Technical Report UT-CVIS-TR-96-002
Perceptual organization of two-dimensional patterns
Wilson S. Geisler and Boaz J. Super

Abstract

A theoretical framework and computational model of perceptual organization in two-dimensional patterns was developed and tested. In the model, input images are initially processed by a dense array of cortical neurons, whose properties are consistent with recent analyses of single neuron responses in primary visual cortex. Image structure is then discovered by interleaving pattern-matching processes with grouping processes, while enforcing a generalized figure-ground principle. To test the model and estimate critical parameters, a series of three-pattern grouping experiments measured how the stimulus dimensions of shape, spatial position, size, and orientation combine in determining perceptual grouping strength. The model accounted for the measured trading relations among these stimulus dimensions, as well as a number of perceptual organization demonstrations.