Face detection based on support vector machines

Face detection is a key problem in building an automatic face system such as face recognition and authentication. A number of approaches have been proposed for face detection. Recently, a novel statistical machine learning method, support vector machine, has been employed. Generally, the current SVM-based methods can be divided into two categories: component-based and whole face-based. It is difficult for the component-based method to extract the small face due to no enough information for each component exists. On the other hand, the whole face-based method is too much computationally expensive to build an effective system. In this paper we present a fast system named wavelet-SVM method to extract a wide range scales of faces from grey-scale images or color images with a preprocessing using a TSL color model. The system is not only accurate and effective, but also largely speeds the system up by applying a TSL B-G color model and multiresolution wavelet decomposition.