Adaptive face detection based on normal distribution model of skin color
HUA Xiaobin;YUAN Mingxin;WANG Binbin;ZHANG Limin;School of Mechanical Engineering,Jiangsu University of Science and Technology;Zhangjiagang Industrial Technology Research Institute,Jiangsu University of Science and Technology;
To realize adaptive detection of different types of faces,and improve accuracy and efficiency of face detection,using good clustering of skin color distribution in color space,an adaptive face detection algorithm is presented based on normal distribution model of skin color. First,on the basis of luminance compensation of a face picture,a normal distribution model of skin color is built in YCb Cr color space,and the probability that face pixels belong to the zone of skin color( namely the similarity between them) is calculated through the model.Then the skin similarity is subjected to median filter and gray processing. Finally,a face is quickly segmented through an adaptive threshold method and located. The experimental results show that the proposed algorithm improves the detection rate and reduces the error detection rate of different types of faces; furthermore,its location accuracy and real-time performance are further improved.