Lateral inhibition and the area operator in visual pattern processing

• Main Author: Connor, Denis John
• Language: English
• Published: University of British Columbia 2011
• Subjects: Vision > Mathematical models
• Online Access: http://hdl.handle.net/2429/36053

The static interaction of the receptor nerves in the lateral eye of the horsesoe crab, Limulus, is called lateral inhibition. It is described by the Hartline equations. A simulator has been built to study lateral inhibition with a view to applying it in a pre-processor for a visual pattern recognition system. The activity in a lateral inhibitory receptor network is maximal in regions of non-uniform illumination. This enhancement of intensity contours has been extensively studied for the case of black and white patterns. It is shown that the level of activity near a black-white boundary provides a measure of its local geometric properites. However, the level of activity is dependent on the boundary orientation. A number of methods for reducing this orientation dependence are explored. The activity in a lateral inhibitory network adjacent to a boundary can be modelled by an area operator. It is shown that the value of this operator along an intensity boundary provides a description of the boundary that is related to its intrinsic description — curvature as a function of arc length. Since the operator is maximal on an intensity boundary, this description has been called the ridge function for the boundary. A ridge function can also be obtained using a lateral inhibitory, network. The properties of this function are discussed. It is shown how ridge functions might be incorporated into a pattern recognition algorithm. A novel method for detecting the bilateral and rotational symmetries in a pattern is described. === Applied Science, Faculty of === Electrical and Computer Engineering, Department of === Graduate