The Use of Spatial Distribution of the Local Histogram Based Features for Finger's Veins Biometrics

Abstract

Finger vein authentication is a new biometric technique utilizing the vein patterns inside of fingers for personal identity verification. Vein patterns are different for each finger belong to each person; and as they are hidden underneath the skin’s surface, this makes finger vein detection a secure biometric for individual identification. The vein grid images are acquired using infrared (IR) cameras. The acquired images are of low contrast and blurred in nature; so, an effective contrast enhancement step is required to expand the values of brightness range in the input vein image. The deal with low quality finger vein image represents the major concern of this work, beside to the selection of proper features to efficiently distinguish between individuals. In this paper a feature vector of the local histogram moments of gray finger image is proposed to represent the veins attributes; the main reason for used local moments is their ability to reflect the statistical behavior of veins variation at each part of finger image.  The extracted features are assembled as a feature vector; which, in turn, is used to distinguish different individuals. Nearest Neighbor classifier are used to make recognition decisions in the matching stage. The system is tested using a database consisting of 3,816 images. This dataset was constructed by capturing 6 samples for each of the 3fingers (i.e., index, middle and ring) that belong to one of the 2 hands of the 106 subjects. The achieved identification results of the proposed system indicate high recognition performance which is 99.52%, while the verification test results indicate error rate 0.003%.