95.96,108
sotlrce-to-PCS.519
timing of. 2i3-274.275-280.48R
using profiles for, 94-117
Convert to Rofile command
InDesign, 358.359
Photoshop. 358-359
cop)+htTag, 516
CorelDRAW,393-404
Apply ICC Profile dialog box, 401
Color Management preferences. 394-399
and ColorSync,396
Composite Printersenings. 39i
exporting from. POI402
Lmpon/Fxponsettings. 397-398
imponing files into. 399-400
Internal RGB settings. 395-396
manttal controls, 400-401
Monitnr settings, 396-397
priming in, 402404
profile embedding in. 402,404
ScaanerlDigitalCamera settings, 398
Separations printer settings, 393.396397.398-399.402
similaritiesto FreeHand,393
strenglhslweaknessesof. 393.404
cornea. 16,17
carrelared color temperature. 13
myola-jet printers. 198
crayon-spitters,198
CRDs. 425426,474,539
Creo Prinergy, 4M. 471472
CRI. 539
cross-rendering.293.294
CUT monitors. See alsomonitors
and ambient light, 212
and black level settings. 140
calibrating, 132 133,136.%cnkocalibration
cleaning, 137
and color temperature settings, 142-143
controlling variabilityin. 119

Index
connols Tor. 127-128
defined. 539
and gamma settings, 133
increasing apparent contrast of, 214
as Ugh1sources, 11
warm-up period for. 136
andwhite luminance settings. 138
and white point senings, 133
CSh. (25426,474,539
C3P worIdlnws, 205
ClJ PS, 325
curlng time, inkjet, 188-190,539
curves. spenml. Y,10
cyan, 53. SeenL-oCM model
Qmbolics Lighrlet. 191

11 values. 40
D50 light hoxes, 214
D50 specification. 533
IXx5 sptxlAcation. 539
Dallm, 455
data averaging, 178
data loss, conversions and. 97-98
daylight
measuring. 215
simulating, I I
wavelength composition of. 10-1 1
DDCZ-compliant monitors. i30
delta-Evalues. 43,186,267,533
densitometers. 3840,540
dansilometrj',3R-40
clensiry
ofhlack. 65
defined. 38,540
as lngariUunicfunclion.34-40
measuring, 38.39
of pixels. 56
nf primaries, 64
destination profiles
defined. 108,540
role of, in conversions. 85.94-95,96. 108

supplying CMS Hrith. 95

vs. output profiles, 100, 108
device attributes, 514
device.dependent colormodels, 540
devicedrift. 114-119.122
device-independent color models 69-72.278-219.540
device profiles.See profiles
device signams. 51 4
device-specificcolor models. 67-69
device-to-device transformations.82
device-to-PCStransbrmations,
102
Devi1:eLink profile%464467
and color-managed applications. 467
and ICC specification. 454.520
maling. 466-467
and prepresssululions. 473
purpose of, 464
remons for using. 4651166
support for,467
unldlrectiond nature of, 485
vs, device profiles, 46441%
devices
calibrating. 114,115, 119.121-122.12M46.Seealso
calibration
evaluarlng.120
monitnring.122
optimizing,121, 154
prnfiling. See p t o f d i i
simlllating. 122
sources ofvariability in, 115-1 19
diffusers, perfect, 39
ditfusinn dither. CMYK. 55
digital camera profiles. Seealsoinput prufiles
building. 157-1131
Camera Raw's built-in, 161,376377
difficulties associateduith. 147-148.244
editing.246249
digital amera targets, 151-152. 158-159
digital mmeras
caphlringcolorwith, 54-55, 157
and color constancy. 47
and color filters. 152

563

564

Real World Color Management, Second Edition

and color gamut, 73

and colar mixing. 73
controlling responses of, 157-158
and dynamic range. 73.74, 157
gray-balancing in. 157
and llghl soarres. 8,152,157
andmetamerism. 157
profiling. See digital camera profiles
variables affecting performance of. 63. 152
white-balancing in, 157-158
Digital ICE. 153
dlgital presses, 194
digital proofers. 202
Digital ROC,153
Digital Swatchbook. 169
directional metamerism, 193
display calibration. 126128,144. Seeulsocalibntinn
display conversions. 274
display engines. 317
displav profiles
defined. 540
evaluatin~,222-224
and ICC specification, 518
importance of, 125-126.146
matrix. 18s. table-based, 103, 12M46
purposeof, 100
role Of calibration in, 126128. 114
saving. 144-146
tonls for creating. 1211-132
two-way nature af. 101-102
vs. other types. 126
displays. Seemonkon
Distiller. 424,127, 430437.441443
dither, CMYK diffusion. 55
dithering 56.57, 540
documents
assiming profiles in. 285-286.289
color management between. 287-288
moving nhjerts between. 287
opening lagged, 2W28:
opening untagged. 285286
ragglng. 282.552

dot gain
compensating for, 205.4Q
curves. 66
defined, 541
dots perinch, 56,541
dpl. 56,541
driver-level color managemenr.32i-338
drivers
CUPS.325
PostScript.313.334-337
printer. 320,329-337
raster. 313.327.329-333
scanner, 337-338
DTP-41 srrlp reader. I7 1
Durst Lambda. 191
d y . 54 1
dye-sub printers, 192-193
dye sublimation. 57
dynamic range
and hlack point. 65
defined, 541
and densimmeters. 40
device limitationson, 72-i3,74
and device profiles. 100
measuring, 40
and monitorcallbration. 138
w. bit depth. 63

L
early-binclmg workflow
and CMYK us. RGR workflows, 277
defind. 541
and ot~tputconversions. 501
pros and cons of. 275-276
vs. intennediate-space workflows. 284
M.lale-bhdingworkflows.275
ECI 2002 target, 175,176,203
editing profiles
alternatives t a 210.266-2fiR
for input devices. 242-249
for oulput devices. 182.2W266
editing space. 541

