Fourier Descriptors for Iris Recognition

Abstract

Automated identification of individuals is one of the most popular works today. Human verification, especially the iris pattern recognition is widely applied as a robust method for applications that demands high security. Many reasons are considered for choosing the iris in human verification, for example, the stability of iris biometric features throughout human life and unaffected of its patterns by human genes (genetic independency). This paper presents a new contribution in iris recognition biometric system. It uses Fourier descriptors method to extract the iris significant feature for iris signature representation. The biometric system is proposed and implemented using four comparative classifiers. It involves four sequential processes: the image enhancement process; iris feature extraction and patterns creation process; template construction process; and finally, the recognition and classification process. The mathematical morphology operations and canny edge detector are both applied for best outer/inner boundary localization procedure. The system satisfied 100% accuracy result regarding iris-pupil boundary localization for CASIA-v1and CASIA-v4 dataset. Also, when the identity of 30 persons were verified, the maximum matching result was %96.67 for Back propagation and lowest rate was %83.5 for Radial basic function.