Text preview for : EDL-92-1_Device_Independent_Color_Reproduction.pdf part of xerox EDL-92-1 Device Independent Color Reproduction xerox parc techReports EDL-92-1_Device_Independent_Color_Reproduction.pdf
Back to : EDL-92-1_Device_Independe | Home
Device Independent Color Reproduction
Maureen C. Stone
Device Independent Color Reproduction
Maureen C. Stone
EDL-92-1 April 1992 [P92-00055]
C> Copyright 1992 Xerox Corporation. All rights reserved.
Abstract: There are many digital color devices: monitors, printers, film and video
recorders. The goal of device independent color reproduction is to provide a
single, standard representation of color that can be rendered on different media to
produce the same visual result. Such representation must go beyond the CIE
standards to include information about inter-color relationships and media
appearance. The goal of these notes is to provide an overview of current practice
in this area, plus information on colorimetry, additive and subtractive color
systems, and color reproduction practice, all as applied to the problem of device
independent color.
CR Categories and Subject Descriptors: Color Reproduction.
Xerox Corporation
XEROX Palo Alto Research Center
3333 Coyote Hill Road
Palo Alto, California 94304
DEVICE INDEPENDENT COLOR REPRODUCTION
Device Independent Color Reproduction
Maureen C. Stone
Xerox Palo Alto Research Center
Palo Alto, CA 94022
Abstract
Abstract: There are many digital color devices: monitors, printers, film and video
recorders. The goal of device independent color reproduction is to provide a single,
standard representation of color that can be rendered on different media to produce
t.he same visual result.. Such a represent.ation must. go beyond the CIE standards to
include information about inter-color relationships and media appearance. The goal of
these notes is to provide an overview of curren t practice in this area, plus information
on colorimetry, additive and subtractive color systems, and color reproduction practice,
all as applied to the problem of device independent color.
1 Introduction
Classic color reproduction systems tightly couple a specific input device (ie. scanner) to a
specific output device (ie. printer). Color reproduction is then the process of transforming
from the input colors to the output colors. In a device independent color reproduction
system, the input and output transformations are defined relative to a standard represen-
tation for the color. This standard representation is usually based on some form of color
measurelnen t defined by CIE standard colorimetry.
The goal in color reproduction is to duplicate the appearance of the original on the
target media. Colorimetry alone does not provide sufficient perceptual information to au-
tomate this process. That is, simply having a colorimetric definition for each color does not
guarantee an acceptable reproduction. This statelnent has several implications. While cur-
rent colorimetric standards are a significant improvement on the process-specific standards
of the past, more information needs to be added to fully describe the desired appearance
of a color. Furthermore, existing systems must have some way to compensate for this in-
herent limitation; either an interactive interface for making on-line adjustments, or a set of
auxiliary transformations that compensate for appearance differences in different devices or
classes of devices.
Colorimetric color reproduction has been explored in the prepress and computer graphics
communities. In the prepress dOlnain, it is often called "soft proofing" as the focus is to
make a color Inonitor an acceptable proof for some printing process[35, 10, 11, 17]. In the
computer graphics domain, the emphasis has been on reproduction of digital originals[39,
37,32, 33, 42, 20], or desktop publishing style scan-modify-print systems[41, 1, 12]. This is
still a very active field so these notes cannot provide a definitive solution or set of solutions.
Furthermore, many solutions are being pursued in proprietary, industrial environments
XEROX PARC, EDL-92-1, APRIL 1992
DEVICE INDEPENDENT COLOR REPRODUCTION
2
Color Input Color Output
~----------------- p-----------------
Standard
Input Device Colors Rep resentation
Input Gamut
Conversion Mapping
Standard Output
Representation Conversion
________________ J Output Device Colors
________________ J
Figure 1: System diagram for input and output components of a colorimetric color repro-
duction system
rather than scientific ones, so a complete taxonomy is difficult to generate. However, some
common problems and principles have been developed and will be presented here.
The structure of these notes is to describe a canonical colorimetric color reproduction
system, then explore each of the pieces of the system in further depth. The intention is to
motivate the particular emphasis used in describing the many disciplines that contribute to
such a system.
