Graphics Reference
In-Depth Information
There's also the widely used Munsell color-order system [Fi76], in which a
wide range of colors are organized in a three-dimensional system of hue, value
(i.e., lightness), and chroma (i.e., saturation or “color purity”), and in which adja-
cent colors have equal perceived “distance” in color space (as judged by a wide
collection of observers).
We have observed that monospectral lights provoke a wide range of sensor
responses, plotted on the horseshoe-shaped curve. We've also seen that choosing
three monospectral lights in the red, green, and blue areas of the spectrum (we'll
call these primaries for the remainder of this section) allows us to produce, by
combining them, many familiar color sensations, but not by any means
all
.Aswe
said earlier, when we consider a color like orange, we find that no combination of
our red, green, and blue primary lights gets us light that we perceive as orange. We
can, through subterfuge, still express the orange light as a sum of the red, green,
and blue primaries, however. What we
really
want is to say that “orange looks like
about a half-and-half mix of red and green, and then move
away
from blue.” In
equations, we'd write something like
orange
=
.45
red
+
.45
green
−
0.1
blue
.
(28.11)
Of course, we can't take away blue light that isn't there, but we can add blue light
to the orange. If we find that
1.0
orange
+
0.1
blue
=
.45
red
+
.45
green
,
(28.12)
in the sense that the color mixes on the left and right produce the same sensor
responses, then we'll express that numerically with Equation 28.11. In this way,
we can find what mixes of our primaries are needed to match
any
monospectral
light
L
, and plot the result as a function of the wavelength of
L
; the result has
the shape shown in Figure 28.16. These three “color matching functions,”
r
,
g
,
and
b
, tell us how much of our red, green, and blue primaries need to be mixed
to generate each monospectral light. For example, to make light that looked like
500 nm monospectral light, we'd have to combine about equal parts of blue and
green, and subtract quite a lot of red (i.e., we'd use
r
(
500
)
,
g
(
500
)
, and
b
(
500
)
as
the mixing coefficients). To make something resembling 650 nm light, we'd use
lots of red, a little green, and no blue.
What about a 50-50 mix of 500 nm and 650 nm light? We'd use a 50-50 mix of
the two color matches above. Because such a mix has all coefficients positive, it's
actually possible to make it with our red, green, and blue standard monospectral
lights. In general, if we have a light with a spectral power distribution
P
, we can
find the “mixing coefficients” by applying the idea above to each wavelength, that
is, we compute
c
r
=
700
400
P
(
λ
)
r
(
λ
)
d
λ
,
(28.13)
c
g
=
700
400
P
(
λ
)
g
(
λ
)
d
λ
, and
(28.14)
c
b
=
700
400
)
b
(
P
(
λ
λ
)
d
λ
,
(28.15)
