Derivative of Gaussian functions as receptive field models for disparity sensitive neurons of the visual cortex

Abstract

Non-symmetric Gabor functions are particularly useful neurophysiologic models of simple cell receptive fields which respond to disparity characteristics of visual stimuli. The left and right visual channels of these simple cell have receptive field profiles which are phase-shifted alterations of the same one-dimensional Gabor function. These profiles resemble derivative of Gaussian (DOG) functions, although DOG functions have never been evaluated using neurophysiologic data from disparity sensitive neurons. Here the authors demonstrate (a) the space frequency response characteristics of DOG functions as a model of a single receptive field, and (b) the space-frequency response characteristics of the combinations of two same-order DOG functions as a model of a simple cell disparity neuron which combine left and right receptive fields. Combining left and right visual fields in four different ways resulted in two characteristic patterns for both even and odd ordered derivatives. Each pattern appeared to be useful, in a different way, for detecting disparity information. These results suggest that DOG functions can be used to produce a set of equations for detecting disparity information.