Journal of Vision (2009) 9(11):5, 1–23 http://journalofvision.

org/9/11/5/ 1

An object-color space
Department of Vision Sciences, Glasgow Caledonian
Alexander D. Logvinenko University, Glasgow, UK

Putting aside metaphorical meanings of the term, color space is understood as a vector space, where lights having the
same color (i.e., subjectively indistinguishable) are represented as a point. The CIE 1931 color space, empirically based on
trichromatic color measurements, is a classical example. Its derivatives, such as CIELAB and sRGB, have been
successfully used in many applications (e.g., in color management). However, having been designed for presenting the
color of self-luminous objects, these spaces are less suitable for presenting color of reflecting objects. Specifically, they can
be used to represent color of objects only for a fixed illumination. Here I put forward a color space to represent the color of
objects independently of illumination. It is based on an ideal color atlas comprising the reflectance spectra taking two
values: k or 1 j k (0 e k e 1), with two transitions (at wavelengths 11 and 12) across the spectrum. This color atlas is
complete; that is, every reflecting object is metameric to some element of the atlas. When illumination alters, the classes of
metameric reflectance spectra are reshuffled but in each class there is exactly one element of the atlas. Hence, the atlas
can uniquely represent the metameric classes irrespective of illumination. Each element of the atlas (thus, object color) is
specified by three numbers: (i) 1 = (11 + 12)/2, which correlates well with hue of object color (as dominant wavelength
correlates with hue of light color); (ii) % = ª11 j 12ª, which correlates with whiteness/blackness; and (iii) ! = ª1 j 2kª, which
correlates with chroma of object color (as colorimetric purity correlates with saturation of light color). Using a geographical
coordinate system, each element of the atlas (thus, each object color) is geometrically represented as a radius vector so that
its length equals !, the latitude and longitude being proportional to % and 1, respectively.
Keywords: color, color vision, color space, color coordinates, cone fundamentals, convexity, object-color solid,
object-color stimulus, color atlas, metamerism, metamer mismatching, optimal reflectance spectra,
spherical chromaticity diagram, chromaticity difference, complementary color
Citation: Logvinenko, A. D. (2009). An object-color space. Journal of Vision, 9(11):5, 1–23, http://journalofvision.org/9/11/5/,
doi:10.1167/9.11.5.

More formally, let us define the color signal in response

Introduction to a light with the spectral power distribution I(1) as a
triplet (81(I), 82(I), 83(I)), where
There has been a long tradition to represent color
geometrically as a point in some space (Kuehni, 2003). Z 1max
Although the term “color space” is often used rather 8i ðIÞ ¼ Ið1Þtð1Þpi ð1Þd1; ð1Þ
loosely, it has an exact meaning in colorimetry (Brainard, 1min
1995; Schanda, 2007), where color is understood as a
class of metameric (i.e., visually indistinguishable) lights
that are represented by their spectral power distributions t(1) is the transmittance spectrum of the ocular media
(Koenderink & Doorn, 2003; Krantz, 1975). As these (e.g., the lens and macular pigment), and pi(1) is the
distributions constitute a positive cone in a properly spectral absorption of the ith photopigment, (i = 1, 2, 3),
chosen functional space, the classes of metameric lights that is assumed to be positive throughout the visible
inherit the linear structure due to Grassmann’s laws spectrum interval [1min, 1max]. I will use throughout the
(Koenderink & Doorn, 2003; Krantz, 1975; Suppes, paper the following values: 1min = 380 nm, 1max = 780 nm.
Krantz, Luce, & Tversky, 1989). As a result, color can Two lights I1(1) and I2(1) are assumed to be metameric
be represented as a point in a vector space. While the if and only if the color signals for these lights are equal,
validity of Grassmann’s laws remains an open issue (Brill that is, 8i(I1) = 8i(I2) for each i.
& Robertson, 2007; Logvinenko, 2006), it is widely Following the established terminology (e.g., Judd &
believed that the physiological basis of metamerism (and Wyszecki, 1975; Stockman & Sharpe, 2007), the product
Grassmann’s laws) is the equality of the cone photo-
receptor outputs. In other words, it is assumed that for two
si ð1Þ ¼ tð1Þpi ð1Þ; ð2Þ
lights to be metameric each type of the photoreceptor
cones should respond equally to these lights (Smith &
Pokorny, 2003; Stockman & Sharpe, 2007; Wyszecki & will be referred to as the ith cone fundamental. The
Stiles, 1982). notations S, M, and L will be used for the response of the
doi: 1 0. 11 67 / 9 . 11 . 5 Received February 3, 2009; published October 5, 2009 ISSN 1534-7362 * ARVO

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 2

cone fundamentals with the peak sensitivity in the short-, natural one-to-one relationship between two object-color
middle-, and long-wave range of the visible spectrum, solids induced by different illuminants. To show what is
respectively. meant by natural relationship, consider a particular case
When one represents a color signal as the point in the when there is such. Specifically, restrict ourselves to the
3D space with the Cartesian coordinates (S, M, L), the boundaries of the two object-color solids obtained for
color signals from all the lights form a convex cone in this different illuminants.
space (referred to as the color signal cone). Each color While for every spectral reflectance function mapping
signal represents a class of metameric lights. As meta- into the object-color solid interior, there is an infinite
meric lights are assumed to have the same color, this cone number of metameric spectral reflectance functions, there
can be considered as a geometrical representation of color is no metamerism on the object-color solid boundary.1 To
in the 3D space that will be referred to as the SML color be more exact, there is no spectral reflectance function
space. metameric to the so-called optimal spectral reflectance
Although, in fact, such a color space was designed for functions, that is, those which map to the boundary of the
representing color of lights (i.e., self-luminous objects), it object-color solid. For example, there are infinitely many
has also been used to represent the color of reflecting spectral reflectance functions mapping to the center of the
objects. Specifically, a surface with the spectral reflec- object-color solid, one of them being a function x(1) K
tance function x(1) illuminated by a light with the spectral 0.5, which takes 0.5 at every wavelength within the visible
power distribution I(1) produces a color signal the ith spectrum (written x0.5(1)). However, only one spectral
component of which is reflectance function maps to the north (respectively,
south) pole of the object-color solid, namely, the “perfect
Z 1max reflector” x(1) K 1 (respectively, the “ideal black” x(1) K
8i ðxÞ ¼ xð1ÞIð1Þsi ð1Þd1: ð3Þ 0), which takes 1 (respectively, 0) at every wavelength
1min within the visible spectrum.
It seems natural to map one object-color solid boundary
It seems safe to assume that when the illumination is fixed onto the other as follows. Let (S, M, L) be a point on the
(i.e., all the objects are lit by the same light), reflecting boundary of the object-color solid obtained for the
objects producing the same color signal (referred to as illuminant I(1). There is just one optimal spectral
metameric reflecting objects) will appear as having the reflectance function that produces this color signal.
same color. Therefore, under constant illumination each Denote it xopt(1). As the set of optimal spectral reflectance
color signal will represent a class of metameric spectral functions remains the same for a very broad class of
reflectance functions. The color signals produced by all illuminants (see Optimal spectral reflectance functions and
the reflecting objects under the same illuminant form a object-color solid section), xopt(1) will map to some point
closed convex volume (referred to as the object-color in the boundary of the object-color solid obtained for the
solid) in the color signal cone (Koenderink & Doorn, other illuminant, IV(1), say, (SV, MV, LV). A one-to-one map
2003; Luther, 1927; Maximov, 1984; Nyberg, 1928; (S, M, L) Y (SV, MV, LV) establishes natural correspondence
Schrodinger, 1920/1970; Wyszecki & Stiles, 1982). between the object-color solid boundaries because it puts
At the first glance, the object-color solid seems to in correspondence color signals which represent the same
provide the geometrical representation of object color in classes of metamerism (though, they are singletons under
the SML color space the same way as the color signal cone both the illuminants). Note that this map is non-linear.
does for light color. However, the analogy is rather Extending this map to the interiors of the object-color
superficial because such a representation depends on the solids results in a many-to-many map because of metamer
illuminant. When the illuminant alters, metameric spectral mismatching. Yet, if we replace each metameric class
reflectance functions might cease to be metameric. This with just one of its members, it becomes possible.
phenomenon is known as metamer mismatching (Wyszecki Specifically, a one-to-one relationship between the classes
& Stiles, 1982). Conversely, spectral reflectance functions of metameric reflectances produced by different illumi-
that produce different color signals under one illuminant nants can be established through a color atlas.2
may become metameric under the other. So, after an In the present context, any sample of non-metameric
illumination change, some spectral reflectance functions spectral reflectance functions will be referred to as a color
will remain in the same metameric class while others will atlas. Each element of a color atlas represents a metameric
fall into different classes. Therefore, the reflecting object class to which it belongs. A color atlas that represents all
metamerism as defined above is different for different the classes of metamerism will be called complete. As
illuminants. every optimal spectral reflectance function makes a
Because of metamer mismatching, there is no natural metameric class on its own, a complete color atlas includes
way to establish a correspondence between the metameric all the optimal spectral reflectance functions.
classes produced under one illumination and those Being a color atlas under one illumination, a set of
produced under another. In other words, there is no spectral reflectance functions may cease being a color

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 3

atlas under the other illumination because of metamer the color atlas. Such an illuminant-induced color stimulus
mismatching. Therefore, when using a notion of color shift is discussed in the Illuminant- and observer-induced
atlas, one has always to specify an illuminant for which color stimulus shifts section.
the given set of spectral reflectance functions is a color
atlas. When a sample of spectral reflectance functions is a
color atlas for the whole family of illuminants, we will say Illuminant invariant color atlas
that the color atlas is invariant with respect to this
illuminant family. To summarize, given a family of
illuminants, a complete illuminant-invariant color atlas Optimal spectral reflectance functions and
is a set of spectral reflectance functions such that (i) each object-color solid
class of metamerism contains exactly one element of this
set under any illuminant in the family, and (ii) any two The optimal spectral reflectance functions can take only
spectral reflectance functions from the set do not become two values: 0 or 1 (Schrodinger, 1920). For example, the
metameric under any illuminant in the family. spectral reflectance function x(1) K 1 is optimal. Another
As there is a one-to-one map between an illuminant- example of spectral reflectance functions that can be
invariant color atlas and a set of classes of metamerism optimal is a step function (i.e., a piecewise constant
under any illuminant (from the family in question), a function taking 1 or 0) with transition at the wavelength
complete illuminant-invariant color atlas induces a one-to- 11:
one map between the classes of metameric reflectances
(thus, between the color solids) induced by different

