Combining Color Spaces for Human Skin Detection in Color Images using Skin Cluster Classifier

Skin detection remains a challenging task over several decades in spite of many techniques evolved. It is the elementary step of most of the computer vision applications like face recognition, human computer interaction, etc. It depends on the suitability of color space chosen, skin modeling and classification of skin and non-skin pixels under varying illumination conditions. This paper presents a symbolic interpretation on the performance of the color spaces using piecewise linear decision boundary classifier in color images to find the winning color space (s). The whole task is divided into three processes: analysis of color spaces individually; analysis of the combination of two color spaces; and finally making a comparative analysis among the results obtained by the above two processes. For performing the fair evaluation, the whole experiment is tested over commonly used databases. Based on the success rate, false positive and false negative of each color spaces, the winner(s) has been chosen among single and the combination of color spaces.