Vectorized fingerprint representation using Minutiae Relation Code

Abstract

Minutiae-based vector representation algorithms have been proposed, which allow us not only to speed up matching tasks, but also to easily apply for various template protection techniques, such as Fuzzy Vault, Fuzzy Commitment, and BioHashing. In this paper, we propose a new vectorized fingerprint descriptor called Minutiae Relation Code(MRC), which consists of a set of vector-represented minutiae relation information between arbitrary minutiae. We also evaluate authentication performances using FVC2002 Database(DB1, DB2, DB3, DB4) and we show 0.82% Equal Error Rate(EER) in DB1, 0.82% in DB2, 2.71% in DB3, and 1.49% in DB4.