Imaging a computational process in the visual cortex

Visual perception is a computational process reconstructing the 3D visual world based on the 2D images viewed by the two eyes, which inherently contains ill-posed problems due to information deficiency. A typical example is the correspondence problem in finding partner points between the two retinal images. A neural network model for solution of the correspondence problem was proposed by Marr (1982, Vision. San Fransisco: Freeman and Co.), constructed of groups of neurons selectively responsive to different disparities of visual stimuli shown to the two eyes. This model solves the correspondence problem as a relaxation process of excitation working between the neuronal groups responding to the same disparity antagonized by inhibition working between those responding to different disparity. Optical recording of responses evoked in the rat visual cortical slice by electrical stimulation of the white matter demonstrated the relaxation process of excitation antagonized by inhibition. Copyright (C) 1996 Elsevier Science Ltd.