Parallel algorithm for gray-scale image segmentation
Harvey A. Cohen · 2002
One of the classic algorithms of mathematical morphology is the image segmentation algorithm of A. Rosenfeld and D. Pfaltz (1966). This algorithm in its "sequential" form, seeks to mark with a common label all connected pixels in an image, where the connectivity actually used in practice is 4-connectivity or 8-connectivity. The sequential Rosenfeld-Pfaltz algorithm (RS Algorithm) algorithm was presented for binary images; H.A. Cohen (199) has previously presented a one pass modification to the RS algorithm, and extended the algorithm to gray scale. The RS algorithm is not simple to parallelise in principle, as pixels in a connected subregion may be connected only via a chain of connected pixels in several other sub regions. We present a parallelisable algorithm for segmenting into regions, images of multiple gray scale, based on the one pass algorithm.