0; if 1min e 1 G 11 ;
illuminants. Indeed, let us put in correspondence those x1 ð1; 11 Þ ¼ ð4Þ
metameric classes that contain the same element of the 1; if 11 e 1 e 1max :
color atlas.
It follows that a complete color atlas, invariant with Note that x(1) K 1 is a particular case of x(1; 11) when
respect to some illuminant family, uniquely represents all 11 = 1min. Given 12 Q 11, the step function with transition
the metameric classes, thus, all the object colors under any wavelengths 11 and 12
illuminant in this family. Therefore, any spatial represen-
tation of such an atlas will be also a spatial representation x2 ð1; 11 ; 12 Þ ¼ x1 ð1; 11 Þjx1 ð1; 12 Þ; ð5Þ
of the object colors for any illuminant in this family.
In this article, I introduced a particular complete color can also be an optimal spectral reflectance function.
atlas invariant with respect to illuminants with positive Equation 5 will be referred to as the rectangular spectral
spectral power distribution (Illuminant invariant color atlas reflectance function. Note that x1(1; 11) is a particular
section). The Object-color space section shows (i) that this case of x2(1; 11, 12) when 12 = 1max. Generally, the step
color atlas can be geometrically represented as the unit function with transitions at the wavelengths 1min G 11 G 12
ball in a 3D space and (ii) that this space can be used for G I G 1m G 1max
representing all the reflecting objects illuminated by lights
with positive spectral power distribution. Such represen- X
m
tation will be independent of illumination in the sense that xm ð1; 11 I1m Þ ¼ ðj1Þij1 x1 ð1; 1i Þ; ð6Þ
the coordinate system in the 3D space will remain the i¼1
same for the illuminations in question.
A complete illuminant-invariant color atlas can be used can be an optimal spectral reflectance function.
to investigate how the color of a particular object For any optimal spectral reflectance function xopt(1), the
(represented by its spectral reflectance function) trans- spectral reflectance function 1 j xopt(1) has proven to be
forms with illumination. The problem of color trans- also optimal. Particularly, if xm(1; 11, I, 1m) is an optimal
formation induced by illumination can be reduced to the spectral reflectance function, the spectral reflectance func-
following two. Firstly, the color atlas itself may change its tion 1 j xm(1; 11, I, 1m) is also optimal. A pair of optimal
color appearance when the illumination alters. For spectral reflectance functions xopt(1) and 1 j xopt(1) will be
example, while a white surface remains clearly recogni- called complementary. The optimal spectral reflectance
zable as white under various chromatic illuminations, its functions x(1) K 1 and xm(1; 11, I, 1m) for any integer m
appearance under, say, yellow light differs from that under will be referred to as of type I, the optimal spectral
blue light. This is a matter of experimental investigation reflectance functions x(1) K 0 and 1 j xm(1; 11, I, 1m) as
that goes beyond the scope of the present article. of type II.
Secondly, the color of an object may change because the There is a general belief (Koenderink & Doorn, 2003;
object may move from one class of metamerism to the MacAdam, 1935; Schrodinger, 1920; Wyszecki & Stiles,
other because of an illumination change. As a result, 1982) that the optimal spectral reflectance functions are
the object will become metameric to a different element of step functions with not more than two transitions across

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 4

the visible spectrum. (It will be referred to as the two- the convex hull of the spectral cone. The roots of Equation 8
transition assumption). However, this is not, strictly are the points where the spectral curve intersects the plane
speaking, true. In fact, the number of transitions depends defined by the equation k1S + k2M + k3L = 0.
on the shape of the cone fundamentals (Maximov, 1984; The shape of a projection of the spectral curve to any
West & Brill, 1983). More specifically, a theorem has plane not containing the origin is indicative of the
been proved (Logvinenko & Levin, 2009) from which it maximum number of transitions for the optimal spectral
follows that, for continuous cone fundamentals s1(1), reflectance functions. For example, let us choose a plane
s2(1), and s3(1), and an illuminant with integrable spectral not containing the origin and consider a contour made by
power distribution I(1), if 11, I, 1m are the only roots of the spectral curve projection to this plane (referred to as
the following equation the spectrum locus). Let us complete this contour with the
interval joining the ends of the spectrum locus (the purple
k1 s1 ð1ÞIð1Þ þ k2 s2 ð1ÞIð1Þ þ k3 s3 ð1ÞIð1Þ ¼ 0; ð7Þ interval). If the resultant contour (referred to as the
completed spectral contour) is convex, then, as noted by
where k1, k2, and k3 are arbitrary real numbers (at least West and Brill (1983), the two-transition assumption
one of which is not equal to zero), then a step function holds true.
(Equation 6) with transitions at the wavelengths 11, I, 1m As it is safe to assume that the transmittance spectrum
will be an optimal spectral reflectance function. of the ocular media is everywhere positive on [1min, 1max]
If the illuminant is such that I(1) 9 0 for 1min e 1 e (i.e., t(1) 9 0 in Equation 2), Equation 8 amounts to
1max, Equation 7 is equivalent to the following
k1 p1 ð1Þ þ k2 p2 ð1Þ þ k3 p3 ð1Þ ¼ 0: ð9Þ
k1 s1 ð1Þ þ k2 s2 ð1Þ þ k3 s3 ð1Þ ¼ 0: ð8Þ Therefore, the lens and macular pigment do not affect the
set of optimal spectral reflectance functions. Generally,
Therefore, the shape of the illuminant spectral power any pre-receptor filter with positive transmittance spec-
distribution does not affect the set of optimal spectral trum cannot affect the set of optimal spectral reflectance
reflectance functions unless it takes zero values within the functions. It is fully determined by the photopigment
visible spectrum interval. spectral absorption.
The roots of Equation 8 have a simple geometrical The absorbance spectra of the cone photopigments are
meaning. Recall that the color signal of the monochro- known to be described by a relatively simple analytical
matic light with the wavelength 1 is given by (s1(1), s2(1), expression (Lamb, 1999). The spectral sensitivities of the
s3(1)). When 1 runs through the interval [1min, 1max], a three cone photopigments calculated from the photopig-
point (s1(1), s2(1), s3(1)) makes a curve in the SML color ment optical density template put forward by Govardovskii
space which is usually referred to as the spectral curve with coauthors (Govardovskii, Fyhrquist, Reuter, Kuzmin,
(Figure 1). The cone through the spectral curve is referred & Donner, 2000) are presented in Figure 2. Note that
to as the spectral cone (Figure 1). It represents color of the although the spectral responses, especially the S-cone’s,
monochromatic lights. The color signal cone proves to be are very low in the long wavelength end of the spectrum,
they differ from zero. (This property is of great
theoretical importance. For instance, letting p1(1) = 0
for 1min 9 1V Q 1 Q 1max in Equation 5, will result in
metamerism in the object-color solid boundary.) The cone
fundamentals based on these photopigment spectra are
presented in Figure 3.
The completed spectral contour in the unit plane of the
SML space based on these cone fundamentals is shown in
Figure 4. As one can see, it is not convex. Firstly, the
short wavelength end of the spectrum locus makes a well-
pronounced “beak.” Secondly, at the opposite end the
spectrum locus reverses its direction making a self-
intersection. This results in a so-called hue-reversal effect
described by Brindley (1955). All this leads to that the
purple interval lies inside the color signal cone.
Because of concavity of the completed spectral contour
in Figure 4, there exist many lines intersecting the
spectrum locus in Figure 4 at more than two points. A
plane through any of these lines and the origin will
Figure 1. The spectral curve (red line) and the spectral cone through intersect the spectral curve at more than two points.
it, evaluated for the cone fundamentals depicted in Figure 3. Therefore, the concavity of the completed spectral contour

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 5

Figure 2. Spectral sensitivity shapes, scaled to unity peak, for the Figure 4. The completed spectral contour based on the cone
photopigments with peak sensitivity 1max at 420, 530, and 560 nm. fundamentals presented in Figure 3. The blue line is the spectrum
Each curve is expressed as 1 j 10jDG(1), where G(1) is the locus; the purple interval is marked with corresponding color.
normalized photopigment optical density spectrum suggested by
Govardovskii et al. (2000) and D is the peak photopigment optical generally believed to be the linear transformation of the
density. The value of D was assumed to be 0.3 (Govardovskii, cone fundamentals (e.g., Judd & Wyszecki, 1975; Stockman
personal communication). & Sharpe, 2007). If this were the case, then the optimal
spectral reflectance functions derived from the color
entails the existence of optimal spectral reflectance matching functions should include spectral reflectance
functions with more than two transitions, thus the failure functions with more than two transitions. However, the
of the two-transition assumption. completed spectral contour (in the unit plane) derived from
It follows from Equation 8 that any three linear the color matching functions adopted by the CIE as the
independent functions that are the linear transformation standard colorimetric observer (Figure 5) is convex. This
of the cone fundamentals will determine the same set of indicates that for this observer the two-transition assump-
optimal spectral reflectance functions as the cone funda- tion holds true.
mentals themselves. The color matching functions are The completed spectral contour plotted for the cone
fundamentals proposed by DeMarco, Pokorny, and Smith
(1992) is also convex (Figure 6). The completed spectral

Figure 3. Cone fundamentals evaluated as the photopigment

spectral sensitivities depicted in Figure 2 times the spectral
transmittance of the lens and macular pigment as given by Figure 5. The completed spectral contour based on the CIE 1931
Wyszecki and Stiles (1982, p. 587). color matching functions (Wyszecki & Stiles, 1982).

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 6

are essential differences between them. More specifically,

if the major difference between them were the pre-
receptor filter transmittance, the spectrum loci determined
by them would have been either all convex or all concave.
However, as follows from Figures 5–7, this is not the case.
Furthermore, because of the qualitative difference
between the optimal spectral reflectance set induced by
the cone fundamentals based on the photopigment spectra
and that derived from the color matching functions, it is
unlikely that the latter can be represented as a linear
transform of the former. Therefore, if Govardovskii’s
template describes accurately enough the photopigment
absorption in human cones, then perhaps the decision
about match in the color matching experiments is related
to the cone outputs in a more complicated way than
hitherto assumed.
As the two-transition assumption is taken for granted in
the color literature, of great theoretical and practical
Figure 6. The completed spectral contour based on the DeMarco, importance is how large is the difference between the real
Smith, and Pokorny cone fundamentals (DeMarco et al., 1992). object-color solid and that obtained under the two-
transition assumption. Specifically, let us call the volume
contour based on the cone fundamentals developed by in the SML color space made by the step functions with
Stockman and Sharpe (2000) is not, strictly speaking, not more than two transitions the regular object-color
convex (Figure 7). A departure from convexity similar to solid (Figure 8). It is a closed volume nested into the real
that in Figure 4 takes place at the long-wavelength end of object-color solid.
the spectrum locus. However, no beak at the short- The curves in Figure 8 are the lines of constant 11 and
wavelength end of the spectrum locus is present in Figure 7. 12. Specifically, each such line represents the color signals
It follows that many optimal spectral reflectance functions induced by the spectral reflectance functions x2(1; 11, 12)
with the third and fourth transitions at the short-wavelength (Equation 5) when either 11 or 12 is fixed. Two coordinate
end of the visible spectrum interval will be lost when using lines, corresponding to the two particular cases when
these cone fundamentals. either 11 = 1min or 12 = 1max, will be referred to as the
Therefore, contrary to the view that the different sets of meridians of the regular object-color solid. When 12 =
the cone fundamentals put forward by various authors 1max, the meridian (referred to as the main meridian) is an
differ from each other mainly by the spectral form of the image of the spectral reflectance functions x1(1; 11)
pre-receptor filter (Smith, Pokorny, & Zaidi, 1983), there (Equation 4). When 11 = 1min, the meridian (referred to
as the opposite meridian) is an image of the optimal
spectral reflectance functions 1 j x1(1; 12). The main
and opposite meridians are marked with red and blue,
respectively, in Figure 2, the main meridian being almost

