Fingerprints verification based on their spectrum

The goal of this paper is proposing a novel fingerprint verification approach to increase the accuracy and robustness of fingerprint verification process. The proposed approach is based on spectrum features extraction from the fingerprint image. In the proposed approach, the two dimensions (2-D) fingerprint image after enhancement is transformed into one dimension (1-D) using lexicographic ordering by dividing the image into non overlapped square blocks, each block scanned column by column. The scanned block pixel values are combined into a 1-D signal, and then the spectrum features are extracted from this signal. Support Vector Machines (SVMs) have been used for matching the features to decide whether these features belong to the same person or not. Furthermore, minutia based fingerprint verification approach is carried out in five steps; enhancement, binarization, thinning, minutiae extraction and minutiae matching. The proposed approach is compared with the minutia based approach and with other published approaches. The international fingerprint verification competitions (FVCs) databases have been used for the evaluation. The results proved the superiority of the proposed approach to the minutia based approach.