A model of spatio-temporal coupling for scene analysis in the primate visual cortex

We propose a new model of information processing in the visual cortex for the analysis of motionless images. It uses lateral inhibition and temporally segregated activations between cortical neurons, resulting in predictive coding. It is in agreement with some characteristics of the architecture of the visual cortex of the monkey. The model is based on the segregation of the times of arrival in the visual cortex of signals originating from the retina and relayed in the lateral geniculate nucleus. This distribution of delays depends both on the type of ganglion cell, of its position on the retina, and of its modulation by the cortical network. Indeed the parallel channels into which the retinal image is decomposed are represented as delay lines; and the resulting spatially uniform and discrete distribution of delays is modulated by spatial factors determined by the geometry of the retina.