Abstract: A system for computer vision comprises a source of digitized video signals and a frame encoder circuit connected to the source of digitized video signals. The frame encoder encodes threshold crossing events at at least two threshold levels and stores encoded data for a two-dimensional area within the frame in a cache memory. The threshold levels are programmable. A computer is in communication with the encoder circuit at least via its data bus and address bus for programming the threshold levels and controlling the encoder and accessing the data in said cache memory. The computer has an associated main memory with a stored task for reading the cache memory and interactively controlling the frame encoder circuit and interpreting the data gathered thereby. Data is encoded by event type and pixel address wherein the event type is indicative of threshold crossed and direction of crossing.