Face Recognition Using Hough Transform Based Feature Extraction

Abstract

The sensitivity to illumination variations is a challenging problem in Face Recognition (FR). In this paper, a novel feature extraction method based on Hough Transform peaks is proposed to address this problem. Individual stages of the FR system are examined and an attempt is made to improve each stage. Block-wise Hough Transform Peaks are used for efficient feature extraction and a Binary Particle Swarm Optimization (BPSO) based feature selection algorithm is used to search the feature space for the optimal feature subset. Experimental results, obtained by applying the proposed algorithm on benchmark face databases, namely, Extended Yale B, CMU PIE, CAS-PEAL and Color FERET databases, show that the proposed system outperforms other FR systems by accounting for the illumination variations that are commonly observed in face images.