EFlColnr profiles, 406

Einstein. Alben, 6
electric d i s c h q e lamps. I I
electromagneticspectrum. R
ElectronlBiueEye. 130. 138
electron guns. 127
Electronics lor Imaging, 171
electrophotostatir:devices. 194.541
Pmhedded ProfileMismatch warnings, 350-352
embedded-profile worldlon~.280-283.2M
embedding prnfiles
in cnlor-managementtargets. 155
and color-management wvrldlows, 9 s 9 4
and conveying color meanlng, 280-285
defined. 541
deviceconsiderations,491
rear of. 301
file-forma1considerations.491
and workflowbetween programs, 295
emissive. 541
encodable colors, 62,m
encoding, 542
encodingcolors. 59-62
EnfOCUs PitStop Pro, 5534%
engine
CMh4 as. 83.86
color. 310.325. SeealsoACE
dkplaylprinr. 317
EPS export dialog box. CureiDRAW,401
EPS files. 389-390.453
Epson printers. 188,329-333
errnr diffusion.55, 57.542
European Color Initiative. 175.176.203
evaluating profiles
checking monitorldisplay. 217-224
importance ofmonitorldisplay, 216217
for input devices. 22.5241
hw oUtput devices, 255-264
purpose of. 209-210
viewing environment for, 210-216
events, color. Seecolor events
Excel spreadsheets, 173-174,1iR,z25

Export PDF dialog box, InDesign. 451

Export sellings. CorelDRAW,397-398
exporting
from CorelDW, 401-402
from FreeHand, 388-3W. 392
from InDesign, 448453
PDPs. 448453
eye
color sensors in, 16
how light is focused bv. 16
Illusuated. L i
non-linearityofresponses in, 31-32,41
pan of spectrum visible to, i 4
Eyeone Display, 131
EyeoneMarch. 139.141.143
EyeOne Pro. 169
EveOne targets. 191-192

F
fabrics, brightening, 1S
Fiery RIB. Wecvonics for Imaging, 171
filtcrs
color, 152
digital camerd. 152
infrared. 8
polarizing, 168
Vunrtz. 320422
spectrophotometer, lfi7
ultraviolet, 16. IS7
flags, profile. 514
fluorescence. 8-9. 15,50,542
fluorescent brighteners,15. I67
fluorescent inks. 15
fluorescent lamps. 11.12
TM screening, 5:. 542
Focoltone, 419,460,469,186,49&499
footcandles. 213
fovea,17.542
foveal vision, 342
foveola. 13
FreeHand. 381392
assigning profiles to imported images in, 387-388

566

Real World Color knagemmt, ktond Edition

Color Management Setup dialog box, 31%

glossy media settings. 118

color preferencs. 382-387

Granger, Dr. Ed, 238

and ColorS?nc. 382.384.388

exporting filesfrom, 388-390,392

Granger Rainbow. 238-239.241.251i

graphic arts applications, 30fi

manual conlrnis. 887-3118

graphics, color-managing,36&3M

Missing ltnage Sources warning. 387

in Illustrator, 362-364

Object palette. 387-388

in InDesign,36W362

Preferences dialog box. 382,383

iri Photoshop. 360

printing from, 390-392

Graphics Device Interface. 543

quirkinessofcolarmanagement in, 381

gray-balanced color space, 543

Separationsprinter settings, 384-385,389,390-391

gray-balancing. 157.513

color-negative scanning targets, 147

CretagMacbeth

I:uii

Gray Component Replacemenr. SeeGCR

Pictrogmphy 191

ColorChecker, 151-152,223-223

Velvia. 150

Eyeone Display,138
EyeoneMarch,139, 141, 143

EveOne ho, 169

gatn controls. CRT. 1.12-143

EyeOne targets. 191-192

gamma

1CCol~r
snip reader, 171

curves. 66,542

iQueue, 461-463

defined. 542

MeasureTool, 169,178,186,189.2fil

monitor setlings. 133.135.22~~222

ProNeMaker. 167

rays. 7

Proflehlaker Pro. 156. 15%160,169.175

scanner settings, 151

Spectmlino,169

gamut, 72-78

Specnoscan,170

clipping. 89

target r e f e ~ n c files.
e
174

and colorantslprimarier. 64
compression. 88.543

defined. 88,542

halftones,55,5647,543

and device limitations. 72-3

handheld spectmphotometers. 168-169,171

and device profiles. 100

hard-copy references.300

mapping, 74-78.543

hard proofing

piots. 7%77

and color-managementworkflaws. 489.500-501

and rendering intents. 88.91

and prepress solutions, 469

gamutTBgag.519

in QuarkXPress. 409,418,419

CATFIRHEM Light Checker, 215

HCT target,106,150

GCR. 197.204.543

headers. ICC profile. 512-515

GDI. 317.543

heat. 13

GEM technology, 153

Heidelberg Printopen. 17%181

geometry measurement. IfiFlfifi,546

Helios Imageserver, 47U-471

gmmevysenings, monitor, 137

Hering, b a l d , 21

glossaw, 533-554

Hering theor): 23

Iuero@yphics,79
high-bit devices, 63,154455,543
HLS color model. 543
hockey pucks. 132
HSB color model. 543
HSVcolor model, 543
hubris. 544
hue
as chromadc component of color vision. 34
and color aantes. 48-49.50
defined, 34-35.544
representnlionsof, 3637
hueshifts, 49,50.71
human eye. Seeeye
Hunt effect.212
Hurcheson. Don, 106. 150, 154.2311
Huygens. Christian. 6
hybrid profiles, 517
hybrid scripting. 480
hyhrid wnrkflows. 286285

technical inlormation about. 51 1-521

verifyingintq7ifyof.318
viewing currently installed,318
ICCspecifiration,51 1-521
and Abstract pmfiles, 520
and ColorSpace Con\wsion profiles,520
and DeviceLhk profiles, 520
and display profiles,518
and input profile?,517
IS0 version, 511
and Named Color profiles.521
and output profiles,51&5lY
profile structure,512-515

purpose of,5 l i
required tags, 516
ICColornrip render, 171
ICM, 306.307-308.322-324,544
illuminance,213
iliuminnnts
and CIE colorirnetrksystem. 4 1
CIE standards for, 1 1-12
defined, 11,544

