Journal of Beijing University of Posts and Telecommunications

  • EI核心期刊

JOURNAL OF BEIJING UNIVERSITY OF POSTS AND TELECOM ›› 2015, Vol. 38 ›› Issue (1): 51-54.doi: 10.13190/j.jbupt.2015.01.010

• Papers • Previous Articles     Next Articles

Feature of Local Gabor Spatial Histogram based on Dominant Neighboring Pixel for Face Recognition

LIU Jun, JING Xiao-jun, SUN Song-lin, LIAN Zi-feng   

  1. School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China
  • Received:2014-08-26 Online:2015-02-28 Published:2015-03-30

Abstract:

For face recognition system used in real-time application,the image feature representation should not only have high discriminative capability,but also the low dimensionality. A new face image representation-local Gabor spatial histogram based on the dominant neighboring pixel (LGSH-DNP) was proposed. Firstly,the face image is filtered by the Gabor filter bank. Each pixel of the resulted filtered images is labeled by the positions of the two neighboring pixels with the highest values,resulting in multiple descriptor images. Secondly,the spatial histogram is extracted from these descriptor images. Finally the weighted intersection histogram similarity measure is used to realize face classification. Extensive experiments are performed on face recognition technology face image database. It verifies the effectiveness of the proposed LGSH-DNP method.

Key words: local Gabor spatial histogram, dominant neighboring pixel, Gabor response magnitude image

CLC Number: