Research On DT-CWT Subspace Features Fusion In Face Recognition

Since the19th century, biometric identification technology has been developing rapidly, especially the face, as a very important part of biological features, has been used in authentication and identification, information security, surveillance systems and other fields. But in the actual application process, because of some unfavorable factors the images of the face are vulnerable to be changed, these impacts affect the recognition accuracy.In view of the actual demand, into the appropriate transformation, the face image is mapped to the feature space, and then face is recognized, it can get good recognition effect. This paper mainly has done work as follows:Based on the research of2DPCA image dimension reduction algorithm and nearest neighbor algorithm, the face recognition system can be implemented; By introducing the traditional wavelet and Gabor transform algorithm, it can solve the problem of face image mode with the change of illumination conditions in face recognition system, and reduce the image noise, which is good for many posture face recognition system; Based the wavelet subband can not be good to get face local texture features and direction selectivity, so that you can not be a good features of human faces, and with Gabor wavelets for face nose, mouth, eyes the filter response is not very prominent, it can not remain the useful local texture features of face recognition, and then DT-CWT subspace features fusion algorithm was put forward. it can well solve the shortages of the traditional wavelet and Gabor wavelet, and resolve the problem of the recognition rate of every band is not the same as others.At last ORL face image database was used to test the algorithm, the experimental results show that:compared with other algorithms in the experiment (two-dimensional principal component analysis, traditional Wavelet, Gabor wavelet, tree complex wavelet feature extraction method), this article algorithm can significantly improve face recognition rate and it is an effective face recognition algorithm.