We present a computational model of human texture perception. The model consists of three stages: (1) The image is convolved with a bank of even-symmetric linear filters followed by half wave rectification to give a set of responses (models outputs of V1 simple cells). (2) Inhibition, localized in space, within and among the neural response profiles which results in the suppression of weak responses when there are strong responses at the same or nearby location (models intracortical inhibition in V1), and (3) texture boundary detection using peaks in the gradients of the inhibited response profiles. Unlike previous attempts along these lines, our model is precisely specified, equally applicable to grey scale and binary textures, and is motivated by detailed comparison with psychophysics and physiology. We interpret experimental results on phase discrimination to show that in stage (1) responses of (a) center-surround mechanisms of differing widths, and (b) directional filters of differing orientations and widths which are even-symmetric about their axes are necessary and sufficient; and that odd-symmetric mechanisms are not used in human texture perception.

A computer implementation of this model has been tested on a large number of the 'classic' stimuli from psychophysical literature. Our model makes predictions about the degree of discriminability of different texture pairs which match very well with experimental measurements of discriminability in human observers due to Krose and Gurnsey & Browse.

From a machine vision point of view our scheme is a high quality texture edge detector which works equally well on images of artificial and natural scenes. The algorithm makes use of simple, local and parallel operations which makes it potentially realtime.





Download Full History