Face Recognition Based On Gabor Wavelettransform

Face recognition is a biometric technology of feature recognition processing great develop potentiality. Its main task is to correctly recognize the identification of different people from images or videos. Feature extraction is the key technology of face recognition, which directly affects the recognition rate. Researchers have shown that Gabor wavelet can capture the salient and local features of different frequencies and directions because of its well biometric vision characteristic. Therefore, it is widely used in the feature extraction of face recognition. The aim of this thesis is to study feature extraction based on Gabor wavelet, the main work are summarized as follows:(1)A comprehensive overview about the content and the develop situation of face recognition system are given. Deeply discussion is made on Gabor transformation as the theoretic supporting of the study of the thesis.(2)In order to avoid the lost of discriminated features which are extracted by using down-sampling method to process the Gabor features, a new method is proposed in this thesis. The new method selects the"T"area of the Gabor feature from images as the discriminated feature. Experimental results show that the proposed method is very effective.(3)Although the proposed method is very effective, the feature extracted by it is very high. So, the methods of dimension reduction-Principle Component Analysis (PCA) and Locality Preserving Projection (LPP) are used to reduce the dimension.(4)In the face recognition experiments, Gabor feature,"T"area of the Gabor feature and the feature obtained by using dimension reduction methods with factor of 2 to process the"T"area of the Gabor feature are extracted, the SVM of error correction is used to classify those features. Experimental results have indicated that the proposed methods of facial feature extraction can improve the performance of recognition system and also can raise the recognition rate effectively.