2 System Overview
A canonical systems diagram for the input and output sections of a colorimetric color
reproduction system is shown in figure 1. This diagram highlights the following system
design problems. First, define a standard representation such as CIE tristimulus values or
CIELAB or CIEL UV. Then, develop a way to characterize each device with respect to this
standard. This defines each device's gamut, or set of all possible colors. Finally, define a set
of controls or transformations to compensate for gamut mismatches, differences in viewing
conditions, and other appearance factors. This final step is typically performed only on
output, for it is there that the gamut limitations are completely known.
The next sections provide more detail on each of these three system components. First
is section on colorimetry and the CIE standards for specifying color. The following section
will discuss digital color devices; both the native color representation (additive and subtrac-
tive color), and how to translate between this representation and the standard. Section 5
discusses gamut mapping, or the transformations needed to accommodate different device
characteristics. Finally, some proposed standards for device independent color representa-
tion are reviewed in section 6.
XEROX PARe, EDL-92-1, APRIL 1992
DEVICE INDEPENDENT COLOR REPRODUCTION
3
~P-----------------------------~
~
~
wavelength
Figure 2: Typical spectra for a blue-green color.
3 Colorimetry
Human beings see color when light is reflected off a colored object, transmitted through a
colored filter, or emitted directly as from the phosphors of a color monitor. Light is a physi-
cal quantity, but color depends on the interaction of light with the human visual system and
is thus a psychophysical phenomenon. Research in vision has determined physical proper-
ties of light that are well-correlated with psychophysical properties of color. Colorimetry
is the science of measuring color based on these physical properties, many of which have
been standardized by the Commission Internationale de l'Eclairage (CIE). Because t~ese
standards are well established in both science and industry, they provide a logical basis for
specifying color in a digital system. There are several extensive reference works on color sci-
ence and colorimetrY[46, 3, 21]. These notes will provide only an overview of the important
concepts and terminology.
3.1 Basic Principles of Color Measurement
Color can be defined as the response of an observer to a visual stimulus. This stimulus
can be quantified as the spectrum of the light reaching the eye, that is, the energy of the
light as a function of its wavelength. Figure 3.1 shows a typical color stimulus. The
short wavelengths correspond to blue colors, the midrange to green and yellow, and the
long wavelengths to red. The spectral reflectance of an object is the percentage of the light
energy reflected at each wavelength and the color of an object is most precisely defined as
its spectral reflectance. The .stimulus that reaches the eye, however, is the product of the
spectral reflectance and the spectrum of the light falling on the object, so the perceived
color of an object can never be separated from its illumination.
The Inost basic aspect of a stimulus to quantify is brightness. Human beings see light
when stimulated by electromagnetic radiation between roughly 350nm and 700nm but they
do not see all these wavelengths equally well. The luminous efficiency function defines how
efficiently the eye responds to light at each wavelength. Multiplying a stimulus by the
luminance response curve and integrating produces a measure of how bright a color will
appear called luminance.
The stimulus is a spectrum, but radically different spectra can appear the same color,
an effect that is formally called metamerism. This principle makes it possible to use three
phosphors on a color monitor or three inks in offset printing to produce a wide range of
XEROX PARC, EDL-92-1, APRIL 1992
DEVICE INDEPENDENT COLOR REPRODUCTION
4
~r-----------------------------~
~
e.
(b) (g)
I
(r)
wavelength
Figure 3: An example of metameric spectra. The one on the left is typical for a surface,
the one on the right is three nearly monochromatic light sources.
colors. The number three in these systems is not coincidental; it is a directly related to
the physiology of human vision. Figure 3 shows two different spectra that produce the
same color. However, two objects that match under one set of lights will not necessarily
match under any other set of lights unless the objects have the same spectral reflectance. A
metameric match is one where the objects have different spectral reflectances but happen
to match under a specific illuminant. All common color reproduction processes produce a
metameric match to the original.
3.2 Tristhnulus Values
While a color stimulus is a spectrum, the redundancy introduced by metamerism leads
us to investigate whether color can be uniquely specified with less information. Imagine
a color matching experiment where an observer has three primary lights, (typically red,
green and blue), that can be adjusted to match a given color. Each color produced by
the system can be defined entirely by the power of the primaries. These powers are called
the tristimulus values for that set of primaries. It can be demonstrated that all colors can
be matched (metamerically) by adding three independent light sources, if you allow the
concept of negative light. In the mechanics of the color matching experiment, "negative
light" means shining one of the primaries on the target color. In mathematical terms, it
simply means allowing the tristimulus values to be negative numbers.
