Iris feature extraction using optimized Gabor wavelet based on multi objective genetic algorithm
Hamed Ghodrati, Mohammad Javad Dehghani, Habibolah Danyali · 2011
Iris reputes for its potential to identify the people with high accuracy in large scale. This is not achieved unless the iris patterns are well represented. Gabor filtering is vastly used in iris recognition literature for feature extraction. Conventionally, Gabor parameters value are supplied by pre-knowledgeable values so that the filter bank size is increased to prevent the losing information. In this paper, multi objective genetic algorithm (MOGA) is used to optimize the Gabor-wavelet in order to reduce the filter requirements and increasing the accuracy. The feature vectors are encoded by phase quantization and a novel method based on iris texture variation. Experimental results show recognizing with CRR=99.68% and EER=0.26% for codes with length only 496 bits on a subset including 2125 iris images from CASIA-IrisV3-Interval database.