Face Recognition Technology Research Based On Contourlet Transform And Subspace Analysis

Face recognition is one of main methods of status authentication which based on technology of biologic characters recognition and has become hot topic in the field of patterm recognition and artificial intelligence. Exploring the high recognition rate algorithm for face recognition has great theory significance and application value.Wavelet analysis is an effective way in the field of signal processing and high-tech in the field of information acquisition and processing. Wavelet is an unified processing framework for various signals processing method, such as muti-scale analysis, time-frequency analysis and sub-band coding. The useful information can be extracted by wavelet transform. Wavelet analysis shows excellent character in the one-dimensional space, but this excellent feature can not be simply extended to two-dimentional space or higher dimentional space. In high dimensional spaces, multiscale geometric analysis has more significant advantages.Contourlet transform is a new image representation scheme which not only possesses the main features of wavelets (namely, multi-scale and time-frequency localization), but also has directionality and anisotropy. The subspace methods have been the most popular approach owing to their appealing properties, such as low time-consuming, good performance on expression and separation.In this paper, a method based on contourlet transform and Principal Components Analysis (PCA) for face recognition is proposed. In the method, the features of the low frequency in contourlet transform are extracted and combined with PCA for face recognition. The recognition rate is researched.At the same time, the other method that based on contourlet transform and Kernel Fisher Discriminant Analysis (KFDA) for face recognition is proposed. In this method, the features of the low frequency and high-frequency directional subband each level in contourlet transform are extracted respectively and combined with KFDA for face recognition. The recognition rate and recognition time are researched. The expremental result shows that combining the coefficients of the low frequency with PCA achieve excellent recognition rate and combining the coefficients of the low frequency with KFDA not only achieve higher recognition rate but also decrease the recognition time. High-frequency directional subband is helpful for recognition, but the recognition rate is low. Combining low frequency with high frequency directional subband can optimize the recognition rate.