Feature Extraction And Recognition Of Iris Based On Multiwavelets

Multiwavelets constitute a new chapter which has been added to wavelet theory . Since 1996 , multiwavelets has become a very active field of wavelet research because it can be simultaneously orthogonal, symmetric, short of supported and high vanishing moments, while traditional wavelet can't.In the beginning of this thesis, we introduce the principle and properties of multiwavelets . On this basis, we expatiate the multiwavelets theory proposed by Charles A, Micchelli and Yuesheng Xu. We discuss the derivation of fractal multi-wavelets filter with the interval characteristic of self-affine.According to the ordinary derivation method of the filter, we construct a group of biorthogonal and symmetric multiwavelets filter. This filter has the shortest length ofsupported, in the course of formation of transform filter matrix H, there is no overlap ofsupported.Meanwhile, it can be reconstructed after wavelets decompose without boundary distortion, so we avoid boundary continuation. In the image processing, the multiwavelets is a perfect feature extraction, and it can describe local feature information of the image.Iris possesses abundant texture features, which is protected by cornea and stable over time once form .Then iris is stable, reliable, unique and noninvasive .Iris identification is one of the most representative biometrics technology,Which also shows huge potential of commerce . Every parts, principle and the preprocessing of the image of iris recognition system is introduced in this thesis. Two methods of extracting of iris texture feature, encoding and matching methods are mainly studied.The first, 2D-Gabor is applied to extract iris texture feature after the characteristic is discussed . Hamming distance is designed the classifier.The second, Biorthogonal multiwavelets filers on the interval characteristic of self-affine is proposed to extract iris texture feature and local with global feature is applied to match iris .After using biorthogonal multiwavelets transform to process iris images, local adaptive coarse quantization encoding is adopted in the low frequency parts of coefficients. Hamming distance is designed the classifier. When the Hamming distance is uncertain of its desicsion for effect of eyelids,eyelashes and iris deformations, mean and variance is extracted from coefficients of multiwavelets transform and euclidean distance of covariance reciprocal with weight value is designed the classifier .The results show that this approach has good results of feature extraction and fast speed of operation and high credibility to identify...