Anisotropic Diffusion Filtering with A-priori Estimated Orientation Field for Enhancing Low-quality Fingerprint Images

We present a novel combination of classic anisotropic diffusion filters, a-priori estimated orientation fields and a locally adaptive contrast enhancement step for improving the quality of fingerprint images. By applying our new approach to low-quality images of the FVC2004 fingerprint databases, we are able to show its equivalency to state-of-the-art enhancement methods for fingerprints. Linear anisotropic diffusion and three types of nonlinear anisotropic diffusion are examined: coherence-enhancing, incoherenceenhancing and edge-enhancing diffusion. The performance of diffusion filtering is compared to existing methods in verification tests according to the FVC protocol using the freely available NIST matcher BOZORTH3. In addition, combining anisotropic diffusion filtering with curved Gabor filters leads to considerable improvements in terms of lower equal error rates. The achieved performance improvements and the computational efficiency of the method suggest to include anisotropic diffusion filtering as a standard image enhancement add-on module for a future real-time fingerprint recognition system. In order to facilitate the reproduction of these results by other researchers, a Matlab implementation of the anisotropic diffusion filters is made available for download. Carsten Gottschlich is with the University of Göttingen, Institute for Mathematical Stochastics, Goldschmidtstr. 7, 37077 Göttingen, Germany. Phone: +49-(0)551-39172100. Fax: +49-(0)551-3913505. Email: [email protected] Carola-Bibiane Schönlieb is with the University of Cambridge, Department of Applied Mathematics and Theoretical Physics, Wilberforce Road, CB3 0WA Cambridge, UK. Phone: +44-(0)1223-764251 Email: [email protected]