Graphics Reference
In-Depth Information
are identical, but the contrasts in the left column appear less than the contrasts in
the right column.
The signal representative of luminance (the 0.42 power of luminance) seems
as if it should be a part of a video signal; in fact, a video signal starts out as three
values, r , g , and b representing the amounts of red, green, and blue light in a way
that's linear in the intensity (if you double the intensity, then each of r , g , and b will
double). The luma is then a weighted sum of r 0.42 , g 0.42 , and b 0.42 . The difference
between these values and the values determined by computing luminance directly
and raising that to the 0.42 power is generally insignificant (another application
of the Noncommutativity principle) and luma is used as the Y component of the
Y IQ color model described below. The prime on Y indicates that this coordinate
does not vary linearly as a function of the light intensity. Ordinary video cameras
compute R, G, and B values that, for a given aperture and white balance, are
proportional to incoming intensities in the appropriate wavelength ranges, and
raise these to the 0.45 power; the values they produce should therefore be called
R , G , and B , following the naming convention. To recover the original R , G ,
and B values, these must be raised to the 2.2 power. And to transform to other
color spaces, we typically must first recover R, G, and B, and then perform the
conversion, since most color transformations are described in terms of things like
R, G, and B that vary linearly with energy.
The exponent 2.2 that is used to convert video R G B values back to RGB is
often called gamma, and the process of raising values to some power around 2.2
is known as gamma correction. The number 2.2 is by no means universal; other
gamma values have been used in various image formats over the years, and many
image display programs allow the user to “adjust gamma” to modify the exponent
used in the display process.
28.13 Describing Color
In computer graphics, we often need to describe color mathematically. Because
the physical interaction of light and surfaces occurs in ways determined by their
spectra rather than their colors, we don't use the L u v description of light when
we want to model this physical interaction. And because the values we compute
while rendering are typically spectral radiance values (possibly for some fairly
broadband spectra, i.e., the radiance for the bottom, middle, and top thirds of
the visible spectrum), which then must be converted to values that govern three
display brightnesses, it's best to separate the physical models used in rendering
from our description of colors that appear on our displays or printers.
So we'll now present several color models used to describe the colors that
our devices can produce. Typically these color models are bounded, in the sense
that they can only describe colors up to a certain intensity (or generally only
a subset of the colors up to some intensity value). This matches the physical
characteristics of many devices: An LCD monitor cannot produce more than a
certain brightness; the light reflected from a printed page cannot exceed the light
arriving at the page, etc.
The choice of a color model may be motivated by simplicity (as in the RGB
model), ease of use (the HSV and HLS models), or particular engineering con-
cerns (like the Y IQ model used for the broadcast of color television signals or
the CMY model for printing). And with the widespread interchange of imagery
among different devices, there are color models whose design is based on lossless
 
 
Search WWH ::




Custom Search