Cortical simple cells: modeling physiology and image processing applications

Abstract

One of the fundamental questions in visual system physiology is how cells acquire their receptive field properties. The present paper addresses the question of how cortical simple cell behavior originates through the specification of an associated neural circuit which nonlinearly combines inputs from ON and OFF streams. The responses of the model closely follow the static nonlinear behavior observed in the recordings of Hammond and MacKay (1983) and suggest that simple cells should not be viewed as linear devices as commonly assumed since the work of Hubel and Wiesel (1965). We also evaluate the model with both camera and artificial images. Our circuit exhibits a higher positional selectivity (better contrast localization) when compared to a corresponding linear scheme and can be used as an effective "edge detection" algorithm.