Application of neural constraint satisfaction networks to vision

Summary form only given, as follows. The problem of constraint satisfaction is common in computer vision. The author maps this problem to a network where the nodes are the hypotheses and the links are the constraints. The network is implemented as a neural network which is then used to select the optimal subset of hypotheses which satisfy the given constraints. The author illustrates the use of a simple constraint satisfaction network and an augmented multilayered version for solving problems in perceptual organization.<>