Figure 8. Regular object color solid as evaluated for the cone

Figure 7. The completed spectral contour based on the Stockman fundamentals depicted in Figure 3 and the standard CIE illuminant
and Sharpe 2-deg cone fundamentals (Stockman & Sharpe, D65. Axes are in normalized unit. The red and blue lines stand for
2000). the main and opposite meridians respectively. See text.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 7

hidden. The space between the coordinate lines corre- that the optimal color atlas invariance does not mean that
sponds to (1max j 1min) / 30. As can be seen, the the optimal metamer of a particular spectral reflectance
coordinate lines are not evenly distributed across the should be illuminant independent. In fact, many spectral
boundary surface. They can be made equally spaced after reflectance functions have different optimal metamers
reparameterizing the wavelength interval (see Wavelength under different illuminants. An alteration of the optimal
reparameterization section). metamer caused by an illuminant change reflects a shift of
It turns out that the maximal difference between the real this reflectance from one class of metamerism to the other
and regular object-color solids is very small (see Optimal because of the illuminant change. In other words, an
metamers and optimal color atlas section). Therefore, the alteration of the optimal metamer for a particular object
regular object-color solid, which is a great deal easier to indicates the objective alteration of the color of this object
evaluate from the computational point of view, can be (see Illuminant-induced color stimulus shift section).
taken as an approximation to the real object-color solid in Like optimal spectral reflectance functions, the optimal
many practical applications. metamers will be divided into two groupsVof types I and
IIVdepending on shape of their spectral waveform.
Specifically, the optimal metamer (Equation 10) will be
Optimal metamers and optimal color atlas assigned the same type as the optimal reflectance xopt(1)
(see Optimal spectral reflectance functions and object-
Given an illuminant I(1) 9 0, for each spectral color solid section). Generally, a spectral reflectance
reflectance function x(1) there is (i) a unique optimal function will be referred to as of type I (respectively, II)
spectral reflectance function xopt(1) and (ii) a unique if its optimal metamer is of type I (respectively, II).
number 0 e ! e 1 such that spectral reflectance function For each optimal metamer x(1) m x0.5(1), there is a
unique optimal metamer that differs from x(1) only in
ð1j !Þx0:5 ð1Þ þ !xopt ð1Þ; ð10Þ type. Two such optimal metamers will be called comple-
mentary. They map to the points symmetrical with respect
is metameric to x(1). (Recall that x0.5(1) K 0.5.) to the object-color solid center.3 For the sake of general-
The set of all the spectral reflectance functions ity, the optimal metamer x0.5(1) will be assumed to be
(Equation 10) is a complete color atlas. Indeed, given an complementary to itself. The colors of two complemen-
arbitrary x(1), the optimal spectral reflectance xopt(1) that tary optimal metamers will be called complementary. For
determines the spectral waveform of (Equation 10) can be each object color, there is only one complementary color.
found from the condition that the color signal (81(xopt), As the object-color complementarity has been defined
82(xopt), 83(xopt)) is the boundary point of the object-color in terms of the optimal color atlas, it is independent of
solid lying on the same radius as (81(x), 82(x), 83(x)). illumination. Note, however, that two spectral reflectance
As the spectral waveform of (Equation 10) is fully functions that have complementary colors under one
specified by the optimal spectral reflectance functions illuminant may prove to have non-complementary colors
xopt(1), this atlas will be referred to as the optimal color under the other illuminant (see Illuminant-induced color
atlas. A member (Equation 10) of the optimal color atlas stimulus shift section).
metameric to x(1) will be called the optimal metamer of Denoting x~opt, the optimal spectral reflectance function
x(1). It takes two values: 0.5(1 T !), the difference complementary to an optimal spectral reflectance function
between which being !. Therefore, ! indicates how much xopt (i.e., x~opt = 1 j xopt), we get x0.5(1) + !(x~opt(1) j
the spectral reflectance function (1 j !) x0.5(1) + !xopt(1) = x0.5(1)) = x0.5(1) j !(xopt(1) j x0.5(1)). This equation
x0.5(1) + !(xopt(1) j x0.5(1)) deviates from the level of allows us to extend formally the chromatic amplitude
0.5. I will call the quantity ! the chromatic amplitude of range treating an optimal metamer with negative chro-
the optimal metamer (Equation 10). It can be evaluated as matic amplitude as that of positive chromatic amplitude of
the same magnitude and complementary spectral wave-
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi form. For example, the optimal spectral reflectance
uP
u ð8 ðxÞj 8 ðx ÞÞ2
3
function x(1) K 0 can be interpreted as the optimal
u i i 0:5
spectral reflectance function x(1) K 1 with negative
!¼u
tP
i¼1 : ð11Þ
3
chromatic amplitude j1. Hence, using negative chromatic
ð8i ðxopt Þj 8i ðx0:5 ÞÞ2
i¼1 amplitude, we can restrict our consideration to only
optimal spectral reflectance functions of type I. In this
When chromatic amplitude ! goes from 0 to 1, the color case, the spectral reflectance functions of negative
signal of Equation 10 moves from the center to the chromatic amplitude will be exactly the set of the spectral
boundary of the object-color solid remaining on the same reflectance functions of type II. I will use the term purity
radius. for the absolute value of chromatic amplitude.
The optimal color atlas proves to be invariant with The optimal color atlas is, generally, different for
respect to illuminants with positive spectral power different cone fundamentals (thus, color matching func-
distribution (see Appendix A2). It should be kept in mind tions) because it is determined by the set of the optimal

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 8

object-color stimuli (i.e., the optimal spectral reflectance can be specified by only three numbers (11, 12, and !),
functions). For example, when the two-transition assump- using the rectangle color atlas will allow a three-
tion holds true, the optimal object-color stimuli are the coordinate representation of the optimal color atlas.
rectangular spectral reflectance functions. When the two- There are, of course, many other improper color atlases
transition assumption fails there are optimal spectral that provide a three-coordinate representation of the
reflectance functions with more than two transitions. optimal color atlas. For example, given three spectral
Therefore, in this case the optimal color atlas also includes reflectance functions (which do not metameric under any
spectral reflectance functions with more than two transitions. illuminant with positive spectral power distribution) x1(1),
As the set of the optimal object-color stimuli, in fact, x2(1), and x3(1), any element of the optimal color atlas
generates the optimal color atlas, it will play an important x(1) is metameric to some linear combination of these
role in our analysis. I will refer to the set of the optimal functions: k1x1(1) + k2x2(1) + k3x3(1). The weights k1, k2,
object-color stimuli as the chromatic base of the optimal and k3 can be considered as linear coordinates of the
color atlas. It will be shown in the Perceptual correlates of element of the optimal color atlas. When x(1) runs over
the !%1 color descriptors section that the chromatic base the whole optimal color atlas, the resultant set of the
specifies the object-color hues. In the case of the two- functions, {k1x1(1) + k2x2(1) + k3x3(1)}, makes an
transition assumption, the chromatic base comprises only improper color atlas that will be referred to as a linear
rectangular spectral reflectance functions. It will be called color atlas based on the basis functions x1(1), x2(1), x3(1).
the rectangle chromatic base. It should be borne in mind that a coordinate representa-
It is important to note that even when the two-transition tion of the optimal color atlas based on an improper color
assumption fails the rectangle chromatic base can be used atlas depends, generally, on both illumination and cone
to represent the optimal color atlas. More specifically, fundamentals. Specifically, a linear color atlas, strictly
given a spectral reflectance function x(1) producing the speaking, will provide different coordinates for an optimal
color signal (81(x), 82(x), 83(x)) (see Equation 3), consider object-color stimuli under different illuminations unless
simultaneous equations this stimulus is a basis function. Direct calculations show
Z 1max that the variation of the linear coordinates, for example,
ðx0:5 ð1Þ þ !ðx2 ð1; 11 ; 12 Þj x0:5 ð1ÞÞÞ for a perfect reflector can be quite large even for natural
1min ð12Þ illuminants. Therefore, linear color atlases do not, gene-

Ið1Þsi ð1Þd1 ¼ 8i ðxÞ; rally, provide the coordinate representation of the optimal
color atlas that is constant with respect to illumination.
where x2(1; 11, 12) is a rectangular spectral reflectance While the three-coordinate representation of the optimal
function with the transition wavelengths 11 and 12 color atlas based on the rectangle color atlas also depends
(Equation 5) (i = 1,2,3). These equations can always be on illuminant, this dependence proves to be very small.
resolved with respect to !, 11, and 12. If the color signal More specifically, the set of optimal spectral reflectance
(81(x), 82(x), 83(x)) lies within the regular object-color functions has, firstly, been evaluated for the cone
solid, the solution will be such that ª!ª e 1. For those fundamentals based on the photopigment spectra (Figure 3).
color signals that are between the boundary surfaces of the Many of rectangular spectral reflectance functions (with
real and regular object-color solids, we will get ª!ª 9 1. two transitions) have proved to be the optimal object-color
In this case (1 + !) / 2 exceeds 1, and (1 j !) / 2 is stimuli. For these, the coordinates 11, 12, and ! are,
negative. While such profiles cannot be treated as spectral obviously, independent of illumination. Still, a number of
reflectance functions, I will consider them as improper optimal spectral reflectance functions with three and four
spectral reflectance functions. Taking them into consid- transitions have been found. These spectral reflectance
eration allows one to represent the optimal color atlas by functions are, generally, metameric to different rectangu-
rectangle spectral reflectance functions with not more than lar metamers under different illuminants. The rectangle
two transitions. metamers have been calculated for the optimal spectral
So, even when the two-transition assumption is not reflectance functions with more than two transitions for a
satisfied, for each spectral reflectance function x(1), there few illuminants. Their purity has been found to fall into
is a (perhaps, improper) spectral reflectance function the narrow band between 1 and 1.01. Hence, if one
x0.5(1) + !(x2(1; 11, 12) j x0.5(1)) that meets Equation 12. measures the distance between the boundaries of the real
It will be called the rectangular metamer of the object- and regular object-color solids along radii, then the
color stimulus x(1). The rectangular metamer will be maximum distance does not exceed 1%. It follows that
referred to as improper if ª!ª 9 1. Although the set of all the regular object-color solid is a good approximation to the
the rectangle metamers for all the object-color stimuli real object-color solid. Furthermore, the variability of the
cannot be a proper colors atlas (as it contains improper coordinates 11, 12, and ! for the optimal spectral
spectral reflectance functions), I will refer to it as the reflectance functions with more than two transitions has
rectangle color atlas. To avoid confusion, I will call been found to be negligibly small. Therefore, while
improper those color atlases that contain improper being of theoretical importance, the dependence of the
spectral reflectance functions. As a rectangular metamer coordinates 11, 12, and ! on illuminant is unlikely to have