ICC
and CMMs. 86,8!l
and CMSs.83
defined, 544
goals of. 85
meaninj? of acronym. 85
profiles. SeelCC pmfilcc
Website,511
ICCAutoflow,46l.4fi2
1CC-based color management systems. 83
ICC Pmfile Format Specification,85
ICC profiles. See also profiles
and black point compensation. X55
and color-maria@ applications.324
and ComlDRAW. 401
headers for, 512-515
iliusueted. 512
and PDFs. 426
and QuarkWress.407
structure of, 512-515

discounting. 413
most imponant, Tor graphic arts. 41
Illumination. International Commission on, 1I, 69. See also
CIE
Illustratnr
Assign ProRIesconimand.357-358
Color Management Policies settings. 344-34-16
color-managing imported graphics in. 362-364
Color Settings dialog box, 342-356
Conversion Options settings, 354355
convening placed graphics in. 363
cross-rendering in, 293
Document Info Palette bug. 35(l
Inlegration with other Adobe applications.341
Missing Profile warning. 346,349
perfnrmingcolor-nlode conversions in, 359
placed graphics in, 362-364
printing controls, 370-373
Profile Mismatch warnings.352-354
I'roolSeNp feature. 365-3fi6

568

Real World Color Management, k m n d Edition

slmuhtions in. 290.365-366

Image Cnlor Management. SwICh!
images. See nlsographics
assigning profiles to importd. 387-381)
demosaicing. 376
saving, in LAB model, 70
unpmfiied legacy. 485
Imagesetters,58.205
Import dialog box. CoreiDMW 399
Import settings, CorelDiUW,397-398
in-RIP color management, 473474
lnDesign
Assign Profilecornmand.357-3%
Coior Managemen! Policies settings, 14&346
coior-managingimported graphlcs in. 360-362
Color Setrings dialog box. 342-35fi
Conversion Options settings, 354-3513
Conven to Pmfilecommand, 358.359
cmss-rendering in. 293-294
Export PDF dialog box. 451
exporting PDFs from. 448453
Image Color Settings command, 362
Lmage Impon Options dialog box, 3131
integration with crtherAdoheapplications,341
Missing Profile warning, 34fi. 341%-349
PDF Export Presets dialog box, 450
placinggraphics in,36&362
printillgconnols, 370-375
Profile Mismatch warnings 350-351.352-354
PrnofSetup feature. 3663fii
simulations in, 290.366367
sofbprooling in, 369370
infrared filters.8
infrared light, 7. a. 544
ink, metameric, 26,545
ink limits
defined. 552
for prlnters. 195-196. 198. 199.2ilil
for prlntlng presses. 204-205
inkjet printers
color senirtgs for, 18R
connolling variabilityin, 119

cuing time required for. 188-190

media senlngs for, 185-187
profiling
CMYKinkjets, 196191)
RGB inkjets, 184-191
resolution settings for,188
input devices. S4P~Isodigitaicameras; scanners
and colorgamut, 73
dynamic range of. 73.74
prntiling. Sminput protiles
sources ofvariabilityin. 63
variables affectingperformance o t 152
input profiles,147-161
h~ilding
for digital cameras, 157-161
forscannem, 153-156
defined. 544
and devicevariables. 1difficultiesencountered with. 147-148
editing, 242-249
evaluating.ZW41
hybrid. 517
and ICCspecficatiun, 517
matrix- ln. table-based. 103.517
naming. 156
one-way nature of, 10i&iOi
purpose of, IOU, 161
and RGB values. 148. 155
saving. 156
targets for, 148-152
vs. sourceprofiles. 100
input-to-output conversions.81-82
Integrated ColorSolutions. 130. 174,1YO, 191
intensiry. 31,544
intermediare.space $votldlows,280
international Coior Consortium. See ICC
lnternational Commission on Iiluminotion, 11.69. Sccnlso
CSE
International StandardsOrganization,212. Seeolsg IS0
Internet
colorworkfl~~.
303
managing color on,302-303

lnda
and monitor calihracion,303
Internet Explorer. 302
Interpolation, fli.107,376,379,544
ipueue. 461463
IR filters. 8
IR wavelengths. 7.8
iris. I7
lS0.212.427.511.544
I'r8standanl. 149-150.175-176.544

J
InvaScript. 4i5.476.4Ti47R. 481
lPEC
and embedded profiles. 282,491
expordng, 326
and gay-iwhite-balancing, 158,244
and non-color-managed applications. 326
previewing. 323

K
K
as abbmiation for keltrin,13.544
in CMYE; madel. 53
Kelly-Moore Pain! Co.. 211
kelvins. 13.541
KnaLThomas. 249.377
Kodak
DCS Acquire plug-in, 502-503
film. 150
scanner targets, 149-150
and "secret sauce" in profiles,88
Kodak DigIra1 Science. 382,384,388
Korzybski.Aifred, 113

L
1B-based workflows. 497
LAU model. See also CIE IAB
importance of, i l l
limitations of, 71
and opponencylVichromacy. 25
purpase of, 42
as replacement for LW, 42

saving images in, 70

as universal translation language. 70,N
LaCie EleeVonlBlueEye, 130,138
laminated proofs. 216
laser printers. 56. 1 19. 1944%
iate-binding workflows
and CmIRGH worMows. 276
defined, 545
and output conversions. 501
and prepresssolutions. 469
pros and cuns of, 276
m.early-binding workflows. 275
vs. lntermediate-space workflmvs, 28ll
ILD monitors. See also monitors
andambientlight, 212
and black level settings. 140
cdibrating. 132. 133. 136. Seealsonlibration
cleaning. 137
and color temperature settings. 1 4 2
controls for, 128
default brighmpss of, 138
defined, 545
warm-up period for, 136
and white luminance settings. 138
andwhite point settings, 133
Left Dakora. 466.467
legacvimages, unprofiled. 485
lens, in lluman eye, 17
libraries. named-color, 498499
light
absorption 1,s. ~ f l e c t i o nof. 13-14
chancteristia OK 6 9
color as properly of. 5
howobjects interact with, 6. 12-15
inrensiryof,31
measuring. 37-38,213
monochromatic. 9,32
and physics ofrrAor. 5
sources ol: 10-11. See also baht sources
way* of modeling. 6
iight boxes. 214
light checkers, 215

