Color image segmentation by unsupervised 2D histogram clustering and Dempster-Shafer region merging

摘要

In this paper, a color image segmentation method based on a new approach called bimarginal is proposed.Toudovercome the drawbacks of the classical marginal approaches, color components are considered in pairs in order toudhave a partial view of their inner correlation. Working with color images, the three possible combinations areudconsidered as three independant information sources. Each pairwise component combination is firstly analyzedudaccording to an unsupervised morphologic clustering which looks for the dominant colors of a 2D histogram. Thisudleads to obtain three segmentation maps combined by intersection after being simplified. The intersection processuditself producing an over-segmentation of the image, a pairwise region merging is done according to a similarityudcriterion with the Dempster-Shafer theory up to a termination criterion. To fully automate the segmentation, an energyudfunction is proposed to quantify the segmentation quality. The latter acts as a performance indicator and is used alludover the segmentation to tune its parameters.