Computation of vision operators in hypercube processors

Abstract

Five low-level vision algorithms have been implemented on a 16-node iAPX 286/287 processor-based hypercube system (AMETEK S-14) by exploitation of its network embedding feature. This includes edge detection with the Sobel operator, histogramming, one-pass parallel binary image thinning, and noise-cleaning. The primary objective is to parallelize these algorithms in the ring and mesh networks of the S-14 hypercube system and to determine the actual speedup factor of parallel implementation over the sequential programming. Four 512 × 512 gray-level images have been processed concurrently in S-14 and a tenfold improvement in the speedup has been obtained compared to the sequential implementation in an iAPX 286/287 uniprocessor system (e.g. a single processor of S-14). This result implies that parallel implementation of vision operators on the relatively inexpensive array processor S-14 is ten times faster than sequential execution. It is also shown that the mesh topology of S-14 is more suitable for window convolution than the ring network of S-14.