A pose adaptive eye detection method using 3D face information

Abstract

This paper proposes an eye detection algorithm robust to head pose variation. First, face region is detected by Viola-Jones algorithm. Second, in order to detect eye robust to pose variation, a window template is generated from the 3D mean face model. Finally, eye is detected and localized by using this window template. In the experiments, the proposed method is tested on CAS-PEAL database.