A Weighted Skin Probability Map for skin color segmentation
Biplab Ketan Chakraborty, Manas Kamal Bhuyan, Sunil Kumar · 2016
Skin color segmentation is important for several image processing, and computer vision applications. But, the accuracy of a color-based skin detection method is affected by the presence of some skin-like colors in the background regions. So, probabilistic approaches are more suitable for the skin detection as compared to hard decision-based approaches. A Skin Probability Map (SPM) of an image provides the probability of a pixel belonging to skin region. It is observed that the accuracy of a SPM-based skin detection method also depend on the chosen colorspace for the SPM. In this paper, a novel Weighted Skin Probability Map (WSPM) is proposed for the skin color segmentation. The WSPM is represented as a weighted sum of the SPMs obtained from different color spaces. Experimental results based on standard databases show that replacing the single colorspace-based SPMs with the proposed WSPM can reduce the overall detection errors significantly.