ARTeM: a new system for human authentication using finger vein images

A new system (ARTeM) for human authentication using finger vein images is described here. The developed algorithm combines 1) a fuzzy contrast enhancement algorithm with 2) a mutual information and affine transformation based registration technique and 3) a correlation coefficient based template matching algorithm, to detect the identity of a person based on the match-scores with finger vein images stored in the database. For performance assessment of the ARTeM algorithm, the benchmark SDUMLA multimodal biometric database containing 3816 images of 106 persons is used. On the complete database, up to 95.28% classification accuracy is achieved with single finger images; while up to 98.11% accuracy is observed with a consensus of two fingers. On a reduced subset of 86 persons’ database, 98.84% accuracy is achieved with single finger classification and cent percent classification is obtained using a consensus of two fingers. Comparative analyses with other works also validate the effectiveness of the developed methodology.