Feature-Based Improved Face Recognition By Gabor Wavelet

Face Recognition, which has been studied for more than 30 years, is one of the most challenging researches in pattern recognition and machine vision, it can be widely applied in the public security, information security and human-computer interaction. Generally speaking, the face recognition system consists of face detection. feature piont location, image pre-processing, feature extraction and face recognition. The thesis studies the correlative issues based on the former works, the main points are as follows:(1)First. introduce background, significance, development history and the research progress abroad and at home in face recognition. Second, build a face recognition system to discuss the major process of building a work of the various sectors of the technology required, the algorithm is introduced and studied.(2)An improved algorithm for face recognition based on the classical 2D Gabor wavelet in feature extraction phase is proposed. Firstly, using subblock sampling and SFFS algorithms select sampling point of feature extraction and 40 feature matrices which are reconstructed with the same scale and the same directions 2D Gabor wavelet transform results are obtained. Secondly, the dimensionality reduction and denoised technique with PCA are applied to form the new training samples. Lastly. SVM classifier is constructed and the vote decision strategy is used to determine the recognition result.(3)The results of simulations on ORL and Yale face database show that comparing with the traditional 2D Gabor wavelet algorithms, the proposed algorithm improves detection speed obviously and obtains a satisfactory recognition ratio with small sample problem in face recognition. At the same time, the proposed algorithm has strong robustness on parametric selection in PCA and SVM. The advantages not only settle down the difficult problem of selecting the kernel parameters in SVM partly but also expand its application limits.