Bayesian Perception and Representation of Visual Motion

Eero P. Simoncelli

Center for Neural Science, New York University

The pattern of local image velocities on the retina encodes important environmental information. Although humans are generally able to extract this information, there are conditions under which they make substantial "mistakes". We've shown that these are a natural consequence of a system that attempts to optimally estimate velocity from visual input, given that (a) there is noise in the initial measurements, and (b) slower motions are more
likely to occur than faster ones.  I'll develop this model, show that it can account for a variety of data from human subjects, and discuss the implementation of these calculations in the primate brain.