569

570

Real World Color Management, Second Editlon

light sources
achromatic us chromatic attributes of. 32-37

and color temperature, 13

defined, 10
and input devices. I52
hiloring protiles to, 214
typesof. 10-11

raster drivers. 329-330

source pmfle assumptions. 328
Mac OS X 31 1322
AppleScdpt Studio. 478
and ColorSync. 306.311-322
displaytprint engine. 317
Displays panel, 314

lighrbulbs. 13

and FreeHand, 381,392

lighting

generic pmhles provided with. 315.32a329

importance of. 50

as rntlltiuseroperaringsystem, 315

IS0 standards fo~.212

PostScript driver. Xis337

Lightlet,Cymbolics. 191

Prim dialogs, 33SY31

lightness, 34.545

profile locations. 315-316

linear relationships,545

Scripfs folder. 47fi

linearization, 121. 17R-179.1'33. 205,545

Lines Per Inch. 545
Link-o-Inlor,466

source pmfileassumptions.328-329
Macbeth CoiorChecker. 151.223-224
Macintosh. SeenlsoMacOS

logarithmic. 545

color-managedbmwsen, 302

logarithms. 33.39

color-management scripts. 476

Logo ColorLab. Z s . 22a232,261

gamma settings, 302-303

longwavc-radio waves. 7

monitor settings. 310-31 1

lnokup tables, 66,103, IZi, 545

OS-level color management on, 306.308-322.31:!

ipi. 545

profile Locations on. 311.31E316

ltlrnens. 213

saving profileson. 14W46

luminance. 138-139, 168,213,415

maKenfa, 53. SeealsoCMYKmodel

luminous flux.213

mapping, gamut. 7P78.543

luminous intensib', 213,546

Margulis, Dan. 49

luminous power, 213

mathematical color models, 52.69

I.UTs. 66. 103,545,546

matrix, 546

LWcolor space. 42

matrix-based pmfles. 102,103,517

matte media settings, i 18

MatrixColumnTag. 517
Mac OS 9,308-311

meafiurement,incalibrationlprofiling,119-122.16M7i

AppieSrriptr foldel; 476

measurement files. managing, 173-174, 178. 186,225

CMM Incalion. 311

mea~uremenrgeometry, 165-166.546

ColorSync contml panel, 308-310

MeasureTnol. 169.178.186.189.26i

dispiaylprint engine. 317

rneastlring

generic profiles provided with, 315

color, 37-45.51

Monitors control panel, 310-31 i

color conslancy. 47

PnslScript driver. 334-335

colorimeIxic valuer. 38

printer drivers. 329-3313

density, 38

pmfile locations. 311

light, 37-38.213

specaal properties. 38

monitorcalihrators. 104. 130. 132. ,See-nlsocalibmtion

spectral reflectance.44

monitor compensation. 274

tone reproduction. 6667

monitor hoods. 132.214

measuring instruments. 165-171

monitor phosphors. 54

mediaUllnePoinrTag,516

monitor profiles. 546. See a h displayprofdw

memory colors, 49.546

monitor RGR. 54.546

metameric colon, 26

monitor simulation, 290

meti~mericinks 26.546

hlonitor Spyder. 131

metamerism. 25-30

monitor-to-printcomparisons. 215-216

benefits of. 28-29

monitors

camera, 29-30.157. 161.241.535

ambient Ught lwels fnr, 212-215

and color management. 50

and black level settings. 140-14 1

and colorimerers. 43

calibrating. 48.65.126146.217-222.303. Seenlso

defined. 21,25-26,548

calibration

directional.193

clearing. 137

examples of, 28

and color gamut, 72.74.76.88

illustrated. 27.30

md color temperature settings, 142-143

ohsewer, 517

conmllingvariability in. 119

printer. I93

DDC2-compliant, 130

scanner. 73.550

displaying color on. 54

testing,262

dynamic range of, 72-73

why it occUrs.262i

extending life of, 137

metamen. 25.546

factors aKecting color on, 64

Microsoft

and gamma settings. 133,135.220-222

Excel. 173-174.178.225

as light sources. 11

Internet Ewplnter. 30'2

measuring behavior of, 63

Windows. See Windows w t e m s

profiling.Seedisplay profiles

microwaves, 7

and white luminance settings. 138-134

Mismatch warnings. SeeProNe Mismatch warnings

and white point settings. 133-135

mismatched color, i2-78

Monitors control panel. MacOS9.31LL3-311

Missing Image Sources warning. FteeHand. 387

monochromatic light. 9.32.5.1:

Missing Prolile warnings

Illustrator. 346.349

multichannel printers. 207

mystery meat. 297-301)

InDesign, 346.348-349
Photoshop. 286,346347

Mirsubirhi SpemVue, 130

Named Color profiles,521

mnemonics

named-color systems. 498499. Sew nku Focoltone:

for bands in s p e c m . R
for complementary colon. 21
Monaco Systems. 131

Pantone
names, color, 4 W 9

MonacoOPTLY. 131,139.141,143

nanometers. 7,547
NCPs,521

MonacoPROFILER. 179-181,503

neutrals. 48

9 4

Real World Color Managemant, Second Edition

Newton. Sir Isaac. 6.0, 51

nit. 213
nm. 7
non-color-managed applications. 306307,326
non-color-managed environments, 300-:303
non-linear relationships. 547
normalizing color, 484.486488.5OP505
null translorm. 547
number systems (for representing color). 51-78
ambiguities in. 5'2
analog origins of,53-56

device-independent niodels, 651-72

devicelimitations affecting, 72-78
device-specific models. 67-69
digirdl evolution of. 37-59
hmrfthey work, 59-62
number of colon In. color definitions. 63
why numbersvur): 63-67

