Infrared finger biometric using scale invariant feature transform correspondences

Abstract

In this work, we propose a finger vein recognition system using SIFT (Scale Invariant Feature Transform) descriptors. Finger vein recognition has recently attracted considerable attention as a biometric technique for personal identification because of its convenience and high security. We extract SIFT descriptors from illuminationnormalized images and measure the similarity using a number of the matching descriptors. In this method, we eliminate the step of localizing and resizing the finger area, and we directly normalize the captured images to extract SIFT descriptors from them. In the matching step, we use the properties of SIFT descriptors effectively so that we match both the components and the locations of the descriptors. Experimental result showed that the matching performance of 0% equal error rate. Consequently, the SIFT matching recognition is suitable for 1:N matching identification system under the assumption that the processing speed is improved. © 2015 American Scientific Publishers. All rights reserved.