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 9

Direct calculation shows that in this case (i.e., for

rectangular spectral reflectance functions of type II), the
spectral bandwidth is

% ¼ ð1max j 1min Þj k11 j12 k; ð13Þ

and the central wavelength is given by
(
 0:5ðð11 þ 12 Þ þ ð1max j 1min ÞÞ; if ð11 þ 12 Þ G ð1max j 1min Þ;
1 ¼
0:5ðð11 þ 12 Þ j ð1max j 1min ÞÞ; if ð11 þ 12 Þ 9 ð1max j 1min Þ:
ð14Þ
It follows that the central wavelengths of complementary
rectangular spectral reflectance functions lie on the
opposite ends of the diameter of the visible spectrum
circle, the sum of their spectral bandwidths being equal to
Figure 9. Rectangular metamer. It is specified by its purity, !; 1max j 1min. This is in line with our intuition of
 complementarity.
spectral bandwidth, %; and central wavelength, 1 . See text.
The rectangular spectral reflectance function x(1) K 0
has zero spectral bandwidth; however, central wavelength
a significant impact in most practical applications. Thus, is not defined for it. Likewise, for x(1) K 1, the spectral
the rectangle color atlas can be use to provide the three- bandwidth % = 1max j 1min, the central wavelength being
coordinate representation of the optimal color atlas which not defined.
is quite easy to compute. Conversely, a spectral bandwidth and a central wave-
length uniquely determine the rectangular spectral reflec-
tance function. Specifically, given 0 G % G (1max j 1min)
Rectangle color atlas and rectangular  
and 1min G 1 G 1max, a pair (%, 1 ) uniquely defines the
metamers rectangular spectral reflectance function, the transition
wavelengths 11 and 12 of which are given by
The rectangle color atlas is determined by the set of
rectangular spectral reflectance functions. Each such
 
function is completely specified by the transition wave- 1 j 0:5%; if ð1 j 0:5%Þ 9 1min ;
11 ¼   j 0:5%Þ G 1 ;
lengths, 11 and 12, and its type. If it is of type I, it takes up 1 j 0:5% þ ð1max j 1min Þ; if ð1 min
1 within the interval [11, 12], its complementary spectral ð15Þ
reflectance function taking up 1 exactly outside it. The
interval [11, 12] can be specified by its width % = ª11 j

  þ 0:5%Þ G 1 ;
12ª and the position of its center 1 = (11 + 12)/2. I refer to 1 þ 0:5%; if ð1
 12 ¼  
max
% as the spectral bandwidth and 1 as the central 1 þ 0:5% j ð1max j 1min Þ; if ð1 þ 0:5%Þ 9 1max :
wavelength for a rectangular spectral reflectance function
of type I. Thus, the rectangular metamer of a spectral ð16Þ
reflectance function of type I is completely specified by
the three numbers: purity, spectral bandwidth, and central
wavelength (Figure 9).
To generalize these notions for rectangular spectral
reflectance functions of type II, I will make no difference
between 1min and 1max, turning the visible spectrum
interval into the visible spectrum circle (Figure 10). Two
points, 1min e 11 G 12 e 1max, determine two comple-
mentary arcs on this circle. The rectangular spectral
reflectance function of type I takes 1 on the arc where
11 e 1 e 12. The spectral bandwidth and central
wavelength are the length and the center of this arc,
respectively. Its complementary spectral reflectance func-
tion (of type II) takes 1 on the complementary arc. It is the
length and center of this complementary arc that are
taken, by definition, as the spectral bandwidth and central Figure 10. Rectangular metamer as defined on the visible
wavelength for spectral waveforms of type II. spectrum circle. See text.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 10

differ in purity, and thus in apparent purity. For example,

the purity of the gray paper with the spectral reflectance
x0.5(1) is zero whereas that of the white paper (x(1) K 1) is
maximal (i.e., equal to 1). The apparent difference
between various spectral reflectances that have the same
purity will be referred to as that in chromaticity. In other
words, the chromatic base specifies chromaticity of
object-color stimuli.
It was found that the papers of the same Munsell Hue
had rather close central wavelengths. This tendency is
illustrated by Figure 12 where the papers with maximal
chroma (for each value) from five pages of the Munsell
book are displayed. When the papers of the same Munsell
Figure 11. Regular object-color solid same as in Figure 8 except Hue are of the same type, they are arranged in nearly
that the coordinate lines are based on spectral band and central vertical rows that are well separated and ordered in the
wavelength. horizontal direction. However, there is a considerable
misalignment between the papers of the same Munsell
By definition, zero spectral bandwidth (% = 0) defines the Hue of different type. For example, three blue and three
optimal spectral reflectance function x(1) K 0. Likewise, purple papers of type II are shifted clockwise in the visible
the spectral bandwidth % = 1max j 1min defines the spectrum circle relative to the blue and purple papers of
optimal spectral reflectance function x(1) K 1. type I. One yellow and two red papers of type II are

Any pair (%, 1 ) defines a unique location on the regular shifted counterclockwise. This misalignment will be
object-color solid boundary surface. The rectangular discussed below (see Wavelength reparameterization
spectral reflectance functions with the same central section).
wavelength map to a closed curve on this surface. So do While the symbols of the same color are also ordered in
the rectangular spectral reflectance functions with the the vertical direction in Figure 12 (the larger Munsell
same spectral bandwidth. These curves can serve as the Value, the broader the spectral bandwidth), there seems to
coordinate web on the regular object-color solid boundary
surface (Figure 11).
To summarize, the rectangular metamer of a spectral
reflectance function is specified by the three numbers:
purity, spectral bandwidth, and central wavelength. I will
call the purity, spectral bandwidth, and central wavelength
of the rectangular metamer of a spectral reflectance
function the !%1 color descriptors of the spectral
reflectance function. The !%1 color descriptors will be
used to represent spectral reflectance functions in a 3D
space (Object-color space section).

Perceptual correlates of the !%1 color

descriptors

The !%1 color descriptors for the spectral reflectance

functions of 1600 Munsell papers (retrieved from http://
spectral.joensuu.fi/databases) were evaluated for the CIE
illuminant D65 using the cone fundamentals based on the
photopigment spectra (Figure 3). The papers having the
same spectral waveform and differing only in purity were Figure 12. Color descriptors of the Munsell papers of maximal
found to have similar color appearance. Specifically, they chroma from five pages (10B, 10G, 10Y, 10R, and 10PB) after
differ in a perceptual dimension, which will be referred to wavelength reparameterizing. Each paper is plotted as a point

as apparent purity. Intuitively, apparent purity indicates with the Cartesian coordinates % (ordinate) and 1 (abscissae). The
the strength of the chromatic quality (specified by the color of the symbols resembles the hue in the page. The red and
spectral waveform of the rectangular metamer). It should blue lines represent the main and opposite meridians, respec-
be mentioned, however, that apparent purity is different tively. A triangle made by these lines and the horizontal axis
from Munsell Chroma. Indeed, having zero Munsell represents the reflectance spectra of type I. The points above the
Chroma, all achromatic Munsell papers, nevertheless, blue and red lines represent those of type II.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 11

not contributing into producing color signal. Figure 13

presents the length of the opposite meridian of the regular
object-color solid (the blue coordinate line in Figure 8)
as a function of wavelength. As one can see, the length
practically does not differ from zero for 11 G 405 nm and
reaches its maximum approximately at 650 nm. To be more
exact, the length of the fragments of the opposite meridian
for 380 e 11 e 405 is less than 0.01 and more than 0.99 for
380 e 11 e 650. Therefore, given a 1% accuracy, the
values 405 and 650 nm can be taken as effective ends of
the visible spectrum interval.
Alternatively, a rather drastic solution will be to
reparameterize the visible spectrum interval so as to make
the wavelength contribution into producing the color
signal more uniform. For instance, one can take length
of the meridian in the regular object-color solid as a new
Figure 13. Regular-object-color-solid-meridian length in normal- parameter. The length A of the arc of the opposite
ized units. meridian between the points (s1(1min), s2(1min), s3(1min))
and (s1(1), s2(1), s3(1)) (1min e 1 e 1max) is given by

Z 1 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
be no simple relationship between spectral bandwidth and Að1Þ ¼ ðs1 ð2ÞÞ2 þ ðs2 ð2ÞÞ2 þ ðs3 ð2ÞÞ2 d2:
Munsell Value. 1min
Alternatively, it is the strength of the white and black ð17Þ
hue components (referred to as blackness and whiteness)
that seems to correlate with the spectral bandwidth.
Indeed, whiteness was found to increase and blackness Letting A^(1) = A(1)/A(1max), define a parameter 5 =
to decrease with the spectral bandwidth. Consider, for A^(1). Note that as A^(1min) = 0, and A^(1max) = 1, 5 varies
example, a series of Munsell papers of the same central from 0 to 1. Figure 14 presents the same cone fundamen-
wavelength. When the spectral bandwidth is close to zero, tals as in Figure 3 in terms of the reparameterized
the papers look black. When the spectral bandwidth wavelength, 5.
gradually increases, the papers get tinged with some The spectral bandwidth and central wavelength of
chromatic hue, loosing their blackness. For some value of rectangular spectral reflectance functions can be rede-
spectral bandwidth, the hue becomes purely chromatic fined in terms of the new parameter 5. Specifically, given
(containing no blackness at all). Further increase of
spectral bandwidth results in a tinge of whiteness. White-
ness becomes maximal for the spectral bandwidth value
close to (1max j 1min).

