CNN-based modeling of the human early vision system

Abstract

CNN (cellular neural network)-based retina model is introduced to simulate the human early vision system. By considering the retinal activity using CNN, we can begin to think about whole retinal interactions in space/time and consider the mechanism of a large population of cells as the edge detection and detection of moving stimuli. Furthermore, after simulating the output response of each retinal neuron, ERG, which is volume conductor potential recorded at the cornea and used as important diagnostic measure in eye clinic, is calculated. Each wavelet which composes the typical ERG has close relationship with the mechanism of a special retinal layer and by analyzing the each wavelet form we can guess the clinical state of special retinal layer.