On Considering Uncertainty and Alternatives in Low-Level Vision

In this paper we address the uncertainty is­ sues involved in the low-level vision task of image segmentation. Researchers in com­ puter vision have worked extensively on this problem, in which the goal is to partition (or segment) an image into regions that are ho­ mogeneous or uniform in some sense. This segmentation is often utilized by some higher­ level process, such as an object recognition system. We show that by considering uncer­ tainty in a Bayesian formalism, we can use statistical image models to build an approx­ imate representation of a probability distri­ bution over a space of alternative segmenta­ tions. We give detailed descriptions of the various levels of uncertainty associated with this problem, discuss the interaction of prior and posterior distributions, and provide the operations for constructing this representa­ tion.