Person Identification with Iris Recognition Based on Generalized Structure Tensor
S. Joshua Kumaresan, J. Raja, Paul Perinbam, D. Ebenezer · 2015
3 Abstract: An eye image has useless parts such as eyelid and eyelash apart from iris. These useless parts are masked and iris region alone is detected by Generalized structure Tensor (GST). Independent Component Analysis (ICA) is used to extract texture features in the iris. The best features (small subset) are then selected from the extracted texture features using Cuckoo Search algorithm which is an evolutionary algorithm. The best features of a person are then compared by calculating their hamming distance with a number of features of different individuals in the database for identification. The effectiveness is proved by comparing the proposed feature selection approach with other feature selection methods and without feature selection on a CASIA v3.0 database. The experimental results on this database show the proposed approach can achieve higher matching rates in iris authentication.