Perceptual grouping as Bayesian mixture estimation

Perceptual grouping is the process by which the visual system organizes the image into distinct objects or clusters. Here we briefly describe a Bayesian approach to grouping, formulating it as an inverse probability problem in which the goal is to estimate the organization that best explains the observed set of visual elements. We pose the problem as an instance of mixture modeling, in which the image configuration is assumed to have been generated by a set of distinct data-generating components or sources (“objects”), whose locations and structure we seek to estimate. We illustrate the approach with three classes of source models: dot clusters, contours, and axial shapes. We show how this approach to the problem unifies and gives natural accounts of a number of perceptual grouping problems, including contour integration, shape representation, and figure/ground estimation. Highlight: A novel framework for perceptual grouping uses Bayesian mixture estimation to provide a unifying account of grouping problems, including contour integration, shape representation, and figure-ground estimation.