Research On Application Of Wavelet Transform In Biometrics Recognition Technology

As the requirements of system and information safety increases in this informational society, more and more occasions need to verify personal identities. However, the traditional verification methods can not satisfy the requirements to the development in society due to its intrinsic weakness. Biometric recognition is a kind of technology that verifies personal identities through the biological intrinsic feature of human beings. Since the biological feature can not be forgotten or lost and can not be fabricated or simulated, it can overcome the weak points of conventional methods to identity verification effectively.Wavelet transform is an excellent image processing tool that has been widely used in biometrics. Most of the existing algorithms are on the basis of real district wavelet transform which suffers shift-variation and less direction selection ability resulting that the algorithms based on Discrete Wavelet Transform (DWT) are instable and incomplete which leads to the poor recognition rate. Moreover, DWT generates real number results, thus it can not contain phase information that is very important in image processing.Dual-tree Complex Wavelet Transform (DT-CWT) has been introduced into the filed of biometrics recognition to solve the problems of the real DWT due to its good properties of shift-variance and improved directionality. Two algorithms based on the properties of the biometrics images and DT-CWT is proposed in this paper to perform the recognition task. The first algorithm is based on the amplitude of DT-CWT. It sums the amplitude of points in each block of every detailed image transformed through DT-CWT to get part and the whole feature and employs weighted Euclidean distance to match. Compared experiments show that this algorithm is robust and better in feature extraction than that of the real wavelet transform based method. The second algorithm is based on the phase feature of DT-CWT. It utilizes the phase information from DT-CWT to code the detailed image and add the directional information in the code and takes advantage of the improved Hamming distance to match. Results of experiments show the effectiveness of this approach. In the dissertation, it also analyzes the reasons of different results when the DT-CWT based algorithm is used on different biometric recognitions and describes the areas this algorithm fits for.