Image Understanding Research at Rochester

Dana H. Ballard, Chris M. Brown, and Randal C. Nelson
Department of Computer Science
University of Rochester
Rochester, NY 14627

Abstract: Work in animate vision continues at all levels from low-level, preattentive object tracking and segmentation utilities to Bayesian techniques and decision theory for controlling and reasoning about the vision process. We have had success with pose-invariant object recognition, which we pursued using color and projectively invariant geometric features. Learning is taking on a position of increasing importance in our work as we strive for more adaptive behavior that needs less a-priori structuring.