TABLE OF CONTENTS It has been shown that color can be described by many different vectors. The spectral space of N-vectors is the highest dimensional space and most complete. That is, if one knows the spectrum of the radiant source or the reflectivity of an object, one can compute its color coordinates for all practical uses. However, in most cases, the complete spectrum is not needed and only the tristimulus values for a radiant source, or a limited set of tristimulus values under a few illuminants for a reflective object are required. The restriction to tristimulus values greatly reduces the dimensionality of the problem. It also changes the color space used to describe the objects.
There are practical reasons for choosing to describe a color in a particular space. Any color space is adequate for computing a color match, as long as it can represent all of the perceptual colors in the CIE spaces uniquely. There are two additional criteria for a color space that can make it more or less desirable. It is desirable for the computation of the transformation to the color space to be fast. Since perfect matches are almost impossible to achieve, it is desirable for distances of color vectors in the space to correspond to perceptual distances, that is, if two colors are far apart in the color space, they look significantly different. Unfortunately, these two criteria are antagonistic. The color spaces that can best be used to measure perceptual difference require heavy computation.
