The Research Of Iris Recognition Based On Multiwavelet Transform

In modern times, going with the development of science, technology and economy, people think much of identity recognition more and more. But traditional identity recognition technologies such as password etc. have not suit the identity recognition in modern times for their bad ability against forging. For overcoming the disadvantages of traditional identity recognition technologies, biometrics recognition methods were brought forward such as fingerprint, iris and voice. Among these biometrics recognition methods, the research on iris recognition has received increasing attention because of its uniqueness, stableness and difficulty of counterfeiting.Iris image preprocessing and feature extraction play a key role and are the difficulties in iris recognition. Wavelet analysis theory is a so excellent time-frequency analysis tools that many researchers have applied it to iris recognition technology and achieved certain results. In recent years, multiwavelet theory as the development of scalar wavelet technology, has begun to been applied into various fields gradually. It does not only keep the good localization characteristics in time domain and frequency domain of scalar wavelet, but also overcomes the shortcomings of scalar wavelet. It owns a lot of good properties at the mean time, such as the symmetry, compactly support, orthogonality and vanishing moments or higher approximation order , as a result, the multivavelet has a good performance in the expression and extracting of iris texture features.This article, to overcome the disadvantages of iris recognition algorithm in existence, makes a thorough research on image preprocessing and feature extraction of iris recognition system. And the main research works are as follows:1. Based on the detailed analysis of the process of iris image preprocessing, it proposes an improved iris location method, using of morphological algorithms to locate the internal and external edge of the iris, which avoids blind search. Experiment proves this method reduces the computing of Hough transform and improves its reliability.2. It introduces the theory of multiwavelet into iris image feature extraction, proposes the method of directly coding and cosine distance matching based on multiwavelet sub- cingulum and the coding method of two-dimensional iris texture feature measure based on multiwavelet multi-scale information. At last, through experimental data, it compares the method of GHM multiwavelet cosine distance with the method of detection based on SVM, Super sausage neural network detection and wavelet zero-crossing detection on recognition capability and evaluates all of them.