Fast and robust iris recognition

Abstract

Iris recognition is a reliable biometric identification technology. In this study, the one-to-many identification and one-to-one verification using iris recognition are both examined. Our method consists of three phases: pupil and iris location, iris feature image extraction and iris matching. The CASIA iris database was used in this study to evaluate the performance of the proposed method. The database includes 108 iris classes, each including seven images. The identification accuracy of our method is 99.646%; while the proposed method can achieve an equal error rate of 1.085%. On the other hand, feature extraction of an iris image in our method is only 16.78 ms. Each verification test of an iris images takes 1.4 ms and each identification test over 432 database images takes 151.2 ms. Note the experimental accuracies were reported without manual rejection of poor-quality images or invalid iris-location images, thus the efficiency and robustness of the proposed method can be proven.