Neural Network for Dynamic Binding with Graph Representation: Form, Linking, and Depth-From-Occlusion

Abstract

A neural network is presented which explicity represents form attributes and relations between them, thus solving the binding problem without temporal coding. Rather, the network creates a graph representation by dynamically allocating nodes to code local form attributes and establishing ares to link. With this representation, the network selectivly groups and segments in depth objects based on line junction information, producing results consistent with those of several recent visual search eperiments. In addiction to depth-from-occlusion, the network provides a sufficient framework for local line-labelling processes to recover other 3-D variables, such as edge/surface contiguity, edge, slant, and edge convexity.