The paper describes the design and operation of a real-time image processing system and outlines one of its application areas. The system consists of a dedicated hardware processor called RAPAC (a reconfigurable attached processor architecture for convolution) and a host computer which is used for algorithm development and RAPAC control. RAPAC uses hardware processor units and multiple image memories, in a software controlled architecture, to process 5 MHz streams of pixel data. This processing rate allows it to process a 256 × 256 pixel image in 20 ms, one field time of a standard TV camera. The result is either a new 256 × 256 pixel image generated from the old image or a reduced data set which describes attributes of features in the image. These attributes are used by the host computer to calculate a decision output concerning the content of the image.