Efficient parallelization methods of labeling algorithm

Abstract

Digital image processing is a field with broad applications. The development of technology has made it possible to introduce intelligent systems in distinctive areas such as medicine, robotics and astronomy. In this paper, the authors focus on indexing algorithms (also called labeling). Numerous studies have considered the various ways of implementing parallelization and the associated benefits. The indexing process involves assigning the same label to pixels of the same object. For the purpose of this study, a few algorithms proposed by Suzuki et al., Soh et al. and the method described by Tadeusiewicz and Korohoda were implemented. In order to parallelize the algorithms, the indexing algorithm of Niknam et al. was used and a method of partial parallelization was proposed.