Research On Fingerprint Authentication Algorithm

Firstly, the development and application of Biometric Authentication is fully reviewed in this paper, as well as Fingerprint Authentication, one of the most mature techniques of Biometric Authentication. Based on the abroad researches, a system algorithm design is described here, which consists of Fingerprint Preprocess, Characteristic Extraction, and Finger Matching. Fingerprint Preprocess includes Fingerprint Foreground Extraction, Image Enhancement and Thinning. Fingerprint Foreground Extraction in this paper is important for the following algorithms, which is based on Edge Extraction Operator and Mathematical Morphology. As a result, foreground and background of fingerprint images can be distinguished, which reduce much noise from the background for the following fingerprint algorithms. Thus the accuracy is enhanced, the searching area is reduced, and the computing time is saved. Also the foreground is valuable information for Fingerprint Matching. Based on a Gabor wavelet image enhancement, an adaptive enhancement algorithm is proposed, which also considers the information of Singular Points and Foreground, and gets better performance. Because of the hexagonal grid, the connectivity paradox in the rectangular grid can be avoided. The templates here are simpler, and easier to implement on software and hardware. The algorithm is more robust to the noise on the edges, which are better than its counterpart on the rectangular grid. Characteristic Extraction here means Singular Point Extraction. The positions, types, and orientations of Singular Points are extracted from gray fingerprint images, in the step of Singular Point Extraction, by computing the Poincare Index and square orientation field. Algorithm in this paper owns more accuracy, computing speed and robustness, because of hierarchically designed. There are two different levels for Finger Matching--Coarse Matching and Precise Matching. At the coarse level, fingerprints are filtered by the structure information such as Singular Points, which saves much computing time. And then, at the precise level, fingerprints are matched by the minutia templates based on Hough Transform. Because of the more minutia information is considered, the complexity is reduced to O ( n 2) from O ( n 3).