Jürgen Golz
Christian-Albrechts-Universität zu Kiel, Institut für Psychologie

The visual stimulus associated with an object surface varies with the illumination falling on the object. To accomplish the constancy of perceived surface colours under changing illumination, the visual system must in some sense internalize the regularities of the chromatic interaction between observed surfaces and the incident light as these jointly determine the visual stimulus. In this talk I will present a theoretical analysis of the chromatic interaction between surfaces and illuminants in our natural environment. This analysis yields four regularities inherent in the retinal chromatic shifts induced by changes of illumination:
1. A multiplicative change of the cone excitations under changes of illumination, where the excitations of each cone type are scaled for all surfaces by the same factor.
2. A chromaticity-dependent change in luminance: surfaces similar to the illuminant in chromaticity are rendered as lighter than surfaces dissimilar from the illuminant.
3. A resistance of narrow-band surfaces to chromaticity shifts.
4. A compression of the chromaticity gamut with narrow-band illuminants.
I will also summarize psychophysical experiments that address the question which of the above mentioned regularities are taken into account by the human visual system for perceiving surface and illumination colours. In order to obtain almost constant surface colour percepts under different illuminants, the visual system must estimate the chromatic properties of the current illumination. One class of cues to the illumination are the so called chromatic scene statistics: statistics of the chromatic distribution within the retinal image received from the scene. For instance, under a reddish illuminant the retinal image becomes on average reddish and thus a reddish mean chromaticity within the retinal image could indicate a reddish illuminant. However, taking the mean retinal chromaticity as a cue to the chromatic properties of the illuminant faces a difficulty: a non-neutral mean chromaticity cannot be attributed unambiguously to a chromatically biased illuminant because to some degree it also could be caused by a chromatically unbalanced composition of surfaces in the scene. To enhance the estimation of the illumination, it would be helpful to take additional scene statistics into account if they vary systematically with the illuminant. Candidates for such additional scene statistics as diagnostically useful cues to the chromatic properties of the illumination can be derived from the above mentioned analysis of regularities in the chromatic interaction between surfaces and illuminants. The psychophysical experiments presented in the third part of this talk investigate which of these candidate cues are indeed taken into account by the human visual system.