Leveraging spatial analysis on homogonous regions of color images for skin classification

Skin detection is one the most studied subjects in vision literature. Due to the appealing features of skin segmentation algorithms, they are widely used in different biometric applications such as face detection, face recognition, face tracking, hand gesture recognition, etc. However, several challenges such as nonlinear illumination, equipment effects, personal interferences, ethnicity variations, etc are involved which have been yielded to the inefficiency of many color based methods. Even though several interesting ideas have already been proposed, the problem has not been satisfactorily solved yet. This paper introduces a skin segmentation technique that addresses limitations of the earlier works. The proposed system consists of some steps including initial seed generation of skin map, Otsu segmentation in color images and a super two stage diffusion method. The initial seed of skin pixels is provided based on the idea of ternary image as there are certain pixels in images which are associated to human complexion with very high probability. The Otsu segmentation is performed on Cb and Cr of YCbCr color space, I of YIQ, H of HSV, a of Lab, u of Luv and c of Lch and the result accompanying with the edge map of the image is utilized in diffusion stage in order to annex initially unidentified skin pixels to the seed. Both quantitative and qualitative results demonstrate the effectiveness of the proposed system in compare with the state-of-the-art works.