Analyses of the Watershed Transform

International audienceIn the framework of mathematical morphology, watershed transform (WT) represents a key step in image segmentation procedure. In this paper, we present a thorough analysis of some existing watershed approaches in the discrete case: WT based on flooding, WT based on path-cost minimization, watershed based on topology preservation, WT based on local condition and WT based on minimum spanning forest. For each approach, we present detailed description of processing procedure followed by mathematical foundations and algorithm of reference. Recent publications based on some approaches are also presented and discussed. Our study concludes with a classification of different watershed transform algorithms according to solution uniqueness, topology preservation, prerequisites minima computing and linearity