The properties of constancy models based on the proportionality rule of von Kries are examined in a series of simplified examples. It is found that the breadth of receptor-sensitivity functions causes metamerism, thwarting color constancy. Overlap of these functions limits the accuracy of von Kries adaptation for a more subtle reason: it causes nonzero off-diagonal elements in the transformation matrix relating object reflectance to receptor stimulations. Such off-diagonal elements make von Kries adaptation an inexact color-constancy scheme, even when the illuminant is restricted to prevent metamerism.

Two white lights may have the same chromaticity, and yet when used to light an array of objects may differ in their ability to reveal colors. For example, any white light can be matched by a mixture of two narrow-band lights, a yellow plus a blue. In this extreme case, reds and greens become black or brown and the red-green dimension is lost. At the other extreme, a light with three narrow bands, at the proper wavelengths, can brighten reds and greens and increase redgreen contrast, relative to a broad-band light such as daylight. Many commercial lights tend to dull reds and greens, relative to broad-band sources, a central reason that color rendering is a practical concern. A telling example is neodymium glass, a yellow-absorbing filter that is sometimes used to improve color rendering. This article seeks to bring these ideas to life through detailed graphical examples.

No abstract

Lights vary in their ability to render object spectral reflectances into color contrasts. When a light L1 is replaced by another L2, even if L2 matches L1 in chromaticity, systematic color shifts may occur, including a loss or gain of chromatic color. For instance, many familiar lights, when compared to daylight, dull red and green objects, rendering them closer to gray. An opponent colors method is appropriate to this discussion because it brings to the surface the notion of chromatic color, meaning actual departure from white or gray. In this article, an opponent-colors analysis leads to a matrix formulation that serves two purposes. The effects of replacing L1 by L2 are estimated with a 3x3 "rendering matrix" P. Given an object's tristimulus vector under L1, the method makes an approximate prediction of the new tristimulus vector under L2. Thanks to the opponent formulation, matrix element P22 quantifies the gain or loss of redness and greenness, while P33 expresses gain or loss of blueness and yellowness. These in fact are major effects, so the method is both quantitative and explanatory.