It has been empirically proven that different observers with normal color vision will pro-
duce the same tristimulus values for a given color, or, conversely, any stimulus that reduces
to the same tristimulus values will look the same to any standard observer, assuming iden-
tical viewing conditions and observer adaption. By standardizing the primaries, therefore,
we can define any colored light with just three numbers rather than an entire spectrum.
In the terminology of linear algebra, the primaries provide the basis for a three dimen-
sional vector space. Each color is a vector in this space. If two stimuli are combined by
addition (add the spectra wavelength by wavelength), the tristimulus values of the result-
ing color can be obtained by adding the tristimulus values of the stimuli. That is, the
tristimulus values obtained by adding lights can be computed by adding tristimulus values
as vectors. Similarly, scalar multiplication of the stimulus results in a scalar multiple of the
tristimulus values. This formally defines the property of additive mixture and is an inherent
XEROX PARC, EDL-92-1, APRIL 1992
DEVICE INDEPENDENT COLOR REPRODUCTION
5
property of the human visual system.
Given a set of tristimulus values defined relative to one basis (set of primaries), it is
possible to transform them to another basis using a simple, linear transformation. Most
standard primaries are not real light sources but have been designed to have specific prop-
erties such as making the tristimulus values positive for all visible color.
3.3 Color Matching Functions
The color matching experiment requires an observer to make judgements about which com-
bination of primaries colors match a specific color. That is, to generate the tristimulus
values for a color, an observer needs to make a match. It is much more useful to have a
way to define the tristimulus values without using an observer each time.
The color matching functions are a way to capture the essential information about an
observer that is needed to compute tristimulus values from a stimulus. Given a set of
primaries, the color matching functions are defined by a color match to an equal energy
white reference at regular intervals across the visible spectrum. This produces a set of three,
empirical functions of wavelength that define the observer's response. The tristimulus values
for a color stimulus can be computed by integrating the product of the stimulus with the
color matching functions.
The color matching fUllctions are valid only for a specific observer and a specific set
of primaries. However, there is sufficient commonality among human observers that it is
possible to create a standard observer. The standard primaries are arbitrary, though to
actually perform the matching experiment they have to be real light sources
3.4 CIE Standards
The C01nmission Internationale de 1'Eclairage is an international organization that has
standardized a method for computing tristimulus values, denoted X, Y and Z, as a stan-
dard representation for colors. In 1931 the eIE standardized three primaries and a standard
observer to produce color matching functions that are widely used by a variety of industries
to quantify colors. These are specified as tables of data defining functions which are mul-
tiplied by the stimulus and integrated to produce the tristimulus values[46, pp 158-164].
These functions can be seen in figure 4.
Within a reasonable range, multiplying the spectral power distribution by a scalar does
not change the perceived color (effectively, the ambient light changes in brightness). Pro-
jecting the tristimulus values on a plane of constant value factors out the lengths of the
vectors while maintaining their relative positions. The resulting two-dimensional represen-
tation is known as the chromaticity coordinates of the color, and such a projection is called
a chromaticity diagram. The commonly used chromaticity coordinates are x and y, where
x = X/eX + Y + Z) and y = Y/(X + Y + Z).
The familiar eIE chromaticity diagram shown in figure 5 is a plot of x vs. y. Included
on this plot is a horseshoe-shaped curve called the spectrum locus that is defined by the
chromaticity coordinates of the spectral (single frequency) colors. The purple line is the
line connecting the blue with the red end of the spectrum locus. All visible colors lie inside
the region bounded by the spectrum locus and the purple line. The black-body radiation I
curve describes the chrOInaticity coordinates for black body radiators of different color
temperatures. All colors described as "white" lie near the black-body curve. The colors on
the spectrum locus represent pure, saturated colors.
XEROX PARC, EDL-92-1, APRIL 1992
DEVICE INDEPENDENT COLOR REPRODUCTION
6
( X t Y