Face Recognition using Block-Based DCT Feature Extraction

Abstract

Face recognition (FR) with reduced number of features is challenging and energy based feature extraction is an effective approach to solve this problem. However, existing methods are hard to extract only the required low frequency features, which is important for capturing the intrinsic features of a face image. This paper proposes a novel Block-Based Discrete Cosine Transform (BBDCT) for feature extraction wherein each 8x8 DCT block is of adequate size to collect the information within that block without any compromise. Individual stages of FR system are examined and an attempt is made to improve each stage. A Binary Particle Swarm Optimization (BPSO)-based feature selection algorithm is used to search the feature vector space for the optimal feature subset. Experimental results show the promising performance of BBDCT for face recognition on 4 benchmark face databases, namely, ORL, Cropped UMIST, Extended Yale B and Color FERET databases. A significant increase in the overall recognition rate and a substantial reduction in the number of features, are observed.