Wavelength reparameterization
A large shift in central wavelength between the
Munsell papers of the same Munsell Hue but different type
(Figure 12) can be, at least partly, due to that the ends of
the visible spectrum, 1min and 1max, have been inappropri-
ately defined. Note that the length of the visible spectrum
interval (1max j 1min) is involved in the definition of both
spectral bandwidth and central wavelength for spectral
reflectance functions of type II (Equations 13 and 14).
Any change of 1min and/or 1max will affect the !%1 color
descriptor values. Therefore, if, say, the long wavelength
end, 1max, is unduly in excess, this will result in a central
wavelength shift for spectral reflectance functions of
type II similar to that observed in Figure 12.
It seems natural, then, to define more optimally the Figure 14. Cone fundamentals as in Figure 3. The horizontal axis
visible spectrum interval, leaving outside it the wavelengths represents the meridian length parameter 5.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 12

terms of the meridian length parameter 5. The space

between the coordinate lines in Figure 15 corresponds to
1/30. As one can see, the coordinate web is uniform now.
The color descriptors !, %~, and 1~ lend themselves to be
interpreted as coordinates. Dropping off purity !, one can
use spectral bandwidth and central wavelength, %~ and 1~, as
the Cartesian coordinates to represent a spectral reflec-
tance as a point in a plane. Such a representation will be
referred to as the %1 diagram. It shows which element of
the rectangle chromatic base should be taken to produce
(in the sense of Equation 10) the rectangular metamer of
the spectral reflectance in question. Figure 16 presents in
Figure 15. Regular object-color solid with the coordinate lines in
the %1 diagram the same Munsell papers as Figure 12.
term of reparameterized wavelength 5.
Note that, unlike Figure 12, there is almost no misalign-
ment between the papers of different type.
a rectangular spectral reflectance function x2(1; 11, 12)
(Equation 5), define
The !%1 color descriptors of the Munsell
8 and NCS papers
< k51 j 52 k; for x2 ð1; 11 ; 12 Þ;
%~ ¼ ð18Þ Figure 17 displays all the Munsell papers in the %1
:
1j k51 j 52 k; for 1j x2 ð1; 11 ; 12 Þ; diagram. As one can see, the Munsell papers are rather
unevenly distributed. There are large “lacunas” void of the
where 5i = A^(1i) (i = 1,2). Similarly, define Munsell papers. The Munsell collection is believed to be
uniform. Specifically, adjacent papers in the Munsell tree
8 are supposed to be separated from each other by small,
>
< 0:5ð51 þ 52 Þ; for x2 ð1; 11 ; 12 Þ;
~¼
1 0:5ðð51 þ 52 Þ þ 1Þ; for 1 j x2 ð1; 11 ; 12 Þ if ð51 þ 52 Þ G 1;
presumably, equal perceptual distances. It follows that
>
: even after wavelength reparameterizing the %1 diagram
0:5ðð51 þ 52 Þ j 1Þ; for 1 j x2 ð1; 11 ; 12 Þ if ð51 þ 52 Þ 9 1: remains perceptually non-uniform.
ð19Þ As many symbols in Figure 17 overlap, the histograms in
Figures 18–20 give better understanding how the Munsell
In Figure 15, the same object-color solid as in Figure 8 papers are distributed along each of the !%1 dimensions.
is presented with the coordinate lines equally spaced in Figure 18 shows that purity gravitates to values more than
0.5. Most of the Munsell papers have the spectral

Figure 17. The distribution of 1600 Munsell papers in the %1

diagram. Axes as in Figure 16. The papers of the same Munsell
Figure 16. Color descriptors of the Munsell papers of maximal Hue are depicted by symbols of the same color. The papers of the
chroma from five pages (10B, 10G, 10Y, 10R, and 10PB) after same Munsell Value are represented by symbols of the same
wavelength reparameterizing. Each paper is plotted as a point shape. The red and blue lines represent the main and opposite
with the Cartesian coordinates %~ (ordinate) and 1~ (abscissae). meridians, respectively.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 13

Figure 20. Central wavelength (1~) distribution for 1600 Munsell

Figure 18. Purity (!) distribution for 1600 Munsell papers. papers.

bandwidth value less than 0.25 (Figure 19). Central wave-

transforming the Munsell notations into that of NCS and
length is also distributed rather unevenly (Figure 20).
vice versa, giving us a way to go between the system. The
There are four well-pronounced peaksVaround the 1~
practical importance of such a transformation is generally
values 0, 0.4, 0.75, and 1, that correspond to the ends of acknowledged (e.g., Nayatani, 2004).
the visible spectrum interval and the wavelengths 475 and
575 nm. Interestingly, the value 1~ = 0.4 is the one at
which all the three curves in Figure 14 have approxi-
mately equal ordinates. Object-color space
To find out whether such uneven distribution in the %1
diagram is a particular feature of the Munsell collection, The %1 diagram has two major shortcomings. Firstly, it
the !%1 color descriptors have been also evaluated for the does not represent purity. For purity to be represented, one
1950 NCS papers (Hard & Sivik, 1981) for the same needs one more spatial dimension. Secondly, mapping the
illuminant and cone fundamentals. Figure 21 presents the rectangle chromatic base onto a rectangle area in the
Munsell and NCS papers in the %1 diagram. Both the Cartesian plane is incorrect from the topological point of
Munsell and NCS collections are distributed remarkably view. It is discontinuous at both the perfect reflector
similarly. (x(1) K 1) and the ideal black (x(1) K 0). Indeed, for any
Presenting both Munsell and NCS papers in the same %1 central wavelength, a rectangular spectral reflectance
diagram allows one to see how these two different color function converges to x(1) K 0 when the spectral bandwidth
systems relate to each other. Moreover, this allows for goes to zero. However, a point (%0, 10) in the %1 diagram

Figure 21. NCS (red dots) and Munsell (blue dots) paper location
Figure 19. Spectral bandwidth (%~) distribution for 1600 Munsell in the %1 diagram. Axes as in Figure 16. The red and blue lines
papers. represent the main and opposite meridians, respectively.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 14

will converge to (0, 10) when %0 Y 0. Thus, the points with object-color solid, it will be referred to as the !%1 object-
different 10 will converge to different points in the bottom color solid.
side of the rectangle area in Figure 21. One can avoid such In contrast with the real object-color solid that changes
a break of continuity mapping the rectangle chromatic base its shape as a result of the linear transformation of the
onto a sphere as described below. cone fundamentals, the !%1 object-color solid is invariant
with respect to any non-singular linear transformation of
the cone fundamentals. Furthermore, although it is not
Geographical coordinates for spectral invariant relative to an illuminant change, it remains
reflectance functions almost the same for any illuminant that takes non-zero
value across the visible spectrum (see Optimal metamers
Using the geographical coordinates, one can represent a and optimal color atlas section).
rectangle color atlas as follows. A rectangular metamer
with purity !, spectral bandwidth %~, and central wave-
length 1~ is represented as a point in a 3D space with the Spherical representation of the Munsell
Cartesian coordinate system such that the distance > of the papers
point from the origin equals the purity, i.e.,
As its maximum deviation from the unit ball (in purity
unit) does not exceed 0.01, the !%1 object-color solid for
> ¼ !; ð20Þ
the cone fundamentals based on the photopigment spectra
(Figure 3) is practically indistinguishable from the unit
the geographical latitude " and longitude E being
ball (Figures 22 and 23). Since the Munsell papers have
purity of the magnitude less than 1 (Figure 18), they all lie
" ¼ :%~ j :=2; E ¼ 2:1~: ð21Þ inside of the unit ball. For each Munsell paper, a point on
the !%1 object-color solid boundary (i.e., the chromaticity
The 3D space endowed with the geographical coordinate diagram) was determined, which lies on the same radius.
system such that the coordinates are retrieved from the In other words, all the Munsell papers were projected on
!%1 color descriptors by Equations 20 and 21 will be the unit sphere dropping down the information of their
referred to as the !%1 object-color space. purity (Figures 22 and 23). As in Figure 17, symbol shape
The equator (as well as any circle parallel to it) encodes Munsell Value and symbol color Munsell Hue. In
represents the visible spectrum circle. The meridian spite of the metric difference between the %1 diagram and
semicircles are loci of constant central wavelength. the spherical chromaticity diagram,4 Munsell papers are
Therefore, longitude encodes central wavelength. A circle also quite unevenly distributed over the spherical surface
parallel to the equator is the loci of constant spectral in Figures 22 and 23. As can be seen in Figure 23,
bandwidth. Thus, latitude encodes spectral bandwidth.
The north and south poles represent x(1) K 1 (i.e., the
perfect reflector) and x(1) K 0 (i.e., the ideal black),
respectively. Spectral bandwidth and central wavelength
serve as the internal coordinates on the spherical
surface > = 1; thus, every rectangular spectral reflectance
function can be uniquely located on it. Therefore, the
rectangle chromatic base maps onto the unit sphere > = 1,
which will be called the chromaticity diagram, " and
E in Equation 21 being referred to as the chromaticity
coordinates.
Given a spectral reflectance, its chromaticity coordi-
nates determine the direction (the radius of the sphere) to
which the spectral reflectance belongs, and its purity
determines the distance from the center at which the
spectral reflectance is on this radius.
The rectangle color atlas for the cone fundamentals
satisfying the two-transition assumption will be geometri-
cally represented in the !%1 object-color space as a 3D
unit ball. The rectangle color atlas induced by cone
fundamentals not satisfying the two-transition assumption
is represented in the !%1 object-color space as a 3D Figure 22. The !%1 object-color solid and the spherical represen-
volume deviating from the 3D unit ball. As this volume tation of Munsell papers (view from top). The red and blue lines
can be considered as a smooth transformation of the represent the main and opposite meridians, respectively.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 15

Figure 24. Effect of illuminant on purity (CIE D65 vs. CIE A). Each
Figure 23. The !%1 object-color solid and the spherical represen- of the 1600 points represents the purity, !, of a Munsell paper as
tation of Munsell papers (view from bottom). The red and blue evaluated for D65 (abscissae) and A (ordinate).
lines represent the main and opposite meridians, respectively. and 26 present the spectral bandwidths and central
wavelengths as evaluated for the CIE illuminants D65
and A, respectively.
Munsell papers gravitate toward the south pole of the For the purpose of its quantification, an illuminant-
sphere. Still, there are vast areas on the sphere void of induced color stimulus shift can be decomposed into three
Munsell papers. component shifts. Specifically, let !, %~, and 1~ be the
purity, spectral bandwidth, and central wavelength of the
rectangular metamer for a spectral reflectance x(1) under
Illuminant- and observer-induced one illumination, and !V, %~V, 1~Vunder the other. Then the
color stimulus shifts illuminant-induced purity shift for x(1) can be quantified
by the difference ! j !V, the illuminant-induced spectral
bandwidth shift by %~ j %~V and the illuminant-induced
Illuminant-induced color stimulus shift central wavelength shift by 1~ j 1~V.
When illumination changes an object-color stimulus can
move from one class of metamerism to another. Hence,
the color coordinates of a reflecting object may alter when
the illumination changes because of the shift of its
rectangular metamer from one metameric class to the
other. Such a shift of an object-color stimulus over classes
of metamerism induced by a change in illumination
(referred to as the illuminant-induced color stimulus shift)
results in an alteration of object color. For instance, as
mentioned in the Optimal metamers and optimal color
atlas section, object-color stimulus having complementary
colors may get non-complementary colors because of the
illuminant-induced color stimulus shift.
The illuminant-induced color stimulus shift caused by a
replacement of the CIE illuminant D65 with the CIE
illuminant A has been examined for Munsell and NCS
papers. Specifically, using Equation 12 the !%1 color
descriptors of 1600 Munsell papers were calculated for the
illuminants D65 and A using the cone fundamentals Figure 25. Effect of illuminant on spectral bandwidth (CIE D65 vs.
depicted in Figure 3. Figure 24 shows how purity CIE A). Each of the 1600 points represents the spectral
evaluated for the CIE illuminant A covaries with that bandwidths, %~, of a Munsell paper as evaluated for D65 (the
evaluated for the CIE illuminant D65. Likewise, Figures 25 abscissae) and A (the ordinate).

