A computational model of early vision based on synchronized response and inner product operation

Abstract

A new early-vision complete computational model is proposed in this paper, which mainly based on probabilistic reasoning for visual information processing and combination of the synchronized response with the sparse representation. The model consists of multi-scale filtering, phase synchronization and inner product operations, in which the spatial localization, orientation and band-pass characteristics of the functional columns distributed in the cortex V1 can respond to local features in image patches, which obtained from division of retinal image by means of orthogonally divided into image patches according to the size of the receptive field of a ganglion cell. Theoretical analysis and simulated results show that at the system level, the model reflects the nature of the excitation of neurons in the V1 cortex by local characteristics of the external stimuli. Therefore, it may be having some use as a reference in investigations of neural mechanisms in visual information processing.