| In this work, a new Automatic Fingerprint Identification method as a biometric technique for personal identification is proposed. A review on the fingerprint identification problem is presented. It is shown that similarly to the local “minutiae” based fingerprint representation, a simpler global ridge-based fingerprint representation may be successfully used for fingerprint identification and verification, if an appropriate global metric of similarity between fingerprints, can be found. A new global representation of fingerprints suitable for fingerprint recognition is obtained by computation of moiré patterns from binary fingerprints. A simple straightforward algorithm for automatic binarization of gray-scale fingerprint images is developed. It is based on a Laplacian edge detector kernel, a Gaussian pre-filter, and a median filter, and do not resort to the use of brightness histograms. Further, a new method of fingerprint recognition based on the computation of moment invariants of fingerprint moiré patterns is proposed. In a preliminary stage, the method is tested with an available database of 12 gray-scale fingerprint images which produce a total of 132 entries. At this stage, a satisfactory performance up to 86.4% of correct identification rate vs. a 12.1% of false acceptance rate and 1.5% of false rejection rate is obtained. |
