Texture correlation feature for Support Vector Machine-based face detection
Human face detection is an important step in any human face processing applications such as human face recognition systems, surveillance systems, face tracking, and expression recognition, et cetera. The problems associated with face detection is the fact that human face is a non-rigid object which is characterized by large variations. Therefore, it is difficult to model. Despite all the past efforts, many face detection methods are still too inflexible to be applicable for real world applications. In this thesis, our goal is to improve previous appearance-based face detection work. We examine the use of the texture correlation I feature for appearance-based face detection. It turns out that our system with the texture correlation feature can help to address some shortcomings of systems using the grey scale feature. Our system consists of two parts. The feature extraction part is responsible for extracting texture correlation features. The second part is a Support Vector Machine which performs the classification task.