Abstract
Biometric techniques used for personal identification have recently raised an increasing interest. Among many biometrics, features of fingerprint images are most frequently used. In this thesis, we develop a fingerprint identification system. Our system first takes a 300 ’ 300 fingerprint image, scanned via a Veridicom FPS110 reader, as input, does image enhancement, segmentation, and thinning to get a thinned image. From which, we find a minutiae pattern consisting of endings and bifurcations with their corresponding flow directions to best match those from a database.Our database contains 308 fingerprint images contributed by 77 persons, each contributes 4 times of his right index fingerprints in a few minutes. Experiments on a Pentium II 400 PC with 128MB SDRAM show that the implemented system spends 6.78 seconds for processing and matching a fingerprint image, and it attains about 88% recognition rate with 3.25% false acceptance rate.