Abstract
This invention is about the human face detection method, in which color video signals of different continuous-pictures are used to judge the pixel difference of the time before and after in order to perform the correspondence of motion probability. The color video signal of single picture is again used to do the correspondence of skin-color probability in order to find multi skin-color assembly groups. Additionally, the correspondences of the motion probability and the skin-color fullness for the multi skin-color assembly groups are performed to decide the one among the multi skin-color assembly groups as the human face block. The color information of this human face block is again used to trace the human face block of different continuous-pictures.