Robust Face Detection Combined Skin Color Features, Template Matching and VQ Histogram Method in Practical Environments

Abstract The face detection in complex background is still difficult by existing techniques. We present an approach to real-time face detection in practical environrnents using multi-module integration. We combine color extraction module, geometrical method, template-based module and face recognition module named VQ histogram method into a single robust face detection system. First, the skin color is used to locate the skin color areas and possible face candidates aze extracted by geometrical constraints, then template matching is carried out using appropriate-sized templates to extract the point of face, fmally a very simple yet highly reliable face recognition method called VQ histogram method is utilized to verify whether it is the face or not. The method is proved very effective in a practical environrnent.