DescriptionPerceptual grouping is the process by which a set of image elements is divided into distinct “objects” or components. In this dissertation I propose a Bayesian framework for understanding perceptual grouping, in which the goal of the computation is to estimate the organization that best explains the observed configuration of image elements. I formalize the problem of perceptual grouping as a mixture estimation problem, where it is assumed that the configuration of elements is generated by a set of distinct components (or ”objects”), whose underlying parameters one seeks to estimate. In the first part of this dissertation I will propose a simplified version of the framework and show how it can be used to estimate the number of objects, more specifically clusters of dots, present in the image. Across two experiments I show how the model gives an accurate and quantitatively precise account of subjects’ numerosity judgments, while at the same time outperforming more standard accounts for dot clustering. In the second part of the dissertation this simplified framework is expanded to estimate a hierarchical representation of the image elements. This framework can easily be adjusted to different subproblems of perceptual grouping. Here I will show how an instantiation of our framework for contour integration, part decomposition, and shape completion can account for several key perceptual phenomena and previously collected human subject data.