Fingerprint orientation alignment and similarity measurement

Abstract

In this paper, an algorithm for fingerprint orientation alignment is proposed in which a fingerprint is aligned to a particular direction according to its ridge flows. The algorithm works equally well for different types of fingerprints, including the type with no reference points. Aligned fingerprint images provide a lot of useful information for various applications. To demonstrate the use of such information, a fingerprint similarity measurement system is developed in which fingerprints are expressed by vectors with different elements derived from aligned images. When given two fingerprints, this system can help determine which of the two is more similar to a third fingerprint and which one is less similar by calculating distances among vectors. The alignment algorithm has been tested on 165 images in the NIST-4 fingerprint database. By rotating a test image in different degrees, 20 images are created. The alignment precision is calculated by an equivalent Euclidean norm, E(I), of the 20 rotated images to the original image. A smaller E(I) indicates a higher accuracy of the algorithm. The experimental results show that the average E(I) of these 165 samples is 1.234, and the spread of these 165 E(I)s is 1.1.

Keywords:

• FINGERPRINT
• GABOR FILTER
• SINGULAR POINT
• FAULT LINE
• SIMILARITY