Central Processing of Visual Information

Weill Medical Coll Of Cornell Univ, New York NY

Investigators

Linked publications & trials

Abstract

Description (provided by applicant): Though current computational models of visual processing provide a good account of processing of simple images such as lines and gratings, our understanding of the processing of natural scenes is substantially incomplete. Several lines of reasoning indicate that the fundamental reason for this is that current models do not account for sensitivity to higher-order image statistics - the distinguishing feature of natural scenes. Until now, sensitivity to high-order image statistics has eluded systematic study because they constitute an enormous set of parameters. The goal of this research is to surmount this barrier. Via psychophysical studies of texture perception, we propose to test two powerful hypotheses that will tame the complex domain of high-order image statistics. Aim 1 will test a hypothesis that simplifies how image statistics combine. Specifically, we hypothesize that the interaction of pairs of image statistics can be described by a quadratic combination rule, and interactions of multiple image statistics can be predicted from their pair wise interactions. Aim 2 will test the hypothesis that only small subsets of image statistics are visually salient. Specifically, we hypothesize that only two kinds of image statistics - "first-order histogram statistics" and "local correlation statistics" - are visually salient, and that a much larger set of image statistics, "high-order histogram statistics", are not perceptually relevant. Aim 3 will combine these two simplifying strategies, to account for perception of the complex multiscale high-order statistics present in natural scenes. PUBLIC HEALTH RELEVANCE: The long-term goal of this project is to understand how the brain analyzes incoming visual information. An enhanced understanding of this process will advance our ability to diagnose and remediate disturbances of perception, which cause significant morbidity in conditions as disparate as amblyopic, Alzheimer's disease and stroke.