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 16

Figure 28. Spectral bandwidth shift ª%~ j %~Vª for Munsell papers
induced by the illuminant shift from D65 to A.
Figure 26. Effect of illuminant on central wavelength (CIE D65 vs.
CIE A). Each of the 1600 points represents the central wave-
lengths, 1~, of a Munsell paper as evaluated for D65 (abscissae)
and A (ordinate).
the spectral bandwidth shift magnitude, ª%~ j %~Vª, and
central wavelength shift magnitude, ª1~ j 1~Vª, respectively.
Both the spectral bandwidth and central wavelength
shifts contribute into the shift in chromaticity. It must be
The component color stimulus shifts produced by said, however, that the latter quantity is rather ambiguous.
switching over from the illuminant D65 to A have been Indeed, a shift of the same magnitude ª1~ j 1~Vª along the
derived from the data presented in Figures 24–26. equator might result in a larger chromaticity difference
Specifically, the color descriptors !, %~, and 1~ have been than along circles closer to the poles. In other words, the
calculated for the CIE illuminant D65 and then !V, %~V, 1~V distance between two points on the chromaticity sphere
for the CIE illuminant A. The distribution of the purity separated by the same difference (1~ j 1~V) depends on their
shift magnitude, ª! j !Vª, is given in Figure 27. Purity of latitudes (i.e., %~ and %~V).
the Munsell collection turns out to be rather robust to the A more appropriate index of the chromaticity shift
illuminant change. Specifically, it has been found that would then be a distance on the chromaticity sphere
1023 Munsell papers (64%) change their purity by 0.01 or between two points determined by (%~, 1~) and (%~V, 1~V). As
less, 1404 papers (88%) changing it by not more than 0.025. known, the distance between two points on a sphere is
Similarly, Figures 28 and 29 present the distributions of measured by the length of the shorter arc of the great

Figure 27. Purity shift ª! j !Vª for Munsell papers induced by the Figure 29. Central wavelength shift ª1~ j 1~Vª for Munsell papers
illuminant shift from D65 to A. induced by the illuminant shift from D65 to A.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 17

For large purity differences, Equation 23 is not

appropriate. Indeed, consider, for example, two stimuli
lying on the diameter of the equatorial circle of the !%1
object-color solid (i.e., " = "V= 0 and E j EV= :): one of
full purity (c = 1) and the other with the purity close
to zero, that is, cV; 0. In this case, Equation 23 yields D ;
c/2. Therefore, despite that these object-color stimuli
differ practically only in purity, the Equation 23 gives a
measure substantially different from zero. To avoid such a
“pathology,” Equation 23 can be modified as follows

minfc; cV g   
D¼ arccos cos"cos"V cos E jEV þ sin"sin"VÞ:
:
ð24Þ

When c j cV is small, both Equations 23 and 24 give

approximately the same value. However, when cVY 0, D
Figure 30. Chromaticity shift for Munsell papers induced by the
in Equation 24 approaches zero for any (%~, 1~) and (%~V, 1~V).
illuminant shift from D65 to A.
Using Equation 24, the chromaticity shift induced by
the illuminant shift (from D65 to A) has been evaluated
for 1600 Munsell papers (Figure 30) and 1950 NCS
circle5 through the points. More specifically, consider two papers (Figure 31). The mean chromaticity shift for 1600
object-color stimuli lying on the same sphere in the !%1 Munsell papers is 0.0118; the maximum chromaticity shift
object-color space (i.e., of the same purity) with coor- being 0.1283. For 57.6% of the Munsell papers, the
dinates (c, %~, 1~) and (c, %~V, 1~V). And let ", "V, E, and EVbe the chromaticity shift does not exceed 0.01, for 83% 0.02, and
geographical latitudes and longitudes of the points (%~, 1~) for 92.3% 0.03. A similar result has been observed for
and (%~V, 1~V), that is, " = :%~ j :/2, "V= :%~Vj :/2, E = 2:1~, NCS papers (Figure 31). Specifically, the mean and
and EV= 2:1~V. Then, the distance between these points on maximum chromaticity shifts are found to be 0.0135 and
the spherical surface of radius c is given by 0.1349, respectively. Nearly half the NCS papers (48.5%)
undergo chromaticity shift of not more than 0.01, 79.1%
changing their chromaticity by not more than 0.02, and
d ¼ carccosðcos"cos"V cosðE j EV Þ þ sin"sin"V Þ: 92% not more than 0.03.
ð22Þ It should be borne in mind that the chromaticity
difference (Equation 24) is an arbitrary stimulus measure.
As one can see, the distance d is proportional to purity, c.
This is in line with our intuition that the chromaticity
difference between the object-color stimuli decreases with
their purity. Equation 22 will be used to quantify the
chromaticity difference between the object-color stimuli
with color descriptors (c, %~, 1~) and (c, %~V, 1~V).
This formula can also be used for the object-color
stimuli of different purities, c and cV, provided that the
purity difference ªc j cVª is not large. In this case, the
average purity, (c j cV) / 2, can be put as c in Equation 22.
More specifically, to measure the chromaticity difference
between the object-color stimuli with color descriptors (c,
%~, 1~) and (cV, %~V, 1~V) when c j cVis small the following
formula can be used:

ðc þ cV Þ    
D¼ arccos cos"cos"V cos E j EV þ sin"sin"V :
2:
ð23Þ

Due to normalizing by a factor : in Equation 23, the Figure 31. Chromaticity shift for NCS papers induced by the
maximum chromaticity difference equals 1. illuminant shift from D65 to A.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 18

produced by both a possible change of the sheet color (due

to the illuminant-induced color stimulus shift) and by the
change in the coordinate system. In the !%1 object color,
these two contributions are well separated because the
coordinate system in this space does not depend on
illuminant. That is, the new space is invariant of
illumination change. Moreover, any change of the coor-
dinates of a reflecting object evaluated for two illuminants
indicates the change of the object color induced by the
illuminant alteration. Hence, the new space can be used to
predict the effect of the illumination on object color that is
hard to achieve with the CIE 1931 color space and its
derivatives.

Observer-induced color stimulus shift

As a given spectral reflectance function can be

Figure 32. Chromaticity difference between papers in adjacent metameric to different rectangle metamers for different
pages of the Munsell book of color. observers (i.e., different sets of cone fundamentals), the
color coordinates of the same object can be different for
different observers. A chromaticity shift similar to an
In order to relate it to subjective color difference, the illuminant-induced color stimulus shift can occur due to a
chromaticity differences (Equation 24) between adjacent change of cone fundamentals. In attempt to ascertain how
Munsell Hues have been evaluated. More specifically, robust to alteration of cone fundamentals chromaticity is,
40 Munsell papersVone from each pageVhave been the chromaticity shift induced by replacing the cone
selected so that each selected paper has maximal Munsell fundamentals depicted in Figure 3 with Smith and
Chroma. Chromaticity differences (Equation 24) have been Pokorny’s cone fundamentals (Smith & Pokorny, 1975)
evaluated for 40 consecutive pairs in this selection for 1600 Munsell papers has been evaluated (Figure 33).
(Figure 32). Maximal difference, 0.145, has been found Only 5 of 1600 papers undergo the chromaticity shift
between the papers 10Y8.5/12 and 2.5GY7/12, the exceeding 0.03, the chromaticity shift less than 0.02
minimal, 0.007, between 5RP5/12 and 7.5RP5/14. The taking place for 88% papers. The mean chromaticity
mean chromaticity difference is found to be 0.0366, the shifts induced by the cone fundamentals change is found
median being 0.0250. Therefore, the mean chromaticity to be 0.0057 that is less than the minimal chromaticity
difference between adjacent Munsell Hues is more than difference between Munsell Hues (0.007). Generally, 57%
three times as much as the mean chromaticity shift Munsell papers exhibit a chromaticity shift less than
induced by the illuminant shift (D65 to A) observed for
all the 1600 Munsell papers. Hence, on average, a change
in color for Munsell papers induced by the shift from the
illuminants D65 to A is, by and large, at the limit of our
ability to discern chromatic differences.
The main motivation for developing a new color space
has been that the CIE 1931 color space, and its derivatives
can be used to represent the color of objects only if the
illumination of these objects is the same because when the
illumination changes the vector representation (i.e.,
tristimulus values) changes as well. As one can see, the
color coordinates of a reflecting object in the !%1 object-
color space vary with illumination as well (because of
illuminant-induced color stimulus shift). Yet, there is an
important difference between these two cases. In the case
of the CIE 1931 color stimulus representation, the
coordinate system itself changes with illumination. For
example, a sheet of paper under (i) the direct sun light and
(ii) the light from blue sky will reflect different lights. Figure 33. Chromaticity shift produced by replacing the cone
These lights will be represented by two different points in fundamentals depicted in Figure 3 with Smith and Pokorny’s
the CIE 1931 color space. This difference is, generally, (1975) cone fundamentals.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 19

