| As one of the most friendly biometric recognition technology, face recognition technology is very widely used in security monitoring, user authentication, man-machine interaction and so on. It has received the widespread attention of most scientists and produced many excellent face recognition algorithms.Wavelet transform with strong time-frequency localization analysis ability and multi-resolution characteristic has been widely used in face recognition, but wavelet basis is isotropic, so wavelet transform can only reflect the point singularity of one-dimensional signal and unable express the edge character of two-dimensional image, it cannot provide an optimally sparse representation of face image included more facial contour and five sense organs’curve information. Curvelet transform put forward by Candes and Donoho largely make up for the defects of wavelet transform. Curvelet transform has not only the multi-scale characteristic of wavelet transform, but also very strong directivity and highly anisotropic, can well represent singular characteristic of line and curve. Compared with wavelet transform, it can more sparsely express the image and make signal energy more concentrated.This paper mainly studies the basic theory and implementation of Curvelet transform, and analyzes the Curvelet coefficients resulting from the decomposition, then study the theory and implementation of principal component analysis, included traditional principal component analysis, two-dimensional principal component analysis, two-directional two-dimensional principal component analysis. Finally, a face recognition algorithm based on second-generation Curvelet transform and two-directional two-dimensional principal component analysis is proposed and realized. We perform the experiment with ORL and YALE face database, the results show that this algorithm has higher recognition rate and shorter recognition time compare to the face recognition algorithm based on traditional wavelet transform. |
PDF Full Text Request