Previous: General Up: From Visual Stimuli to Color Space Next: Modeling Neurophysiological Data

Assumptions and Limitations

In modeling color perception, I will make some limiting assumptions to make the problem tractable:

  1. I am concerned with single-point determination of perceived color only; hence, I disregard spatial adaptation and interactions in color perception [Boynton 1979]: I will not be concerned with such phenomena as induced colors or Mach bands [Boynton 1979]. Context is only taken into account implicitly to the extent that it is necessary for the discrimination and identification of basic color categories. For example, a color like brown is not distinguishable from orange in the so-called aperture mode of viewing (in isolation, through a small hole in a screen), because brown is a ``dark color'' which is only defined with respect to brighter colors in its surroundings. When we examine the spacing of basic color categories in a color space, we implicitly take this into account by comparing the location of brown to that of orange or white, for instance.

  2. I assume a fixed adaptation state of the visual system; i.e., I am not concerned with issues of color constancy (invariant perception under varying light sources), pigment bleaching, temporally induced colors, and the like [Boynton 1979].

  3. I assume foveal photoreceptors as sensors, and their spectral sensitivities as defined in the CIE 1931 standard observer data [Wyszecki \& Stiles 1982]. Although there is some evidence that there are inaccuracies in these data, and several revisions have been proposed and published, the standard is so well established that I have chosen to adhere to it. From a practical point of view, it is also the basis of the NTSC color TV standards and corresponding camera specifications. Moreover, the existing revisions like Judd 1951, CIE 1968, and Vos 1978 [Wyszecki \& Stiles 1982] are not so different that they would have much effect on the relative locations of basic color categories in the color space.

Although these assumptions and limitations represent a considerable simplification in some ways, they do not render my work useless for practical applications, as I will show below.

lammens@cs.buffalo.edu