0.007. Therefore, at least half the papers undergo a transition wavelengths were complementary to each other;
chromaticity shift (due to the cone fundamentals change), or one transition wavelength coincided with the visible
which is most unlikely to be noticed by a human observer. spectrum interval end (i.e., 1min or 1max), and the other lay
between the wavelengths complementary to 1min and
1max. The semichromes prove to map into a closed curve
on the object-color solid boundary (Koenderink & Doorn,
Discussion 2003). As a result, Ostwald’s color atlas turned out to be
not complete; that is, it did not encompass all the object
colors. Moreover, as the complementarity of wavelengths
A spatial representation of object color implies a depends on illuminant, Ostwald’s definition of semi-
sample of spectral reflectance functions, i.e., color atlas chromes depends on illuminant. Therefore, Ostwald’s
(Koenderink, 1987; Koenderink & Doorn, 2003). It is the color atlas is not invariant of illumination.
color atlas that undergoes a spatial order whereby all the The optimal color atlas proposed here is complete and
colors it renders also get spatially arranged. Ideally, a invariant with respect to all the lights with positive
color atlas (thus, the color order system based on it) is to spectral power distribution. It provides a frame of
specify a spectral reflectance in terms of some perceptual reference, independent of illuminant, in terms of which
color dimensions. In fact, the color order systems each class of object-color stimuli metameric under any
available at the present are rather far from this ideal. positive illuminant can be represented. More specifically,
Indeed, the spatial order is usually imposed on the color as shown above, the optimal color atlas makes a three-
atlas either in terms of some perceptual (e.g., Munsell dimensional manifold that can be geometrically repre-
atlas, NCS) or physical dimensions (e.g., Koenderink & sented as a 3D unit ball. A class of stimuli metameric
Doorn, 2003; Ostwald, 1931). Being based on intuitively under a specific illuminant is specified by projecting it
clear perceptual dimensions, the perceptual color order along the radius from the center to the corresponding
systems usually do not provide an easy way to characterize optimal object-color stimuli. Such a frame reference
an arbitrary spectral reflectance in terms of these dimen- remains the same for any light with positive spectral
sions.6 For instance, 200 years ago Runge, with his color power distribution because, first, the set of optimal object-
sphere, anticipated the geometrical form of the object- color stimuli does not depend on illumination (of this
color manifold7 (Kuehni, 2003, pp. 59–62). However, one kind), and second, the spectral reflectance x0.5(1) K 0.5
cannot locate a particular spectral reflectance in Runge’s always belongs to the class of metamerism, which is the
sphere. Nor can one compute Munsell Hue, Chroma, and center of the manifold of all the classes of metamerism for
Value for an arbitrary spectral reflectance function.8 On any illuminant. Note that this way of encoding is similar
the other hand, the dimensions of the physical color order to that which the visual system uses to encode color.
systems do not lend themselves readily to perceptual Achromatic and chromatic hues encode the optimal
interpretation. For example, representing a spectral reflec- object-color stimulus, and purity encodes the distance
tance as a weighted combination of three basis spectral from the center. Therefore, the way of encoding object
reflectance functions, the linear models specify each color underlying the proposed object-color space is a
arbitrary spectral reflectance by the three numbers (for a rough sketch of the end product of the human color vision.
review, see Brainard, 1995; Hurlbert, 1998; Maloney, Furthermore, reparameterizing the optimal color atlas
2003). However, a phenomenological interpretation of by using the rectangle color atlas allows to introduce the
these numbers is hardly possible. spherical coordinate system with clear perceptual mean-
An ideal solution would be a physical color order system ing. Indeed, the central wavelength proves to characterize
whose dimensions allow straightforward phenomenolog- the hue of a rectangular metamer (thus, any spectral
ical interpretation. Ostwald’s color system (Ostwald, reflectance metameric to it) as accurately as the dominant
1931) can be considered as a significant step toward this wavelength does the hue of a monochromatic light.
ideal. This system is based on a set of rectangular spectral An important feature of the rectangle color atlas is that
reflectance functions that differ from the rectangle color the !%1 color descriptors of the optimal rectangular
atlas put forward here in the following. Firstly, Ostwald’s spectral reflectance functions remain practically constant
rectangular spectral reflectance function takes two values, with respect to illumination. Particularly, the chromaticity
a and b, which are not necessarily symmetric with respect coordinates and purity of the perfect reflector (x(1) K 1)
to value 0.5 (that is, a + b can differ from 1). Secondly, and the ideal black (x(1) K 0) are the same for all the
Ostwald’s color atlas does not include the whole set of illuminants (with positive spectral power distribution).
rectangular spectral reflectance functions. More specifi- This is not the case when reparameterization of the
cally, Ostwald employed a set of rectangular spectral optimal color atlas is based on a linear representation
reflectance functions of special form which he called unless the basis spectral reflectance functions themselves
semichromes (Vollfarben) (Koenderink & Doorn, 2003). include the perfect reflector or ideal black. Generally, the
The set of semichromes included only those rectangular linear coordinates of the optimal rectangular spectral
spectral reflectance functions for which either their reflectance functions will change with illumination.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 20

Furthermore, the basis spectral reflectance functions move Illuminant: An illuminant is spectral power distribution.
from one class of metamerism to another when illumina- Remark: Object-color stimulus times illuminant
tion alters. Therefore, the linear representation of a amounts to color stimulus as defined by Wyszecki and
metameric class is performed in terms of the basis that Stiles (1982, p. 723).
is subject to uncontrollable change when illumination Sensor: A sensor 8 is a linear device the response, 8(w),
alters. In other words, the linear models do not provide a of which to a color stimulus w(1) is given by
frame reference robust to illumination change. From this 1Zmax
point of view, using the linear models for representing
spectral reflectance provides no advantage to the common 8ðwÞ ¼ sð1Þwð1Þd1; ðA1Þ
practice of using the CIE 1931 color space to represent 1min
object color.
Likewise, a color atlas based on Gaussian spectral where s(1) is the spectral sensitivity of the sensor and
reflectance functions (MacLeod & Golz, 2003; Nikolaev, [1min, 1max] is the visible spectrum interval.
1985) does not provide a frame reference robust to Color signal: The color signal produced by sensors 81,
illumination change either. Neither does any physical 82, and 83 in response to a color stimulus w(1) is a triplet
finite sample of spectral reflectance functions (e.g., (81(w), 82(w), 83(w)).
Munsell atlas, NCS). Indeed, being subject to color Color signal space: Given sensors 81, 82, and 83, a
stimulus shift, the Munsell papers may change their class color signal space is a 3D affine space in which any color
of metamerism with illumination. Therefore, if some stimulus w(1) is represented as a point with the coor-
spectral reflectance remains metameric to the same dinates (81(w), 82(w), 83(w)).
Munsell paper under two different illuminants, this will Metameric object-color stimuli: Two physically differ-
not necessarily guarantee that its color remains the same. ent object-color stimuli producing the same color signals
It might simply mean that this spectral reflectance changes under some illuminant are called metameric under this
its color.9 The same way as the Munsell paper in question. illuminant.
Object-color: An object-color is a class of metameric
object-color stimuli.
Conclusion Object-color solid: The object-color solid is a set of
points in the color signal space produced by all possible
object-color stimuli for a fixed illuminant.
The new color space is suitable for computational Optimal object-color stimuli: An optimal object-color
representation of the object colors encoded by not only stimulus is a spectral reflectance the color signal of which
humans but also artificial devices (digital cameras, LCD lies on the boundary surface of the object-color solid.
and CRT displays, color printers, etc.). Specifically, the Remark: If x(1) is optimal object-color stimulus, then
proposed new color space can be used as a device- 1 j x(1) is optimal object-color stimulus as well.
independent color space in color management systems for Complementary optimal object-color stimuli: Given an
input (digital cameras, scanners) and output (displays, optimal object-color stimulus x(1), the optimal object-
printers) device calibration and characterization. There are color stimulus 1 j x(1) is called complementary to x(1).
a number of advantages of the proposed color space to the Step function: A step (spectral reflectance) function is
CIE 1931 colorimetric space (as well as the sRGB color a piecewise constant function taking only two values: 0
space and many others based on it) currently used as a or 1.
device-independent color space (e.g., Bala, 2003). Firstly, Remark: Each optimal object-color stimulus is a step
the proposed color space is a space for reflective (rather spectral reflectance function.
than self-luminous) objects. Secondly, the dimensions of Transition wavelengths: Transition wavelengths of a
the proposed color space have clear perceptual correlates. step spectral reflectance function are those where tran-
Thirdly, the proposed color space is independent of sition between 1 and 0 occurs.
illumination. Furthermore, the new color space is robust Rectangular function: A rectangular (spectral reflec-
to not only illuminant but observer as well. tance) function is a step function taking 1 on the interval
[11, 12] (1min e 11 e 12 e 1max) and 0 outside it.
Complete color atlas: A complete color atlas is a set of
Appendix A (not metameric) object-color stimuli such that each object-
color stimulus is metameric to one of its element.
Illuminant invariant color atlas: A complete color atlas
Basic terms and definitions is said to be invariant with respect to the family of
illuminants if and only if (i) each class of metamerism
Object-color stimulus: An object-color stimulus is contains exactly one element of the atlas under any
spectral reflectance. illuminant in the family and (ii) any two elements of the

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 21

atlas do not become metameric under any illuminant in Chromaticity coordinates: The chromaticity coordinates
the family. of an object-color stimulus x(1) are the quantities " and E,
Optimal color atlas: The optimal color atlas is a set of which are related to the spectral bandwidth and the central
spectral reflectance functions expressed as wavelength of x(1) as " = :%~ j :/2 and E = 2:1~, where %~
and 1~ are defined by Equations 18 and 19.
ð1j !Þx0:5 ð1Þ þ !xopt ð1Þ; ðA2Þ Object-color space: Object-color space is a 3D space with
the geographical coordinate system such that each object-
where x0.5(1) is the spectral reflectance function taking 0.5 color stimulus x(1) is represented as a point at a distance
at every wavelength 1 within the visible spectrum interval from the origin equal to its purity, the latitude and longitude
[1min, 1max] and xopt(1) runs over the whole set of optimal being equal to the chromaticity coordinates of x(1).
object-color stimuli. Chromaticity difference: Chromaticity difference
Remark: The optimal color atlas is complete and between object-color stimuli with the chromaticity coor-
invariant with respect to illuminants with positive spectral dinates ", "V, E, and EVand purities c and cV is given by
power distribution. Equation 24.
Rectangular metamer: The rectangular metamer of an
object-color stimulus x(1) is a piecewise constant spectral
reflectance function, metameric to x(1), which is given by Illuminant invariance of optimal color atlas
(proof)
ð1j !Þx0:5 ð1Þ þ !x2 ð1; 11 ; 12 Þ; ðA3Þ
Let x1(1) and x2(1) be two different spectral reflectance
where x0.5(1) is defined as above (Equation A2); x2(1; 11, functions given by Equation 10, that is, x1(1) = (1 j !V)
12) is the rectangular spectral reflectance function with the x0.5(1) + !Vxopt
V (1) and x2(1) = (1 j !W) x0.5(1) +
transition wavelengths 11 e 12; ! is a real number. !WxoptW (1), where xopt
V (1) and xopt
W (1) are optimal spectral
Proper and improper rectangular metamers: The reflectance functions (with respect to some set of cone
rectangular metamer (Equation A3) is called proper fundamentals) such that 0 G !V, !W G 1, and either xopt
V (1) m
(respectively, improper) if ª!ª in Equation A3 is such W (1) or !Vm !W, or both the inequalities hold true. As
xopt
that ª!ª e 1 (respectively, ª!ª 9 1). illuminants with positive spectral power distribution do
Rectangle color atlas: The rectangle color atlas is the set not change the optimal spectral reflectance function set,
of the rectangular metamers for all the object-color stimuli. V (1) and xopt
for any such illuminant, xopt W (1) will remain
Central wavelength: The central wavelength of a optimal. Given an arbitrary illuminant with positive
rectangular spectral reflectance function with the transi- 
spectral power distribution, let us denote x opt V and
tion wavelengths 11 and 12 is given by (11 + 12) / 2. The 
W the points in the object-color solid boundary into
x opt
central wavelength of an element of the rectangle color which xopt W (1) map. Let 
V (1) and xopt x 0.5, 
x 1, and 
x 2 be
atlas (1 j !) x0.5(1) + !x2(1; 11, 12) is the central the points in the object-color solid into which x0.5(1),
wavelength of the rectangular spectral reflectance function  
x1(1), and x2(1) map, respectively. Assume that x 1 = x 2,
x2(1; 11, 12). The central wavelength of an object-color    
that is, (1 j !V) x 0.5 + !Vx opt
V = (1 j !W)x 0.5 + !Wx opt W.
stimulus x(1) is the central wavelength of its rectangular Then, it implies that
metamer.
Spectral band: The spectral bandwidth of a rectangular   
spectral reflectance function with the transition wave- !V x Vopt j !W x Wopt ¼ ð!V j !W Þx 0:5 : ðA4Þ
lengths 11 and 12 is given by ª11 j 12ª. The spectral
bandwidth of an element of the rectangle color atlas (1 j !) If !V= !W, then it follows from Equation A4 that  V =
x opt
x0.5(1) + !x2(1; 11, 12) is the spectral bandwidth of the 
W . As there is no metamerism in the object-color solid
x opt
rectangular spectral reflectance function x2(1; 11, 12).  
V = x opt
boundary, x opt W implies xoptV (1) = xopt
W (1) that is
The spectral bandwidth of an object-color stimulus x(1) is impossible because x1(1) and x2(1) are different. There-
the spectral bandwidth of its rectangular metamer. fore, !Vm !W.
Chromatic amplitude: The parameter ! of an element of Assume for certainty that !V9 !W and rewrite Equation A4
the rectangle color atlas (1 j !) x0.5(1) + ! x2(1; 11, 12) is as
its chromatic amplitude. The chromatic amplitude of an
 
object-color stimulus x(1) is the chromatic amplitude of  !W  !W 
its rectangular metamer. V ¼ V x opt
x opt W þ 1j V x 0:5 : ðA5Þ
! !
Purity: The purity of an object-color stimulus is the
absolute value of its chromatic amplitude. W 
Complementary object-color stimuli: Two object-color As 0 G !!V G 1, it follows from Equation A5 that x opt
V is an
stimuli are complementary if their rectangle metamers interior point of the object-color solid that is impossible
differ only by sign of the chromatic amplitude !. V (1) is optimal.
because xopt