0
objem
color as property of..I
reflective, 14
role of, in color events 12-15
transmissive. 14
observer, in color events. 5. 1 6 3 7
obsen~ermetamerism.547
OD. values. 40
DmniWeb. 3u2
one-itlput-one-cn~tp~~t
workflow, 81
one-wavprnfilps, 100-101
OneVision. 455
oppanency, D-25,46.50,547
opponent-color theorg, 23-24
optic nerve, 17
OptiCAI.. 131.138. 141
optical density, 40
optimizarion, device. 121. 154
ordered dither, 57.547
OS-level color management. 305-324
on Macintosh, 306.307-322
pros and cons of, 305

on Windm'ssystems. 306, a2-324

OS X. See Mac OS X

out-of-gamut colors, 74.83.80-92

output devices. Seealsospecilic dwices
andcolorgamut. 72,74.88
convening colors for, 81-82,4138-189
dynamic range of, 72-73,74
limitations of, 72-73.88
measuringhehavior of, 63
profiling. Seeoutput profiles
soulrff ofvariabiUtyin, 116
oulput profiles
building
forcomposite CMYK printers, 193-200
for multichannel printers. 207
f i r printing prerws. 200-207
for RGB inkjet printers, 184-191
for three-coiorrtye-sub printers, 192-193
for true HGBprinters, 191-192
and CMSs. 163-164
colleciingdata For. 165171. 182
cost considerations. 171-172
and data averaging. 178
defined, 547
downloading targets for. 173
editing. 182.264-266
evaluating, 255-261
and iCC specification. 510-519
importance of accuracy in, 164.182
and linearization. 178-179
and measuringinsments, 155-171.172.182
openinglprinting targets for. 182-184
packages for creating. 171-182
and paramelercontrols, 179-18?
purpose of, LOO, 163
tahie-hased n a m e of, 103. I 7 3
targets used for, 175177
two-way namreot 102
vs. dertination profiles. 100
Uarputlnlent. PDFIX 428

meaning ofanonym. 4UW24

page-layout applications.290.421

PDFlXas standard for. 427

PageMaker, 290

and print worMows, 423-424.427,455

paint recommendations.211

thirfl-partytools for working with. 428,453455

Panther, 306.31 1-322.336337.381. S e aka Mac OS X

perceivable colors, 62.64

Pantone. 419,4M),469.486.49a-499

perceptual intent. 547

parameter Controls, proNing tool. 179-182

perceptual rendering, 4748.50,8e-89, 90.92

panicle, modeling light as, 6

perceptually uniform, 547

Paste Rofiie Mismatch warnings, Adobe. 352-354

perfect difirtsers. 39

ITS
and CIE LAB. 84,513

phosphors, 54.64, 127,548

PhotoCAL, 131,138

and CIE XYZ. 84.513

photometry. 213

as component of color-managementsystems. 83

photons

defined,547

ahsorption of. 15

as hub fur device-to-device transformations.82.84

bfendingof, 9

meaning of acronym. 82

counting, 213

signature. 513

defined. 6.548

PCS-to-destination conversions, 519

energy level o t 6

PCS.to.device transforms. 102

and fluorescence,8-9

PDI: Export Presets dialog box, InDesip. 450

PDF Handshake. 47M71

and light sources, 10

hlgh-energy,8-9

PDF Inspekror. 455

photoreceptors. 1617, 18.21.548

PDF workfiows. 423424,427,455

Photoshop

PDFIX

Actions. 476,477

and Acrobat Preflighttool. 445.447

assigning pmfiles in. 357

creating, with Distiller, 441443

Camera Raw plug-in.375. Se~alforamera

Raw

exporting. from Mesign. 448-452

choosing rendering intents in, 293

and lSO, 427

Color Management Policies settings. 344-346

as standard for PDP-based print wurkflws, 427

color-managingimported gaphics in. 360

verification of, 441-443

Color Settings dialug hox. 342-356

versions. 427-43fl

Conversion Options senlngs, 354-356

PDFIXCheckup.455

Convert to Profile command, 358-359

PDFIXlnspektor, 455

croswenderingin. 293-294

PDPS, 423437

and embedded-profile workflows, ZH4-'85

and color management. 42W24.430

integration with other Aduhe applications. 3 4 1

considerations for maldng. 4 3 ~ 3 7

IavaScript support, 475

converting PostScript files to. 430

Missing Prnfilewarning,286.346-347

andembedded pmfiles. 425.491

openlngtarger scans in, 155

exporting from IoDesign. 445453

printing conuols, 37c-373

f l a ~ ruf,
j 424430

printing profiling targets from, 183

and FreeHand. 389-390

Profile Mismatch warnings. 287.288.35U.352-354

and ICCpmfiles. 426

ProbfSetup feature.217.367-369.505

574

Real World Color Management, Second Edition

simulations in. 290.367-369

soft-proofingin, 369-370
physics. colorevents and. 5 . 6 7
I'ictmgraphv. Fuji, 191
pipmenl. 548
Pitstop Pm 453454
Pitshurgh Paint. 211
pixels
capturing, 502-503
definingcolor for. 503-504
densityof. 56
normalizing color for, 504
pixels per inch. 56.548
placed graphics
in Ulustrator. 362-364
in InDesign. 3fin-362
in Pholoshop,360
Planck Max,6
plateseners. 205
plos. gamut, 75-77
plnners. X(. 170-171
polarization,548
polarizing filters. 16R
Poiamid Sprintscan. 153
Portable Document Format, 423424. See also PDFIX:
PDFs
posterization, fin, 217,221, 256. 548
Po3tScript
colnr-management philosophy, 425427
conveningto PDE 430
drivers. 193,313,334-337
and in-RIP color manaRement. 473-474
printers, 199.327
ppi, 56,548
I'raxisoft. 419.461.462
I'reW, 138, 139. 143
premature binding. 281
prepress solutions. 5 6 M 7 3
wmmon charac~eristimok 468
downsides of. 470
examples of, 470-172
questions to askabout. 472473

