Fast face detection and localization from multi-views using statistical approach

Abstract

Window-based face detection methods are fast. However their results are coarse, pose dependent and require fine face alignment for face analysis. Recently a statistical approach is introduced by Toews and Arbel [1], which is able to detect faces in multiple poses and does not require face alignment. However, their method is slow compared to the window-based method. In this paper, we proposed a method, which capable of detecting faces in multiple poses in near real time and also does not require face alignment. Experimental results show that our proposed method has comparable accuracy with the Toews and Arbel’s method but has significantly lower processing time.