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 22
  9
Therefore, x 1 and x 2 are different points in the object- It is a shift from one class of metamerism to another
color solid. So we conclude that different elements x1(1) rather than a phenomenological alteration that is meant
and x2(1) of the optimal color atlas cannot become here by changing color.
metameric under illuminants with positive spectral power
distribution.
References
Bala, R. (2003). Device characterization. In G. Sharma
Acknowledgments (Ed.), Digital color imaging handbook (pp. 269–379).
Washington, DC: CRC Press.
This work was supported by a research grant EP/ Brainard, D. H. (1995). Colorimetry. In M. Bass (Ed.),
C010353/1 from EPSRC. It has greatly benefited from OSA handbook of optics: Vol. 1. Fundamentals,
fruitful discussions on various aspects of color theory with techniques, and design. New York: McGraw-Hill,
M. Brill, A. Chernavski, V. Levin, L. Maloney, and Inc.
V. Maximov. I am grateful to V. Govardovskii for Brill, M. H., & Robertson, A. R. (2007). Open problems
providing me with the photopigment templates. Thanks on the validity of Grassmann’s laws. In J. Schanda
are due to T. Lu and R. Tokunaga for assistance in (Ed.), Colorimetry. New York: John Wiley and Sons,
preparation of the manuscript. I also thank D. Brainard, Inc.
M. Brill, D. Foster, B. Funt, J. Koenderink, and R. Kuehni Brindley, G. S. (1955). The colour of light of very long
for their helpful comments on earlier drafts of the paper. wavelength. The Journal of Physiology, 130, 35–44.
Commercial relationships: none. [PubMed] [Article]
Corresponding authors: Alexander D. Logvinenko. DeMarco, P., Pokorny, J., & Smith, V. C. (1992). Full-
Email: [email protected]. spectrum cone sensitivity functions for X-chromosome
Address: Department of Vision Sciences, Glasgow Cale- linked anomalous trichromats. Journal of the Optical
donian University, City Campus, Cowcaddens Road, Society of America A, Optics and Image Science, 9,
Glasgow, G4 0BA, UK. 1465–1476. [PubMed]
Govardovskii, V. I., Fyhrquist, N., Reuter, T., Kuzmin,
D. G., & Donner, K. (2000). In search of the visual
Footnotes pigment template. Visual Neuroscience, 17, 509–528.
[PubMed]
1
Although this intuitively clear statement has been Hard, A., & Sivik, L. (1981). NCSVNatural color system:
made by several authors before (e.g., Ostwald, 1923; A Swedish standard for color notation. Color
Wyszecki & Stiles, 1982; although see Maximov, 1984, Research and Application, 6, 129–138.
p. 73), the formal proof has not been known until recently Hurlbert, A. C. (1998). Computational models of color
(Logvinenko & Levin, 2009). constancy. In V. Walsh & J. J. Kulikowski (Eds.),
2
See Koenderink (1987) for a notion of color atlas. Perceptual constancy: Why things look as they do
3
Note that the symmetrical location of complementary (pp. 283–322). Cambridge: Cambridge University
optimal metamers in the object-color solid follows from Press.
the definition of complementarity rather than underlies it.
Izmailov, C. A. (1995). Spherical model of discrimination
4
The distance between Munsell papers in the plane %1
of self-luminous and surface colors. In R. D. Luce,
diagram is different from that in the spherical chromaticity
M. D. D’Zmura, & A. K. Romney (Eds.), Geometric
diagram.
5 representations of perceptual phenomena (pp. 153–168).
A great circle is that whose center is coincident with
Mahwah, NJ: Lawrence Erlbaum Associates.
the center of the sphere.
6
At best, it can be done only on the ground of visual Judd, D. B., & Wyszecki, G. (1975). Color in business,
(subjective) inspection provided the spectral reflectance is science, and industry (3rd ed.). New York: John
implemented as a real surface. Wiley and Sons, Inc.
7
Needless to say that at the time, no experimental data
Koenderink, J. J. (1987). Color atlas theory. Journal of the
corroborated Runge’s guess. Yet, more recently, it was
Optical Society of America A, 4, 1314–1321.
supported by multidimensional analysis of Munsell papers
(Izmailov, 1995). Koenderink, J. J., & Doorn, A. J. van. (2003). Perspectives
8
Strictly speaking, it is only Albert Munsell himself on color space. In R. Mausfeld & D. Heyer (Eds.),
who could do this provided that the reflectance function is Colour vision: From light to object (pp. 1–56).
implemented as a surface available for visual inspection. Oxford: Oxford University Press.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018

 Journal of Vision (2009) 9(11):5, 1–23 Logvinenko 23

Krantz, D. H. (1975). Color measurement and color Ostwald, W. (1931). Colour science. London: Winsor and
theory: I. Representation theorem for Grassmann Newton.
structures. Journal of Mathematical Psychology, 12, Schanda, J. (Ed.) (2007). Colorimetry. New York: John
283–303. Wiley and Sons, Inc.
Kuehni, R. G. (2003). Color space and its divisions. Schrodinger, E. (1920). Theorie der Pigmente von grosster
Hoboken, NJ: John Wiley and Sons, Inc. Leuchtkraft. Annalen der Physik, 62, 603–622.
Lamb, T. (1999). Photopigments and the biophysics of Schrodinger, E. (1970). Outline of a theory of color
transduction in cone photoreceptors. In K. R. measurement for daylight vision. In D. L. MacAdam
Gegenfurtner & L. T. Sharpe (Eds.), Color vision: (Ed.), Sources of color sciences (pp. 134–182).
From genes to perception (pp. 89–101). Cambridge: Cambridge, MA: The MIT Press. (Reprinted from
Cambridge University Press. Annalen der Physik, vol. 63, pp. 481–520, 1920).
Logvinenko, A. D. (2006). Does colour matching follow Smith, V. C., & Pokorny, J. (1975). Spectral sensitivity of
Grassmann’s laws? In Symposium on 75 years of the the foveal cone photopigments between 400 and
CIE Standard Colorimetric Observer (pp. 133–138). 500 nm. Vision Research, 15, 161–171. [PubMed]
Vienna, Austria: CIE Publication x030:2006, Com- Smith, V. C., & Pokorny, J. (2003). Color matching and
mission Internationale de l’Eclairage. color discrimination. In S. K. Shevell (Ed.), The
Logvinenko, A. D., & Levin, V. L. (2009). Foundations of science of color (2nd ed., pp. 103–148). Amsterdam:
colour science. Unpublished manuscript. Elsevier.
Luther, R. (1927). Aus dem Gebiet der Farbreiz-Metric. Smith, V. C., Pokorny, J., & Zaidi, Q. (1983). How do sets
Zeitschrift fur Technische Physik, 8, 540–558. of color-matching functions differ? In J. Mollon &
L. Sharpe (Eds.), Colour vision: Physiology and
MacAdam, D. L. (1935). The theory of the maximum
psychophysics (pp. 93–105). London: Academic Press
visual efficiency of colored materials. Journal of the
Ltd.
Optical Society of America, 25, 249–252.
MacLeod, D. I. A., & Golz, J. (2003). A computational Stockman, A., & Sharpe, L. T. (2000). The spectral
analysis of colour constancy. In R. Mausfeld & sensitivities of the middle- and long-wavelength-
D. Heyer (Eds.), Colour vision: From light to object sensitive cones derived from measurements in observers
(pp. 205–242). Oxford: Oxford University Press. of known genotype. Vision Research, 40, 1711–1737.
[PubMed]
Maloney, L. T. (2003). Surface color perception and environ-
mental constraints. In R. Mausfeld & D. Heyer (Eds.), Stockman, A., & Sharpe, L. T. (2007). Spectral sensitiv-
Colour vision: From light to object (pp. 279–300). ities. In T. D. Albright & R. H. Masland (Eds.),
Oxford: Oxford University Press. The senses: A comprehensive reference: Vol. 2. Vision
(p. 87–100). San Diego, CA: Academic Press.
Maximov, V. V. (1984). Transformatsii tsveta pri izme-
nenii osvescheniya [Transformations of colour under Suppes, P., Krantz, D. H., Luce, R. D., & Tversky, A.
illumination changes]. Moscow: Nauka. (Eds.) (1989). Foundations of measurement: Vol. 2.
Geometrical, threshold, and probabilistic representa-
Nayatani, Y. (2004). Transformation from Munsell Space tions. San Diego, CA: Academic Press.
to Swedish NCS through Nayatani Theoretical Space.
Color Research and Application, 29, 151–157. West, G., & Brill, M. H. (1983). Conditions under which
Schrodinger object colors are optimal. Journal of the
Nikolaev, P. P. (1985). A colour constancy model with Optical Society of America, 73, 1223–1225.
continuous spectral functions. Biophysics, 30, 112–117.
Wyszecki, G., & Stiles, W. S. (1982). Color science:
Nyberg, N. D. (1928). Zum Aufbau des Farbkorpers im Concepts and methods, quantitative data and for-
Raume aller Lichtempfindungen. Zeitschrift fur mulae (2nd ed.). New York: John Wiley and Sons.
Physik, 52, 406–419.
Ostwald, W. (1923). Die Farbenlehre. II. Buch: Physika-
lische Farbenlehre (2nd ed.). Leipzig: Unesma.

Downloaded From: http://jov.arvojournals.org/pdfaccess.ashx?url=/data/journals/jov/933556/ on 04/08/2018