da/sec scientific talk on Biometrics
Topic: SIFT based Iris Recognition
Keywords — Biometrics, SIFT, Iris Recognition
SIFT (Scale-Invariant-Feature-Transform), introduced by David G. Lowe, is an algorithm for the detection and description of local features in images. Through it’s invariance to image scale, rotation, illumination and viewpoint, SIFT is considered a robust algorithm in object tracking and other pattern recognition applications like panorama stitching or robot localization. Since it’s introduction in 1999, it was also applied on several biometric characteristics. For recognizing iris images, researchers report unpractical biometric performance paired with a good deal worse comparison time. In recent years, researchers put in a lot of effort to speed-up the comparison time of SIFT descriptors by binarization or involving the GPU.
This talk introduces a new binarization of SIFT keypoints and feature descriptors. By the use of lookup-tables, this implementation achives a competitive biometric performance providing comparison times comparable to Hamming-distance based systems. Experiments are carried out on three different publicly available datasets, namely CASIAv1, CASIAv4 and Biosecure.