Skip to Main Content
A hybrid fingerprint matching framework is proposed in this paper that handles fingerprint distortions and has less computational complexity. For this purpose a dual part algorithm is proposed that processes the global matched fingerprints and local matched fingerprints individually. Two polygons are generated from the outer boundaries of the Delaunay triangulation of the fingerprints. Then, they are compared to determine whether the fingerprints are globally matched or not. If they are globally matched, the central Voronoi cells of fingerprints are compared and after that, the triangles and minutiae corresponding to the matched cells are analysed. If the fingerprints are not globally matched, they are compared locally. This part of the algorithm starts with random triangles from the input and the template fingerprints. If they are matched, the cells and minutiae corresponding to the matched triangles are compared. The algorithm continues processing till the number of matched triangles reaches a predetermined number, so that the fingerprints are reported as matched. The thresholds are dynamically selected according to the characteristics of the fingerprints. On the other hand, the features proposed in this manuscript are invariant to translation and rotation. The algorithm is evaluated on three different databases and the results indicate better performance than the previous methods.