Iris Localization And Feature Extraction Algorithm

With the increasing emphasis on security, automated personal identification based on biometrics has been receiving extensive attention over the past decade. Iris recognition, a kind of biological characteristic recognition, springs up as a result of the social and economical development, and it shows great advantages. Iris recognition technology with its great potential is relatively new among different kinds of biometric recognition technologies. A typical iris recognition system includes iris acquire, iris pre-processing, iris feature extraction and feature pattern match. This dissertation fucous on iris recognition algorithm research whose key steps are iris localization and iris feature extraction. The main works of this dissertation are as follows:1. Firstly, after analysising the existed iris localized algorithm, a new iris localized algorithm is proposed which is based on directional operator mask in this dissertation. The algorithm includes coarse localization and fine localization. The method is: define the center of coarse inner circle as the origin of the rectangular coordinate system, and divided the acquired iris image into several areas. In each area, the only corresponded directional mask is used to compute edge intensity so that the computed complexity is reduced. For outer circle localizing, the bigger masks are applied to compute the edge intensity because the gray-scale transition zone between iris and sclera usually is wider than the one between pupil and iris. It could enhance localization precision with increasing little computation. In the end, a recursive method is proposed for this algorithm to accelerate the localization speed further more.2. The wavelete packet transform is introduced and then the iris feature extract algorithm based on wavelete packet zero-crossings is put forward. The algorithm firstly applies wavelet packet decomposition to the normalized iris image, and then the sub-images which have good effect on irirs recognition are selected for encoding. In the end, the weighting hamming distance is applied to measure the similarity degree of two iris classes.3. In this dissertation, DLDA (Direct Linear Discriminant Analysis) is introduced and DLDA which combines with wavelet transform is proposed to extract the iris feature. To reduce the data dimension, firstly, we apply wavelet decomposition to the normalized iris image, and just choose the coefficients of the approximation part of the second level wavelet decomposition to represent the iris image. And then make use of DLDA to extract the iris feature from this approximation part so that can be easy to calculate.All the proposed algorithms are programmed with MATLAB7.0 and tested on the second version CASIA iris database. The results of the experiment show that the algorithms put forward can recognize people effectively.