Parallel analog algorithms for processing polygonal images on a systolic screen

Abstract

The systolic screen is a very natural parallel architecture for image processing. A square root n* square root n systolic screen consists of a square root n* square root n mesh-of-processors with each processor representing a pixel in a grid. The author studies computational geometry problems for polygonal images on such a screen. The algorithms are analog in that they simulate 'physical' processes based on the homeomorphic representation of polygons on the systolic screen. He obtains simple optimal parallel analog algorithms, subject to some restrictions due to the limited resolution of a systolic screen, that run in O( square root n) time on a square root n* square root n systolic screen, for various problems on polygons.<>