A novel coding scheme for indexing fingerprint patterns

Abstract

Indexing is the process of assigning a numerical value to a database entry in order to facilitate its rapid retrieval. Indexing a fingerprint database can reduce the search space and improve the response time of an identification system. We discuss a novel method for generating index codes for fingerprint images by using a small set of pre-determined reference fingerprints. In the proposed method, the match scores generated by comparing an input fingerprint with the reference fingerprints are subjected to a discretization function, which converts them into an index code. A search mechanism based on the Hamming distance identifies those index codes in the database that are similar to the code of the input image. The proposed technique has several advantages: it obviates the need to extract complex features from the fingerprint image; it utilizes the matcher that is already associated with a particular application; and it can be used to index any biometric database irrespective of the trait or matcher being used. Experimental results on two fingerprint databases (NIST-4 and WVU) indicate that the proposed encoding scheme generates index codes that are well-scattered thereby allowing noisy query images to be indexed correctly. © 2008 Springer Berlin Heidelberg.