A Face Recognition Method Based On Wavelet Transformation And Principal Component Analysis And Its Realization

With the rapid development of the intelligent information technology in nowadays society, there are increasing demands on fast and efficient automatic identity recognitions by most social fields. As an intrinsic property of human, face will almost never be forgotten, lost or stolen, compared with other biometric features. Meanwhile, the characteristics of the face image acquisition were non-contact, non-invasive and non-compulsory, which attracts more and more people’s attention. The aim of this research is to design and develop a face recognition system based on wavelet transformation and principal component analysis (PCA) on matrix laboratory (MATLAB) platform. The main works of this research are as follows:Firstly, it is introduced the research background, application, development, present situation and main technologies of facial recognition, and briefly illustrated the face detection and location technology. Secondly, there are the detailed introductions on the pre-processing methods of face recognition, in consideration of the influences of the size, angel, resolution of the face images. Thirdly, there is an in-depth study on the theory of wavelet transformation, with the introduce of commonly used wavelet function, the discussion of how to extract the relatively stable subband and achieve the effect of the image vector dimension reduction at the same time, and the realization of the simulation experiment. Then, the principal component analysis (PCA) is fully studied; the principle of Karhunen-Loeve transform (K-L) and singular value decomposition (SVD), and some methods of image distance measurement are introduced in details; the experiment simulation of PCA face recognition algorithm is also performed on MATLAB. Finally, our algorithm simulation is realized on Matlab platform; the key influence factors on the recognition rate are mainly focused and analyzed, including the selection of wavelet base, wavelet decomposition levels and wavelet decomposition sub-band in wavelet transformation, the choice of the threshold e in PCA method, and the number of faces in PCA face recognition algorithm; the simulation experiment on the low pixel and high pixel images was carried out. In conclusion, compared with the traditional PCA method, this proposed algorithm, after considering the wavelet transformation, greatly reduced the computation complexity, significantly improved the recognition rate, and effectively decreased the unwanted noise emission level; its extracted features might more accurately and clearly reflect the difference between faces.