Face Detection and Segmentation for Video Surveillance

A fast approach of face detection for video surveillance was proposed.A symmetrical frame difference was used to acquire the motion area.A simple and fast segmentation algorithm,i.e.mid-point-guided growing,was developed to search the skin-like pixels and segment the face-like regions.The distribution information of shape and feature in each face-like region was examined and those regions satisfying with the knowledge of a human face were located and marked.The experimental results show that the accuracy and processing speed are very promising.The approach can be used to detect multiple faces with different poses and sizes.It is a practical method for developing visual surveillance system and can be used as a start point for face tracking research.