Describes the design of a complete, flexible system for perceptual organization in computer vision using graph theoretic techniques, voting methods, and an extension of the Bayesian networks called perceptual inference networks (PINs). The book also reviews and classifies prior work in the area.