Previous: From Visual Stimuli to Color Space
Up: From Visual Stimuli to Color Space Next: Assumptions and Limitations
General
The first ``half'' of the mapping 
needs to transform a visual
stimulus 
into a point in a color space. In Shepard's terms,
this is a psychophysical function that maps physical parameter space into
psychological space [Shepard 1987]. The reasons for using a color-space
transformation are many:
- The psychophysical literature on color perception commonly uses a
color-space model to represent color percepts; see
Chapter
(p. ). If we want to
evaluate the psychological validity of our model, using similar
representations allows us to compare results easily.
- The color-space model reflects important constraints on color
perception; e.g. the constraint that similar spectra by and large evoke
similar percepts is reflected in the mapping of similar spectra onto nearby
points in the color space. This is of course a special case of the
continuity constraint introduced by [Marr 1982].
- The color-space model allows us to easily model some of the tasks a
cognitive theory has to explain (Section p. ), e.g.,
- Discrimination: Two stimuli are discriminable if and only if
they map into distinct points in the color space (data on the minimum
differences people can discriminate under varying circumstances allow us to
adjust the ``resolution'' of the space to match human performance, if
desired).
- Similarity judgment: We can straightforwardly model similarity
between different stimuli as the distance between the corresponding points
in the color space (the geometry of the space need not be Euclidean,
however).
- Identification: we can associate regions of the color space
with foci and boundaries of color categories, as shown in
[Boynton \& Olson 1987], which is of course a consequence of the continuity
constraint mentioned above.
The new color space I will describe is based as closely as closely as
possible on known data about the neurophysiology of color vision. In
particular, an opponent model of color vision as described in
[Dow 1990][De Valois \& Jacobs 1968][De Valois et al. 1966] seems to be
appropriate, since it conforms with widely accepted psychophysical theories
of opponent color systems, particularly the Young/Hering theory, cf.
[Boynton 1990][Boynton 1979].