Skin Segmentation Based on Human Face Illumination Feature

A novel skin segmentation scheme based on human face illumination feature is proposed in this paper. First, we perform a face detection in order to measure the face position and amount on image, and then analyze each face´s illumination feature. According these parameters, we classify the skin pixel and generate skin probability map, and finally fetch out a prefect skin mask for skin segmentation. Experimental results based on common dataset show that the proposed method can achieve 92.83% true positive rate (TPR) with 15.82% false positive rate (FPR), outperforming the traditional GMM skin segmentation method.