A Novel Approach for Skin Region Extraction in Color Images

In this paper, a novel approach for skin regions extraction of color images is presented. In this approach, a new method has been proposed that operates independent of a specific threshold value. For this goal, initially, a set of 13731 skin color samples from different databases such as CBCL and IFD is selected. A mixture function of three Gaussian functions is applied on these samples in order to create a skin color feature extractor. The result of applying this function on an input image is a gray level image called Skin Probability Map (SPM) image. Then, a method based on the contour plot on the gray level SPM image is used to extract the skin areas. Simulation results show that with the same true positive rate, our proposed algorithm has smaller false detection rate.