A fast and robust algorithm for face detection and localization

This paper presents a fast and robust face detection and localization approach. The major techniques employed in method are the support vector machine (SVM) and the generalized symmetric map. The proposed face detection system consists of 3 stages. In the first stage, a feature detection technique is used to hypothesize candidate regions by looking for gross features. The facial features of the concern are the pair of eyes. By making use of the symmetrical property of eyes, a generalized symmetry approach is used. The exhaustive pairing of eyes gives the positions of potential face candidates. In the second stage, a view-based approach is used to verify the face candidates obtained in the first stage. A SVM is trained to do the face candidate verification. The last stage of the system precisely locates the eye positions. The proposed detection system was evaluated using the Bern face database. A detection rate of 96.67% was obtained.