Home // International Journal On Advances in Software, volume 8, numbers 3 and 4, 2015 // View article

Recognition of Human Faces in the Presence of Incomplete Information

Authors:
Soodeh Nikan
Majid Ahmadi

Keywords: face recognition; block based; effective subregion; partial image; incomplete information

Abstract:
Proposed face recognition in this paper is a block based approach. Gabor magnitude-phase centrally symmetric local binary pattern has been used to extract directional texture characteristics of face at different spatial frequencies. Centrally symmetric local binary pattern is applied on the sub-blocks of magnitude and phase responses of Gabor images, which is the important contribution of the proposed work. Sparse classifier is employed as local classifier to find the sub-blocks class labels. We have evaluated the performance of the proposed algorithm on AR and ORL databases. In real world scenarios, partial face images are available to recognize the identity of an unknown individual. By comparing the recognition accuracy on the recognition results of image sub-blocks, we find the location and size of the most effective face sub-region for identification. Moreover, Chi-Square weighted fusion of image sub blocks at decision level leads to significantly improved recognition accuracy. We also evaluate the performance of the proposed algorithm in the presence of incomplete information for low resolution and occluded images.

Pages: 450 to 456

Copyright: Copyright (c) to authors, 2015. Used with permission.

Publication date: December 30, 2015

Published in: journal

ISSN: 1942-2628