Toward a computational theory of shape: An overview

Abstract

Although the shape of objects is a key to their recognition, viable theories for describing shape have been elusive. We propose a theory that unifies the competing elements of shape—parts and protrusions—and we develop a framework for computing them reliably. The framework emerges from introducing conservation laws to computational vision, and has application in areas ranging from robotics to the psychology and physiology of form.