when to use, 468469

presses Seeprinting presses
PressReady, 473
Preview. Apple, 323
primaries, 548.Seealro prinlary cnlors
primarycolors
additive. 19-20.52.533
density of. 64
subtractive, 20.52,64.552
and trichromacy. 19
primary platform signature,514
Prinergy, 468.471472
Print dialog box
Acrobat. 449
CorelDRAW,402403
FreeHand. 391
lllusvalor. 372
LnDesign. 373-375
Photoshop,372
QuarkXF'ress. 416418
print engines. 317
printer C M ,55-56
printer drivers, 320.327-337
printer targets. 224
printers. SeeaIsospeclficlypes
black-generation settin@for, 197
calibrating, 194-395
and rolor gamut. 72,88
contmllingvariabiiilyin. 119
dynamic range of. 72-73,74
impmvingcolorlblackpolnton, 65
pmfiling.105,184-200. Sw~1so.outp~t
profiles
recommended ink limits for. 195196,IYR. 199,200
printing
in Acrobat. 4 3 9 4 0
in CarelDRAW, 402-404
disabling color management prior to. 292
in Freeliand, 390-392
in lllostrator. 370-373
in inDesign, 370-375

in Photoshop,370-373
in QaarkXPress,41-18

targets. 183. I84

printing presses

convening colon with. 85,9447

defined, 82,549

approaches to n~nning.201

editing. 210,242-249.2fi4-268

and black-generation strategies. 205-207

embedding. Seeembedding profiles

choosing profiling targets for, 202-204

evaluating. SePevaluating profiks

and contract pmofers. 202

generic. 108-109,320-329

conwl~ing~riabilily
in, 119

importance or. 113

inklimits for, 204-205

incompatibility of. 85,88

optimizing, 200.202.204

information contained in. 100

profiling. 201-207

limitationsot, 107

I'rint[lpen. Heidelberg, 179-I81

matrix- IW.table-based, 102, 103

process color, 548

namins 156

process control. 38. 114

one-way vs, two-way, 100-101

protile classes. 100-102,513.515,520-521

purpose of, 82.83.84-85, 100

Profile Connection Space. SePCS

roleolmewurement in, 119-120, lliE171

PrufileCrealnr signatwe. 514

saving. 144-146, 156.311

Profile FirstAidutility,316.318

source us. destination. 85, 1118

pmffleflags. 514

storage locations for, 311,324

profile headers, 512-515

testing. S@eevaluating profiles

types of, 100-102

profile ID. 514

profile mismatch. 549
ProfileMismatch warnings
nlustrator. 352-354

visualizingprocessof making. 104-106

ProfleSeque~iceDescTag.520
profiling. SenalsoproNes

InDesign. 350-351,352-354

cost u)nsiderations. 171-172

~hotoshnp,2~7.
ZRR, 35n. 332-354

defined. 114.549

pmfile versions, 513

iterative nature of, 268

profiled captures. 485

packageslluuls. 128-132. 156. 159-161. 171-182.

