Face recognition using Homomorphic Filtering as a pre-processing technique

The appearance of the face will vary drastically when illumination, pose and expression change. Variations in these conditions make Face Recognition (FR) an even more challenging and difficult task. In this paper, we propose three novel techniques, viz., Homomorphic Filtering (HF), Image Flip (IF) and Raster Scan Mapping (RSM), to improve the performance of a FR system. HF is a pre-processing technique used to normalize the illumination variations. IF is used to neutralize pose and expression variations. RSM is used to map 2D to 1D, to increase the ability to store the intensity value of all the spotted lines. DWT and DCT are used for efficient feature extraction and a Binary Particle Swarm Optimization based feature selection algorithm is used to search the feature space for the optimal feature subset. Experimental results show the promising performance of the proposed techniques for FR on four benchmark face databases, namely, CMUPIE, Color FERET, UMIST and ORL databases.