A Neural Model of Visual Contour Integration

Part of Advances in Neural Information Processing Systems 9 (NIPS 1996)

Bibtex Metadata Paper


Zhaoping Li


We introduce a neurobiologically plausible model of contour inte(cid:173) gration from visual inputs of individual oriented edges. The model is composed of interacting excitatory neurons and inhibitory in(cid:173) terneurons, receives visual inputs via oriented receptive fields (RFs) like those in VI. The RF centers are distributed in space. At each location, a finite number of cells tuned to orientations spanning 1800 compose a model hypercolumn. Cortical interactions modify neural activities produced by visual inputs, selectively amplifying activities for edge elements belonging to smooth input contours. El(cid:173) ements within one contour produce synchronized neural activities. We show analytically and empirically that contour enhancement and neural synchrony increase with contour length, smoothness and closure, as observed experimentally. This model gives testable predictions, and in addition, introduces a feedback mechanism al(cid:173) lowing higher visual centers to enhance, suppress, and segment contours.