On parallel channel modeling of retinal processes

Abstract

Based on previous work on parallel processing for visual motion detection and analysis ([1], [2], [3]) in which interesting relationships among number and random distribution of cells, overlapping degree and receptive field size with fault tolerance, accuracy of computations and performance cost were shown in practice, we now focus our attention on modeling a two parallel channel visual peripheral system consisting of three layers of neuron-like cells that account for motion and shape information of perceived objects. A tetraprocessor UltraSparc SUN computer is used for simulation and video-outputs in false color show the two-channel activity of the system. A train of input images presenting a moving target is analyzed by the neuron-like processing layers and the results presented as a video movie showing a colour coded version of neural activity