Abstract: Machine Vision Research at the University of Rochester centers around the concept of active or behavioral vision. Vision is considered as an interactive process between a system and its environment, and this relationship is explicitly invoked to provide constraints to make machine-vision applications tractable. Work in this area has focussed on four main areas: integration of visual and motor control, learning for robot skill and visual model acquisition, task dependent allocation of computational and physical resources, and parallel real-time systems integration and support. We also continue to develop real-time low-level primitives for integration into larger systems.