Original scientific paper
Sparse representation for pose invariant face recognition
; Department of Electronic and Information Engineering, Taizhou Polytechnic College, Taizhou, Jiangsu Province, CHINA
Wei Shen ; Department of Electronic and Information Engineering, Taizhou Polytechnic College, Taizhou, Jiangsu Province, CHINA
Yumin Zeng ; School of physical science and technology, Nanjing Normal University, Nanjing, Jiangsu Province, CHINA
Face recognition is easily affected by pose angle. In order to improve the obustness to pose angle, we need to solve the pose estimation, face synthesis and recognition problem. Sparse representation can represent a face image with linear combination of atom faces. In this paper, we construct different pose dictionaries using face images captured under the same pose angle to estimate pose angle and synthesize front face images for recognition. Experimental results show that sparse representation can estimate pose angle accurately, synthesize near frontal faces very well and significantly improve the recognition rate for large pose angles.
Visits: 612 *