A novel optimal gabor algorithm for face classification

Over the past decade, most of the research in the area of pattern classification has emphasized the use of Gabor filter (GF) banks for extracting features. Typically, the design and the choice of GF banks are done on experimentation basis. In this paper, an attempt is made on obtaining an optimized set of GFs for improving the performance of face classification. The bacteria foraging optimization (BFO) is utilized to get optimized parameters of GF. The proposed BFO-Gabor technique is utilized to derive the feature vectors from the face images based on Gabor energy. These feature vectors are then used by probabilistic reasoning model (PRM) to perform the classification task. The ORL and UMIST datasets are utilized to investigate the superiority of the proposed approach. In addition, the experimental results of the proposed approach and the classical methods are compared. It is observed that the proposed BFO-Gabor method is superior than the classical methods.