DWT-based Face Recognition using Morphological Edge Detection as a pre-processing technique

The appearance of the face will vary drastically when expression, pose and illumination change. Variations in these conditions make Face Recognition (FR) an even more challenging and difficult task. In this paper, we propose two novel techniques, viz., Thresholded Wavelet Edge Enhancement Transform (TWEET) and Morphological Edge Detection (MED), to improve the performance of a FR system. TWEET is used to extract the approximation coefficients along with the prominent detail coefficients of the 1D-DWT of an image, thereby selecting only relevant features and enhancing FR. MED is a pre-processing technique used to normalize pose and illumination variations. The resulting pre-processed image contains the salient edge details of the face and prepares the ground for feature extraction. 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 Cambridge ORL, UMIST, Color FERET and Extended YaleB face databases, shows that the proposed system outperforms other FR systems. A significant increase in the recognition rate and a substantial reduction in the number of features are observed.