Skin Detection Using Color Pixel Classification with Application to Face Detection: A Comparative Study

This paper presents a comprehensive study of the pixel-based skin color detection techniques. Two main issues of the skin detection are the selection of the best color space and skin color pixel classification algorithm. A large set of XM2VTS face database is used to examine whether the selection of color space can enhance the compactness of the skin class and discriminability between skin and non-skin class in thirteen color spaces and six different skin color pixel classification algorithms. The results show that 1) the selection of the color space can improve the skin classification performance 2) the segmentation performance degrades only when chrominance information is used for classification 3) Bayesian classifier is found to perform better as compared to other classification algorithms. Piecewise linear decision boundary classifier algorithm outperforms all the other skin classification algorithms when it is used for images with good illumination conditions.