J4 ›› 2011, Vol. 38 ›› Issue (2): 61-65+111.doi: 10.3969/j.issn.1001-2400.2011.02.011

• Original Articles • Previous Articles     Next Articles

New algorithm for normalization of the face image

JIN Wei1,2;ZHANG Jianqi1;LIANG Jianxia2;WANG Jiaqiang2   

  1. (1. School of Technical Physics, Xidian Univ., Xi'an  710071, China;
    2. Beijing Institute of New Technology Application, Beijing  100012, China)
  • Received:2010-06-23 Online:2011-04-20 Published:2011-05-26
  • Contact: JIN Wei E-mail:jindaweizi2006@sina.com

Abstract:

This paper presents a novel face normalization method by integrating the location of eyes and the corner of the mouth for the normalization in the vertical direction. In the first step, the regions of eyes and the mouth are roughly detected by making use of Adaboost all through the whole face image. The Grey Integral Projection and the improved Local Gradient Operator algorithms are then applied to find the accurate positions of the eyes and the corner of the mouth in the face image, respectively. Finally, the face image is rotated and zoomed according to the relative position of eyes and the mouth in vertical and horizontal directions. Experimental results demonstrate that the proposed method can locate the feature points accurately and present better normalization performance by introducing the zoom criterion in the vertical direction.

Key words: face normalization, face recognition, mouth corners location, local gradient operator

CLC Number: 

  • TP391