Face Recognition Using Mutiple-Channel Gabor Phase Feature

Gabor phase feature-based multiple channel assembling algorithm is presented for face recognition,in which Gabor phase features of face in different scales and orientations are firstly extracted and then combine these features for matching.Four main steps are involved in the proposed algorithm:(i) Gabor phase features of different scales and in different directions are extracted by the convolution of Gabor filter banks and original gray face images;(ii) two-directional two-dimensional principal component analysis is made to conduct dimensionality reduction of Gabor phase features from all directions;(iii) assembling the reduced feature matrixes from all directions,an enhanced feature matrix is formed,and quantizing the enhanced feature matrix,the final binarry face template is achieved;(iv) the Hamming metric based nearest neighbor classifier is used for classification.The results of face recognition experiments by the ORL and Yale face databases show the effectiveness of the proposed method and theoretical analysis proves its role of lowering the amount of computation.