ACTIVE CONTOUR BASED IMAGE SEGMENTATION: A PARALLEL COMPUTING APPROACH

摘要

Segmentation is a fundamental task in image processing, but its hard computational requirements represent a not easily defeated drawback, mainly for tasks in which real time computing is a primary requirement, such as critical medical applications. In the present paper a parallel algorithm for image segmentation will be proposed, which allows fast convergence for active contour algorithm. A parallel computing approach to this problem has led to develop a multithreaded application based upon a parallel algorithm that allows reducing processing time. The algorithm has been implemented on a shared memory multiprocessor machine, using ANSI C++ language and Posix Threads libraries, in order to exploit code portability. Two basic schemes have been proposed for parallelism development and performance issues have been analyzed for both implementations.