Developing Iris Recognition System Based on Enhanced Normalization

Iris texture can be considered as a physical password since it has unique features for every person. Iris recognition is an important and reliable biometric system for access control. Therefore, it is one of the preferred and distinctive biometric methods for identification purposes. In this paper, the performance of the iris recognition is enhanced by developing the normalization process. It is based on daugman's algorithm with some changes by dividing the iris into two halves and both are processed simultaneously. The project is implemented in MATLAB 2018 using CASIA iris-lamp V4 and UBIRIS V2 databases. The proposed normalization algorithm is compared with Daugman's rubber sheet model and the results demonstrate that the proposed algorithm is three times faster than the traditional one. In addition, the results also demonstrate that UBIRIS database has higher recognition accuracy of 98.7% compared to CASIA database which is 98.5%. For other steps, Circular Hough transform is used in the segmentation process, Ridge energy direction and hamming distance are used for feature extraction and matching.