Wavelet Based Rotation Invariant Fingerprint Recognition

Fingerprint recognition is a popular method for person identification over 100 years back which has been now very widely used by every forensics and law enforcement agency. Automatic fingerprint identification system (AFIS) is capable of performing human identification very efficiently. One of most commonly used approach for fingerprint matching is based on minutiae matching. Minutiae based recognition needs time-consuming pre-processing operations for finding minutiae points. This paper proposes image based rotation invariant algorithm for fingerprint feature extraction. A fingerprint recognition approach based on statistical features extracted from discrete image using wavelet has been presented here. Fingerprint recognition can be done efficiently using texture classification approach. We propose an effective combination of features for multi-scale and multi-directional recognition of fingerprints. The features include standard deviation, mean of wavelet decomposed image. We have used distance vector formula for similarity comparison between the texture classes. Database used for evaluation is of Biolab, University of Bologna. Fingerprint verification is done with a total of 2592 fingerprint images obtained by rotating each of the fingerprints of sixty subjects from 0° to 360° in steps of 10° each Experiments are carried out on FVC 2000 and FVC 2004 databases. Proposed work provided success rate up to 96.87% on standard database and can recognize one fingerprint in average time of 0.7ms.