profilellescriptionTag.516

190-191

ProfileMaker. 167

role ofmeaumment in. 119-122.165471

ProfileMaker Pro, 156. 159-160, 169. 175

vs. calibration, 84.1 15, 12.5126, 129

profiles,99-109. SeealsoICCprofiles

h o f S e t e p reamre

accessingmotely, 480481

Acrubat,445

accuracy o l ll0.ll3.227

Ulustrator, 365366

assigning. Seeassigningpmfiles

InDesign. 366367

assuming, 282,2~5,?95
building specific Iypes
display profiles. 125-146

Photoshop, 217,367-369,505
proofers, digital. 202
proofing

input profiles. 157-161

in color-managemrnt worWlows. 484,489,500-501

output profiles, 163-208

defined. 484

classes of, 100-102.513.515.520-521

and CMSs. 82,99
convening. 288

in hllesim, 369-370
in Photoshop. 369-370

9 6

Real World Color Management, Second Edition

hi Q u a r W m s . 409.418.419

reflective ohjec& 14,549

remote, 471

refresh rate. monitor. 137

proofing systems. 419.492

PmPhoto RGR Color Checker, 25CL2-251

relative colorimetric rendering, 89,91,92,291

remote profile access, 480-481

psychological color attributa, 48-49

rendering
colorimetric. Seecolorimeiricrendering

cross-. 293.294

quantum physicr. 6

dictionaries. SeeCRDs

QuaraPress. 105-421

perceptual. 4748.50,88-89,90.92

built-in color management. 407410

Color Management preferences,411414

saturation. 88.89.90.92
rendering intents, B8-92

and Conlpass Pro XT,4 l W 2 1

andcolor-spaceconversions.9R. 519

ERColor profiles. 406

defined, 549

Get Picture dialog box, 415

descriptions of. 88-89

hard proofingin.409,418.419

and ICC specification. 514.515

and ICC prufiles. 407

illustrated,90-92

limirationsofcolor management in. 405.421

and out-of-gamut colon. 88

manual contrais. 4LS-41fi

purposeor, 83

printing in. 416418

reproducible colors. 62

Profile Information palette. 416

required tags, in ICC profiles,516

simulation controls, 290

resalution

soft-proofingIn. $09

monitor. 137

rurningul~colormanagement in, 40M07

printer, 188

versions. 405,407410

scanner. 154-155

Quartz. 317,329,549

retina. 161R.23-24.514

Quarw filters, 32&322

RGB devices, 52,184-192

QuickDraw: 317,328329,549

RGB Explorer,238-239.256

HGB model

UGH inkjel p ~ l e r s184-191

,
14 ratio, 39

analng origins of, 52.53

ratinr. 7

device-specificnaNIe of, 67

radinuon. 13

dlgital evolution of, 57-59

radiators, blackhntly. 10, 13

RGB numbers

radio waves, 7

attaching color meaning to. 811,84.93-94

Rampage S3.stems. 468

changing,for specific devices, 80

raster driwn.313.327,329-333

convertin&wilh profiles. 94-97

Raster lmnge Procesor. SeeRlP

Rpul I\brId Cnmem RNw I

I ~ IM~ nI l w

RGB printers. 184-192

Wwroshop CS.379

RGBlo-CMYKconversions

receptors. 1617.21

automatic. 472473

reference light source. 213

and embedded pmtiles. 302

reflectaoce. 14,39.44,549

in PitStop Pm,454

reflecrion. 13-14

nnd press profiles, 203

in Qt~arkXPress.
409,410,413,421

variables affecting performance of. 63. 116, 11:. 152

RGBEXPMRERB. 238-239,256

"Scnnnir~g
Guide,'"on

IW 171. lR5.473-474.549

scans, color-negative,147. 153

Hutchesonk 154

mds, In human eye, 17

screen, cleaning monitor. I37

Rosena stnne. 79

screen angle. 550

roundingerrors, cnnversion, 97-!I8

screening

Rumney, Ralph. 113

algorithms,57
AM, 56.57.533

defined, 550

samplesperinch. 5fi. 551

FM. 57,542

saturation

stochastic. 55.57.551

as chromatic component of color vision. 34

scmensavers, 136. 144

defined,35.550
rendering. 88.89, 90,92

scripting 474481

representations of.3637
saturation intenl. 550

defined. 475
examples, 476481
hybrid. 480

!canner autoexposurefeature. 116,117,153

learning, 478

scanner drivers. 337-338

pros and cons of, 475475

scanner metamerism. 2930.73.550

scanner pmliles. Swalso input profiles
building, 1%-156

purpose of, 457,474

sepmtion parameters, 179-182
separaljuns

editing. 244-245

ndjustingpress h r questionable, 201

evaluating, 225.235

and blackgeneration. 197. 199

stimulus/response for. 106.153-154

CMYI(. 53.179, 199

vs. digital camera profiles, 148.244

compensating for dot gpln in, 205

*canner RGB. 54-55

and contract pmofen 202

scanncr targets, 149-150.224

creating. 204

scanners

GCR. 197

as artificial trichromats. 21

in.HII! 275,276,474

atltom~licfeatures for. 116.117. 153

capturingcolor with. 54-55

multichannel. 207

cIlan@ngdensifyof whitelblack points on, 66

and color encoding,63
and color gamut, 73
nntl cnlor mixing. 73

simulating. 386
UCR, 197
Separations printer
CorelnRAW settings for. 393,39&39i, 398-399. JUZ
FreeHand settings for. 384-:385 389.390-391

default tone curves on, 15(

server-based automation products. 458464

dvnamic range of, 63.73

Shenuin-Wffliamspaint, 2 1 I

factors affecting coior on. 64

signature.

Ramma settings for. 154

"hi~h-bit."63.154-155

coior space, 51.3

nprimking, 154

device msnufacturprirnodei. 514

platioml. 514

profiling Spe scanner profiles

preferred CMM, 513

578

Real World Color hnagement, Second Wiion

pmmeconnmian space. 513

pmlile creator. 514
simulations
CMM;,son
device. 122
In Illustrator, 290.3653fifi
in InDesign,290,366367
monitor. 290
in Pit~eMaker,290
in Photoshop. 290,367469
in QuurkXl'ress, ZYO
and softproofs. BO,499-5110
sinlultaneous contrasf, 22.46.550
skin tones, 49
soft-proofing
in color-managementworkflon~.288-%IS, 489
in inDesign. 369-370
in Photoshop, 369-370
in QuwWPress,409
solid ink printers, 198-199
S o l ~ ulamps,215
Son). Artisan Color Reference System. 130.138
source profiles
assuming, 328-329
and CMSs. 85.94-95.96
and coior-managedapplications, 326
deducing.29li
role or. in color conversions. 85.96. 108
from unknown sources, 297
w. input profiles. 100
source-10-PCSconversions,519
space
space, 551. .Sre~lsocolor
Specifications for Weh Offset Publications, 109. Scrnlso
SWOP
spectral curves. 9. 10.13
spectral data, 551
spectral energy 9,11
specml power dislrihution. 551
spectral properties, 38
spectral reflectance. 14.44. SeeaL~oorrflectance
SpectraVue, Mluublshi. 130
Spectrocam. 169

Spectmfiler, 170
Spectrolino. 167, 169
specrrophotomrters
a p e m e size for, 1661fii
automated. 170-171
cnmparin~colors with, 186
defined, 551
and delta-e values. 186
handheld, 168-169,171
measurement geonietry for, 165-166
UV filters for, 1137
vs, colorimeters.131
and white point adaptation. 16:
spearophotomeq, -5
Specrroscan. 170
Spemstar, I69
spectrum
and additive 1%. subtractiw primaries. 21
defined. 7,551
how colors are listed in. 21
mnemonic for labeling hands In, 13
spi. 56,551
spot color, 486,551
spreadsheets, managingmeasurement files with. 173-174.
178.186.225
Sprintscan. 153
sftGB. 551
Standard Inuminanrs.CIP. 11-i2,11,551
StmdardObsewer~slimulusresponse. 41.551
steradim 213
Stevens effect, 212
stochastic screening, 55.57.551
strip readers, 171,551
subtractive primaries. 20,21.52.4.552
successive contrast. 23, 46.552
suhlight.13. Seedsodaylight
surfaces, paint recommendafionsfor. 211
SWiNG. 455
SWOP
mnverting objects in PIX to, 321
and C~oPrinergy.471
and Distiller, 442

