Human Face Detection Based On Skin Color And Statistic Method

Face Detection is defined the process confirming the the location, size and of all faces(if there is) in the input image.As a key technology for information processing, face detection has become a popular area of importance, very active research topic in Computer Vision and Pattern Recognition in recent years.The approach detecting the faces can be devided into two kinds: knowledge-based methods and statistical learning-based methods. Knowledge-based methods have high speed but low detection rate and can't detect multi-view faces. Statistical-learning based methods have high detection rate but low speed and are capable for multi-view faces. By combining the strengths of both, we have designed a system which has a higher detection rate and is capable for multi-views faces. The paper chooses Cb,Cr,H,I which has wonderful color clustering property to form"mixed skin model"and segment the skin, greatly narrowing the search area in the next step classifier.The paper adopts the statistical learning-based method proposd by Viola in his paper. We have made some improvements to the original algorithm, first used a new rectangular feature and modified the regular of weight update to make the adaboost adaptive to difficult samples.