Face Recognition using Dual Wavelet Transform and Filter-Transformed Flipping

Face Recognition (FR) under varying pose and illumination conditions is challenging, and extracting pose and illumination invariant features is an effective approach to solve this problem. In this paper, we propose three novel techniques, viz. Filtered Dual Wavelet Transform (FDuWT), Raster Scan Discrete Wavelet Transform (RDWT) and Filter Transformed Flipping (FTF), to improve the performance of the FR system. FDuWT is used for edge enhancement and denoising. RDWT is used to extract salient features. FTF is used to neutralize the pose variant features. Individual stages of the FR system are examined and an attempt is made to improve each stage. After efficient feature extraction through RDWT, a Binary Particle Swarm Optimization 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 three benchmark face databases, namely, Color FERET, Extended Yale B and Pointing Head Pose, show that the proposed system outperforms other FR systems.