NEURAL VISUAL CODING: IMAGE TO OBJECT REPRESENTATION

Johns Hopkins University, Baltimore MD

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Funds are requested to continue our studies of image parsing mechanisms in the visual cortex. Contrast border responses in V2 are selective for figure-ground direction (border ownership coding). The aim of the proposed research is to understand the role of this neural border ownership representation in the visual process. Three broad aims are identified for the next five years: The first is to analyze the mechanisms in border ownership assignment. Evidence shows that border ownership signals reflect the image context far beyond the scope of the conventional receptive field, but emerge with short latency and do not require attention. The aim is to study the global form influence (1) by using displays in which some critical features (curvature, corners, X- and T-junctions) are variably occluded, and analyzing the border ownership signal as a function of the exposed features, and (2) by adding visual noise to a low-contrast figure and correlating the noise pattern with the border ownership signal. The second aim is to study the dynamics of the border ownership signal, its rise time and persistence, whether there is memory for border ownership or if the signal decays passively, and whether border ownership assignment is preserved when a figure is moved. The fixation task will be used for aims 1-2. The third aim is to study the role of the image parsing stage as an interface for top-down processes in visual cognition. To study the interaction of border ownership signals with selective attention, monkeys will be trained to perform shape discrimination with partially overlapping figures, so that correct performance of the task requires border ownership assignment. To determine if central processes can influence the figure ground representation in the visual cortex, border ownership signals will be studied while the monkey is performing a visual search task. The studies will focus on area V2, but areas V1 and V4 will also be explored.