i
i
i
i
i
i
i
i
21.3. Chromatic Adaptation 551
of the adapting white, carried out in cone response space:
⎡
⎣
L
a
M
a
S
a
⎤
⎦
=
⎡
⎢
⎢
⎢
⎢
⎣
1
L
w
00
0
1
M
w
0
00
1
S
w
⎤
⎥
⎥
⎥
⎥
⎦
⎡
⎣
L
M
S
⎤
⎦
.
In many cases, we are interested in what stimulus should be generated under
one illumination to match a given color under a different illumination. For ex-
ample, if we have a colored patch illuminated by daylight, we may ask ourselves
what tristimulus values should be generated to create a matching color patch that
will be illuminated by incandescent light.
We are thus interested in computing corresponding colors, which can be
achieved by cascading two chromatic adaptation calculations. In essence, the
above von Kries transform divides out the adapting illuminant—in our example,
the daylight illumination. If we subsequently multiply in the incandescent il-
luminant, we have computed a corresponding color. If the two illuminants are
given by (L
w,1
,M
w,1
,S
w,1
) and (L
w,2
,M
w,2
,S
w,2
), the corresponding color
(L
c
,M
c
,S
c
) is given by
⎡
⎣
L
c
M
c
S
c
⎤
⎦
=
⎡
⎣
L
w,2
00
0 M
w,2
0
00S
w,2
⎤
⎦
⎡
⎢
⎢
⎢
⎢
⎢
⎣
1
L
w,1
00
0
1
M
w,1
0
00
1
S
w,1
⎤
⎥
⎥
⎥
⎥
⎥
⎦
⎡
⎣
L
M
S
⎤
⎦
.
There are several more complicated and, therefore, more accurate chromatic
adaptation transform in existence (Reinhard et al., 2008). However, the simple
von Kries model remains remarkably effective in modeling chromatic adaptation
and can thus be used to achieve white balancing in digital images.
The importance of chromatic adaptation in the context of rendering, is that we
have moved one step closer to taking into account the viewing environment of the
observer, without having to correct for it by adjusting the scene and rerendering
our imagery. Instead, we can model and render our scenes, and then, as an image
post-process, correct for the illumination of the viewing environment. To ensure
that white balancing does not introduceartifacts, however, it is important to ensure
that the image is rendered to a floating-point format. If rendered to traditional 8-
bit image formats, the chromatic adaptation transform may amplify quantization
errors.
i
i
i
i
i
i
i
i
552 21. Color
21.4 Color Appearance
While colorimetry allows us to accurately specify and communicate color in a
device-independent manner, and chromatic adaptation allows us to predict color
matches across changes in illumination, these tools are still insufficient to describe
what colors actually look like.
To predict the actual perception of an object, we need to know more informa-
tion about the environment and take that information into account. The human
visual system is constantly adapting to its environment, which means that the per-
ception of color will be strongly influenced by such changes. Color appearance
models take into account measurements of the stimulus itself, as well as the view-
ing environment. This means that the resulting description of color is independent
of viewing condition.
The importance of color appearance modeling can be seen in the following
example. Consider an image b eing displayed on an LCD screen. When making
a print of the same image and viewing it in a different context, more often than
not the image will look markedly different. Color appearance models can be
used to predict the changes required to generate an accurate cross-media color
reproduction (Fairchild, 2005).
Although color appearance modeling offers important tools for color repro-
duction, actual implementations tend to be relatively complicated and cumber-
some in practical use. It can be anticipated that this situ ation may change over
time. However, until then, we leave their description to more specialized text
books (Fairchild, 2005).
Notes
Of all the books on color theory, Reinhard et al.’s work (Reinhard et al., 2008) is
most directly geared towards engineering disciplines, including computer graph-
ics, computer vision, and image processing. Other general introductions to color
theory are given by Berns (Berns, 2000) and Stone (Stone, 2003). Wyszecki and
Stiles have produced a comprehensive volume of data and formulae, forming an
indispensable reference work (Wyszecki & Stiles, 2000). For color reproduction,
we recommend Hunt’s book (Hunt, 2004). Color appearance models are compre-
hensively described in Fairchild’s book (Fairchild, 2005). For color issues related
to video and HDTV Poynton’s book is essential. (Poynton, 2003).
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.