green light produces a reddish-purple hue (red + blue, or magenta), while subtracting the red results
in cyan, a combination of blue and green.
Thus, the cyan-magenta-yellow set comprises the subtractive primaries—typically used in printing.
However, since real-world subtractive primaries don't really absorb all the light at any given
wavelength, combining the three doesn't generally give a very good black. So, black ink is commonly
added to the C-M-Y set, resulting in "CMYK" printing. ("K" is used for black, since "B" was already
taken for the blue in the RGB set).
Over the years, many different ways of quantifying color have been developed. Generally, a model
that represents how color is represented is known as a color space. Different devices and applications
can (and do) use different color spaces.
In 1931, the International Commission on Illumination (or CIE as it's known by its French initials)
published a specification that described standardized curves based on the sensitivity of normal human
vision vs. wavelength for the three types of receptors in the eye. These are known as the CIE 1931
color-matching functions, and they lead directly to a space defined by three standardized primaries
known as X, Y, and Z. A simplified "two-dimensional" model was derived from the XYZ system, and is
referred to as the CIE xy color space, or more correctly the xy chromaticity coordinates or
The CIE 1931 system takes into account humans' response to different wavelengths of light, and has
become a standard in the electronics industry for specifying the ability of an electronic device to
reproduce color. However, the CIE 1931 chart is not a perceptually uniform color space. In other
words, the same change in distance on the xy chart in two different regions generally does not mean
the same change in perceived color as a human eye would see it.
In 1976, a system similar to CIE 1931 was developed that partially corrects for the non-uniformity of
CIE 1931—it is called the L*u*v* color space, and has an associated u'v' color chart (Figure 2; the
"color gamut" is discussed in the next section). The challenge (as we will explore) is mapping color
spaces over a range of input devices, applications, and output devices while maintaining (or
supporting) color range accuracy and predictability.