A cortical architecture for the binocular perception of motion-in-depth

A model for the generation of cortical cells selective to motion-in-depth is presented. The model relies upon the computation of the total rate of change of the disparity through the combination of the outputs of monocular cortical units characterized by spatiotemporal receptive fields extracting temporal variations of phase information on the left and right retinal images. Each monocular unit of the cortical architecture can be directly compared to the Adelson and Bergen's motion detector, thus establishing a link between the information contained in the total derivative of the binocular disparity and those hold in the interocular velocity differences. Experimental simulations on stereo sequences evidenced that the model can quantitatively predict motion-in-depth information.