next up previous contents
Next: Is the Mapping Up: Mapping the Spectrum Previous: Mapping the Spectrum

Is Perceptual Color Space a Metric Space?

The first question that should be considered is, ``Does the perceptual color space have a metric?'' For a space Y to have a metric, there must exist a function for that has the properties (1) if and only if ; (2) ; and (3) for all [Itô 1993, page 1014,]. The third property is called the triangle inequality.

Color researchers have often made an implicit assumption that the perceptual color space has a metric. This is one of the foundation blocks upon which all colorimetry is based. The three formal conditions for a metric space can be reformulated as the following assumptions for the existence of a perceptual color metric space.

  1. If there is no noticeable difference between two stimuli, they occupy the same point in perceptual color space.
  2. The difference between color A and color B is the same as the difference between color B and color A.
  3. The difference between color A and color C is less than or equal to the difference between color A and color B plus the difference between color B and color C.
These three assumptions seem reasonable enough, although the triangle inequality requires empirical evidence. According to Indow [Indow 1991] subjects show subadditivity in judging color differences, that is even when , and are adjusted so as to try to approach collinearity. This evidence would tend to support the existence of the triangle inequality in the perceptual color space.

A stronger form of a metric space is a Euclidean metric space. An n-dimensional metric space Y is considered to be Euclidean when for all points , in the space. There have been efforts to create a representation of perceptual color space such that it conforms to a local Euclidean metric [Wyszecki & Stiles 1967], a so--called uniform color scale. However, a fully Euclidean representation of perceptual color space has as yet eluded researchers.

A weaker form of a metric space is a Riemann space. The metric in a Riemann space is the full quadratic form. Thus in an 3--dimensional Riemann space R, the metric is defined as follows:

where for . When the cross--product terms are zero, a Riemann space reduces to a Euclidean space.



next up previous contents
Next: Is the Mapping Up: Mapping the Spectrum Previous: Mapping the Spectrum



Steven M. Boker
Sun Feb 12 19:24:36 EST 1995