An application of scale-invariant feature transform in iris recognition

Abstract

Scale-invariant Feature Transform (SIFT) is an algorithm to find local features in images. SIFT uses Difference-of-Gaussian (DoG) to locate candidate keypoints and performs a detailed fit to locate keypoints, then orientations are added to keypoints and keypoint descriptor is generated for each keypoint. Iris recognition is one of the most reliable biometric authentications. In this paper, we propose a reliable method of iris recognition by applying SIFT. It includes segmentation, matching and evaluation. Other than the conventional method, Normalizing and encoding are removed since SIFT is rotation-invariant and scale-invariant. Our proposed method is tested on CASIA and self-obtained images. Experiments show the proposed method is fast and accurate.