A Self-Organizing Neural Network for Contour Integration Through Synchronized Firing

Abstract

Contour integration in low-level vision is believed to occur based on lateral interaction between neurons with similar orientation tuning. The exact neural mechanisms underlying such interactions, and their developmental origins, are not well understood. This paper suggests through computational simulations that synchronized firing of neurons mediated by patchy lateral connections, formed through input-driven selforganization, can serve as such a mechanism. Furthermore, we argue that different degree of such patchy connections established during development may explain why different areas of the visual field show different degrees of contour integration in psychophysical experiments. Introduction Contour integration in low-level vision means forming a single coherent percept (i.e. a continuous contour) from a discontinuous sequence of line segments. Humans are very good at contour integration; understanding the underlying mechanisms can give us insights into how perceptual...