Visual Processing Seminar

Bayesian Cortical Networks for Contextual Integration in Visual Processing, by Kyungim Baek. One of the challenges faced by the visual system is integrating cues within and across processing streams for inferring scene properties and structure. This is particularly apparent in the inference of object motion, where psychophysical experiments have shown that integration of motion signals, distributed across space, must also be integrated with form cues, such as figure-ground segregation.

In this work, we present a probabilistic network model in which local figure-ground representations encoded by direction of figure? (DOF) modulate the degree of certainty of local motion signals at line terminators. The strength of local motion suppression at extrinsic terminators is determined as a function of the belief in the local DOF, which is defined as the strength of the evidence for surface occlusion.

We first demonstrate the network model for inferring DOF, and show that motion integration can be mediated via a probabilistic representation of DOF. Simulation results show that the model can account for the continuum of perceptual bias seen for motion coherence and perceived direction of motion in psychophysical experiments.

Event Sponsor
Information and Computer Sciences, Manoa

More Information
Stephen Itoga, 956-3500, itoga@hawaii.edu

Ongoing Events