Extraction of high confidence minutiae points from fingerprint images

Fingerprint is one of the well accepted biometric traits for building automatic human recognition system. These systems mainly involves matching of two fingerprints based on their minutiae points. Therefore, extraction of reliable or true minutiae from the fingerprint image is critically important. This paper proposed an novel algorithm for automatic extraction of highly reliable minutiae points from a fingerprint image. It utilizes frequency domain image enhancement and heuristics. Experiments have been conducted on two databases. FVC2002 which is a publicly available fingerprint database containing 800 fingerprint images of 100 persons and IITK-Sel500FP which is an in-house database containing 1000 fingerprint images of 500 persons. Minutiae points extracted by the proposed algorithm are benchmarked against the manually marked ones and the comparison is done with an open source minutiae extractor mindtct of NIST. Experimental results have shown that the proposed algorithm has significantly higher confidence of extracted minutiae points being true.