Indexing fingerprints using minutiae quadruplets

The computational complexity of matching an input fingerprint against every entry in a large-scale fingerprint database can be prohibitive. In fingerprint indexing, a small set of candidate fingerprints is selected from the database and only images in this set are compared against the input probe fingerprint thereby avoiding an exhaustive matching process. In this paper, a new structure named “minutiae quadruplet” is proposed for indexing fingerprints and is used in combination with a clustering technique to filter a fingerprint database. The proposed indexing algorithm is evaluated on all datasets in the Fingerprint Verification Competition (FVC) 2000, 2002 and 2004 databases. The high hit rates achieved at low penetration rates suggest that the proposed algorithm is beneficial for indexing. Indeed, it was observed that for 50% of the fingerprints, in most of the datasets, the penetration rate was less than 5.5% at a 100% hit rate. The robust performance across different databases suggests that the indexing algorithm can be adapted for use in large-scale databases.