Computing the organizations and shapes of two-dimensional dot patterns: a perceptual-level approach

Abstract

A simple perceptual-level scheme for computing the organizations and shapes of two-dimensional dot patterns is described. In this approach we use n neighborhood concept based only on simple proximity considerations (in contrast to more complex definitions of neighborhood based on directional constraints, etc.) to develop pictorially oriented algorithms for the generation and grouping of clusters in dot patterns, for the recovery of their perceptual boundaries, and for their shape articulation. A wide range of experimental results substantiates the validity of this approach as well as the utility of PAX as a well-structured parallel processing language suitable for modeling perceptual-level processes.