Decision Making Journal Club
Local Binocular Motion Perception as Bayesian Estimation
Local image velocities on the retina of the left and right eye reflect important 3D environmental information. The human visual system encodes this information binocularly before it is integrated into a single 3D percept. Here it is shown that the aperture problem of local line motion is solved by a system that estimates 3D velocity under uncertainty. In a probabilistic model of binocular motion perception it is assumed that (i) velocity constraints arise from inverse projections of 2D image constraints in a binocular viewing geometry, (ii) noise in image and depth processing is independent, and (iii) slower motions are more likely to occur than faster ones. We found that instantiation of this Bayesian model successfully accounts for a range of perceived 3D motion directions under ambiguity.