Face Recognition Using Gabor Filter Based Feature Extraction with Anisotropic Diffusion as a Pre-processing Technique

Face Recognition (FR) across pose, illumination and expression is a problem of fundamental importance in computer vision. In this paper, we propose two novel techniques, viz., Anisotropic Diffusion (AD) based pre-processing and Gabor filter based Feature Extraction, to improve the performance of a FR system. AD preserves the edges leading to facial image smoothing and enhancement. Gabor filter is used to capture facial features aligned at specific angles. Along with these, a Binary Particle Swarm Optimization based feature selection algorithm is used to search the feature space for the optimal feature subset. Individual stages of the FR system are examined and an attempt is made to improve each stage. The proposed approach has been tested on four benchmark face databases, namely, ORL, Color FERET, Cropped Yale B and FEI datasets, and demonstrates superior performance compared to existing methods in the presence of pose, illumination and expression variations.