Personal Authentication with an Iris Image Captured Under Visible-Light Condition

In recent years, security has been improved in various scenes by biometric authentication, and iris authentication is a highly accurate system. However, in the case of the iris image capturing under visible-light environment, there is a problem that the corneal reflection of surrounding background covers over the iris pattern. For this reason, iris authentication requires a near-infrared lighting and a near-infrared camera. However, it is difficult to use the iris authentication system installed on a smartphone on a daily basis. Therefore, in this study, we develop an iris authentication method under visible-light environment using a normal RGB camera. In this paper, several preprocessing were applied to images captured in an environment with relatively little corneal reflection. We show the results of comparing accuracy of characteristics of local luminance variation with accuracy of gradient direction feature of iris pattern. From the experimental results, EER (Equal Error Rate) of the gradient direction feature which is an index of performance evaluation in biometric authentication was 5.03%. In addition, we confirmed that the performance of iris authentication decreases when the capturing condition changes.