Invited Session V: Binocular vision and interactions: The case against probabilistic inference: A deterministic theory of 3D visualprocessing.

Abstract

I will describe a new computational theory of 3D cue integration and introduce a novel theoretical framework to study 3D vision in humans. The proposed computational theory differs from the current mainstream approaches to the problem in two fundamentally different ways. First, it assumes that 3D mechanisms are deterministic processes that map a given visual stimulus to a unique 3D representation. Second, the proposed theory posits that 3D processing is heuristic, finding correct solutions to the problem only in ideal viewing conditions. The deterministic and heuristic nature of these mechanisms is inconsistent with Bayesian approaches that model brain mechanisms as processes of Bayesian inference aimed at deriving the most accurate and precise representation of 3D structures. These two main features of the proposed theory are implemented in a computational model that allows quantitative predictions of new phenomena. First, it provides an entirely different interpretation of Just Noticeable Differences, a hallmark measure of perceptual uncertainty. Second, it predicts specific perceptual distortions that are at odds with what previous accounts would predict. I will also discuss how the deterministic and heuristic nature of the proposed computational model points towards a re-evaluation of fundamental theoretical assumptions in perception research.