Parallel shrinking algorithms using 2-subfields approaches

Abstract

A class of parallel shrinking algorithms is introduced which uses two subfields arranged in a checkerboard pattern. For this class of algorithms sufficient conditions are derived for maintaining image connectivity and guaranteeing algorithm convergence and correct shrinking to single pixel residues. In comparisons with other parallel shrinking algorithms instances of the 2-subfields class are shown to exhibit substantially higher parallel speed for specific artificial and real test images.