Iris Recognition Based On2D Log Gabor And Research Of Feature Information Protection

In recent years, with the rapid development of technology and the increasing desireddemand of security, biometric recognition technology and biometric template protection arepaid more attention than before. Iris recognition is regard one of the latest biometricrecognition technology, it gains great advantages. Iris recognition system includes: iris imageobtaining, iris localization, iris normalization, enhancement, feature extraction, iris matching.Feature extraction is the most important part of iris recognition, this paper pays hugeattention on feature extraction methods and a method based on2D Log Gabor, the methodgives a better performance than others.2D Log Gabor features are used to feature templateprotection, under the experiments and analysis of Compress Sensing attack, the method whichbased on Compress Sensing give a good performance, lower EER.The contribution of this paper:1. Analyzing the algorithms of iris localization, Canny operator and Hough transform areto localize the inner circle. Several art-of-out algorithms pointed out that the features whichare closer to pupil are much rich than others. We spread32pixels from the inner circle asouter circle. The method is simple easily complete and the complexity is small.2. Analyzing2D Gabor filter,1D Log filter and2D Log Gabor filter which are proposedto extract iris feature.3. A brief instruction of sparse represent includes sparse represent, sparse recovery andbrute attack with sparse recovery，proposing a improving method based on OMP of sparserecovery. The experiment results show that the EER of2D Log Gabor is1.03%, the bruteattack and sparse recovery only get the CCR of0.44643%. OMP is60times faster than BP,The CCR of the method base on improvement of OMP is99.1071%.All the algorithms are programmed with Matlab7.11b and tested on IITD Version1.0database. The results show that2D Log Gabor is the best iris feature.