Individuals Authentication from Both-Eye Images Using Feature Level Fusion and Score Level Fusion

Abstract

In this paper, an efficient method that allows us to authenticate individuals from both-eye images using feature level fusion and score level fusion is presented. The proposed method consists of three main steps: In the first one, the iris images are segmented in order to extract the iris disc. The segmented images are normalized by Daugman rubber sheet model. In the second step, the normalized images are analyzed by a bench of two 1D Log-Gabor filters to extract the texture characteristics. The encoding is realized with a phase of quantization developed by J. Daugman to generate the binary iris templates. In the third step, feature level fusion is applied by concatenation of both binary irises templates and score level fusion is utilized using Dempster Shafer rule. For the authentication and the similarity measurement between both binary irises templates, the hamming distances are used with a previously calculated threshold. The proposal method using two fusion techniques has been tested on a subset of iris database CASIA-IrisV3-Interval. The proposal method using score level fusion based on Dempster Shafer rule shows better performance with accuracy of 99.97%, FPR of 0% FNR of 4.49%, EER of 1.4% and processing time of 15.39s for one iris image.