Image recognition: Visual grouping, recognition, and learning

Vision extracts useful information from images. Reconstructing the three-di mensional structure of our environment and recognizing the objects that pop ulate it are among the most important functions of our visual system. Compu ter vision researchers study the computational principles of vision and aim at designing algorithms that reproduce these functions. Vision is difficul t: the same scene may give rise to very different images depending on illum ination and viewpoint. Typically, an astronomical number of hypotheses exis t that in principle have to be analyzed to infer a correct scene descriptio n. Moreover, image information might be extracted at different levels of sp atial and logical resolution dependent on the image processing task. Knowle dge of the world allows the visual system to limit the amount of ambiguity and to greatly simplify visual computations. We discuss how simple properti es of the world are captured by the Gestalt rules of grouping, how the visu al system may learn and organize models of objects for recognition, and how one may control the complexity of the description that the visual system c omputes.