and generic profiles,109,315,486

Young-Melmolrz, 18

meaningofacronym, 109
presses. 84.202.204

zone, 24
thermal energy, 13

simulations,410

three-colordve-sub printers, 192-193

and soit-proofing. 486

three-component theory of color vision. 18

and unknown outputs. 493

TIFF files. 154-155.225.389.491

synd~etictargets, 238-241

tonal compression,74,552

tonelevels, 59

.r ratlo, 39

tone reproduction, 63.6647.100

tone curves. scannel; 154

rable-hasedprofiles,102, 103,517

tone reproduction curves. 6667,517,552

tag table. 512.515

total ink limit. 552. S e e a h inklimits

tagped documents. 282.266287

transmissive objects. 14.552

tagged element data. 512.515

transmittance. 39,552

tagging documents. 282.552

TRCs. 66.517.552

tags. ICC-requited,5113

'I'HCTag,517

target description files. 148,552

trichromacy, 1621

target reference files,174

and additive us. subtractive primaries. 19-20

targets

artificial,21

custom, 175

in C E models. 25

dlgital camera, 151-152.158-159

and color space. 20-2 1

downloading. 173

defined, 16.553

editing, 243

effects unexplained bv 22-23

formatting. 173-1 74

HCT, 106.150
input profile, 14@-152
opening, 183-184
Output profile, 175-177
printer, 224
printing. 1X3, 184
RGB process control, 187
scanner. 149-350.154-155.224
synthetic. 238-24 1
TDP$148.552
temperature, rolor. See color temperature
termtnology. 282,533-550
testing profiles Seeevaluating profiles
tlleories

how it Works, 16-18

importance of, 16.19
and metarnerism.25
reconcihg opponenq and. 24
training yourselfto Pcognize, 50
vs. tristimulus.18-19
trichromats, artificial. 11
tristimulus, 1R-19. 41.553
trlstimulus response,Standard Observer, 41.551
tungsten lamps, 11.13

TWIST,455
tWo.wayprofiles. 100-101

U
UCR, 197.204.553

~ o ~ o r n s i o18.23-24
n.

IJlrraLinks. 467

Hering, 23

ultravio!et filters, 16. 167

opponent-color. 27-24

ultraviolet light. 7,&9. 15,553

three-component, 18

Under Color Removal. See 1 ICR

580

Real World Color hnagoment, Second Edition

urrprofiled coptum. 485

and additive primaries, 19-20

unprofiled legacy images,4H5

blending of, 9

untagged documents, 285-286.312

defined. 553

Upton. Steve. 78

fordifferent llght sources. 10

IIV brighteners. 8-9.553

low- L~F. high-enew 7

UVfillers, 16. 167

orphotom, 6

lIV Light, 7,O-9. 15.553

LIV wavelengths, 7.8-9

short vs. long. 7

and subtractive primaries. 20
waves, modeling light as. 6

Web browsers. 302.306

variables
in color-management workflows,273-274

Web OffsetPublications, Specifications for, 109. See nlso

SWOP

conuolling. 115-119

~ i t emlor
, of, 13

in digital cameras. 63, 152

White Ralance controls, Camera Raw, 377

ininkjet prin~ers,I19

white-balance settings. digital camera. 157-158.244

in nlonitors. 63. 1 19

white light I3

in printen, 63, 116, 119

white luminance. 138-139.553

in printingpresses. 119

White point. 63.6W6. 133-135.554

in scanners. 6.3. I I ti, 1 1.: 152

white point adaptation, 65,87,167.554

Vector art. 485486

Windows systems

Vector Pro, 419

displaylprint engine. 317

Velvia. Fuji. 150

gamma settings for. 302-303

videoLUT 12:. 553

OS-level color management o n 306,322-324

violet w;lvelengths, 15. SPPalso ultraviolet lighl

printer drivers. 331-333

visiblespeclrum, 7.553. Seenbospecm~m

profile locations. 324

vision. See nlso eye

saving profiles on, 146

color. Se~colorvirion
and color constancy. 46
fiweal. 542
nun-linearilyof. 31-32.41

setting active display profile for. 322-324

source profileassumptions. 328
workflows
assumed-profile. 283-284.295

\'isuai Basic. 475,480

color.managemenl. Seecolor-managementworktlows

visual calibrators. 129,145,553

Computer-To-Plate ICTP). 205

defined, 272

early-biding. 275-276,277,280,501

Wade. Nathan, 477.480

embedded-profile,280-203,284,295

warnings

hvhrid. 2BQ-285

color not normalized, 497

intermediate-space.280

niisslng imagesaurces. 387

LAB-based, 279.497

missinp profile. 286.346349

late-binding, 275.276277.280,46!4.501

profile mismatch. 287.288.350-354

non-color-managed,300

wavelengths
absorption 1's. retlection of, 13-11

one-hput-one-output, 81
PDE 423421,427,455

lnda
between programs, 272,295
within programs, 272.285-295
working space, 554

WYSIWYG myth, 83

x-rays, 7
?(-Rite
Autoscan Spectrophotometer, 171
Digital Swatchbonk. 169

Dl?-41,167
DTP94.131
Spectrofiler, 170
XPress. SeeQuarkXPress
xy chromaticity chart. 42.554

Wplotters. 170-171,554
XYZ. See CIE XYZ

Y
yeflow light, 22. See also CMlX model

Young,Thomas, 22
Young-Helmoltz theory of color vision. I8

z
zone theoryof color, 24

581

583

Real World Color Management, kcond Wition

GATF RHEM Light Indicator

One ofthe most frequent sources of color-management problems is an
incorrect viewing condition. The GATF RHEM Light Indicator, attached
to the inside back cover of this book, provides a quick, easy way to check
your viewing light. It's printed with two inks that produce a metameric
match under D50 lighting, and a mismatch-where you can see obvious
stripes--under non-D50 lighting.
Place it on the print that you're evaluating and check its appearance. If
you can see obvious stripes rather than a solid color, your lighting is unsuitable for making critical color judgements such as the ones you need
to make for profile editing or proofing. ' b o caveats:
b

This target is designed to invalidate bad lighting conditions, not to

validate good ones.

Different DSO-simulators have different spectra, and some may cause

slight striping-this isn't normally a cause for concern. But if you see
obvious stripes, as in the illustration below, your lighting isn't suitable
for critical color work.