A normalized reference image technique for improving iris recognition systems

Iris recognition is one of the most reliable biometric technologies. In order to improve the accuracy of iris recognition systems, and according to the variations in iris data, due to noises of the eyelids and eyelashes or inappropriate image acquisition environment, multiple iris images per person are enrolled. Therefore, these systems suffer from storage and computational overheads. This paper presents a new technique for generating a reliable reference iris image per person rather than maintaining multiple images. The proposed technique enhances the performance of iris recognition systems regardless of the feature extraction method in use. Moreover, it has the advantage of reducing the amount of data storage, computational complexity and the matching time. The achieved results assure that our proposed approach results in very high recognition rate as well as it decreases the matching threshold.