Image Feature Extraction on Connection Machine CM-5

In this paper, we present a fast parallel implementation of an image feature extraction task on Connection Machine CM-5. We show that, given a 2048\Theta2048 grey level image as input, the extraction of image features, which includes edge detection, thinning, linking, and linear approximation, can be performed in less than 1.2 seconds on a partition of CM-5 having 512 processing nodes. A serial implementation written in C on a Sun Sparc 400 takes more than 8 minutes. Experimental results on various sizes of images using various partitions of CM-5 are also reported. The software has been developed in a modular fashion to permit various techniques to be employed for the individual steps of the processing. Our technique starts by modeling the communication and computation features of the machine. Using this model, scalable algorithms are designed. 1 Introduction Image feature extraction is a fundamental process in vision. In general terms, the primary task of feature extraction is to ex..