Human Face Detecting System Based On Skin Color And Hausdorff Distance

An fully automated face recognition system includes two chief process, face detection and face recognition. Face detection is the first step and key factor. Due to the application of control of secure, vision detection, validation of identity, and new human-machine interface, face detection develops as an alone research subject.This paper is focused on the research of the single face detection based on color image, and a combined method of image segmenting preprocessing based on skin color and template matching is employed. Based on analysis of color space and comparison of different complexion's clustering, a skin color Gaussin probability model based on YCbCr color space is used. With this model, complexion is segmented and smoothly filtered, and two-valued processed. Then size and length-to-width geometrical analysis about object region is carried out by utilizing face's geometrical characters, in order that non-face region is removed and candidate face areas are output as processing result. At last these candidate areas can be determined by using a Hausdorff distance, and face region can be detected availably.The whole algorithm is implemented by using Visual C++. Experimental results show that the correctness ratio of detecting face and tracking speed are improved by combining these two methods, especially for different background and dimensions, And it is also robust and strongly adaptive to small variant pose and variety of expression. Due to the complexity of face itself and non-uniform illumination, missing and false detection is exist in the detection system. In a word, the veracity of face detection is above 89 percent. The system is practical to a certain extent.