An Efficient Supervised Approach for Retinal Person Identification using Zernike Moments

Abstract

Zernike moments map images using orthogonal basis functions. These moments have the advantages of rotation invariance, robustness and minimum information redundancy. In this paper, we focus on distinguishable pattern analysis of the retinal fundus images for person identification using Zernike moments. These moments are used to form 11-D feature vectors and k-nearest neighbor (kNN) classifier is used for person identification on publicly available DRIVE and STARE databases. This method outperforms all the existing methods with accuracy of 100% and 98.64% on DRIVE and STARE databases respectively. Its smaller dimension of feature vector, simplicity and robustness make this method suitable for real-time retinal person identification scheme.General Terms -Biometrics, security, pattern recognition.