Information-Theoretic Estimates of Surface-Colour Coding in Natural Scenes

点击翻译按钮获取中文摘要

Colour vision has several functions, but one of the most important is in helping us to identify objects in the world around us, even though the lighting conditions may vary from occasion to occasion. With just three classes of cone photoreceptors in the eye, however, it is impossible for the eye to estimate uniquely the light-reflecting properties of surfaces. Judgements will therefore sometimes be wrong. Because of this physical limit on performance and the effects of other sensory and perceptual uncertainties, predicting the accuracy of surface-colour judgments, that is, the degree of observers' colour constancy , is difficult, especially when the surfaces are part of complex natural scenes. In working on this and related issues, we have taken advantage of methods of analysis based on Shannon's information theory, which provides an objective, quantitative, operationally oriented way to represent and assess performance. The general aim of this proposal is to validate this approach and to apply it to some specific questions arising from this analysis. These questions concern the extent of differences in the amount of information carried by achromatic and chromatic pathways in viewing natural scenes under different illuminations, the effects of the different spatial resolutions of these two pathways on information-carrying capacity, and the characteristics of opponent-colour coding in the chromatic pathway that are optimal for this task.