Invariant features of contrast detection: an explanation in terms of self-similar detector arrays

Abstract

Human-detection performance for extended spatial contrasts appears to be subjected to several invariants. Such invariant features put general constraints on possible mechanistic interpretations. We show that the hypothesis of scale invariance is sufficient to determine the distribution of elementary sample apertures (receptive field centers) with respect to aperture size. The hypothesis that the elementary sample apertures are virtually identical except for size, which is augmented with general functional forms for pooling of elementary responses and coding properties for the elementary samples, suffices to derive the phenomenological facts. Because of the division of labor among units of different size, the concept of line-spread function loses a physiologically clear interpretation.