This paper presents a new technique for computing polygonal masks for image analysis and computer vision applications. The procedure does not require random access of the image memory and is, unlike its predecessors, highly parallel. It is based on efficient generation of coordinate-reference gray level images (ramps) and other simple architectural features like look-up tables. Even though our algorithm can be efficiently implemented in existing pipeline processors, we hope that future designs of general purpose image processing systems will benefit from this result so that future configurations will provide some parallel hardware for generation and manipulation of gray level ramp images. © 1985.