| In recent years, automatic fingerprint based personal authentication can besuccessfully applied in a myriad of applications. Such as the low cost door openerand Smart phones in everyday life etc. However, this technology applied inlarge-scale or networked environment also faces many challenges.One common ground of all this fingerprint based personal authentication systemsis the necessity for estimation of fingerprint ridge orientation. The importance of ridgeorientation can be deflected from the fact that it is inevitably used for detecting,describing and matching fingerprint features such as minutiae and singular points.Using the ridge orientation, not only the error rates can be improved, but also moreefficient image compression and a speed up in database queries can be achieved.In this thesis we propose a novel method for fingerprint ridge orientationmodeling using wavelet translation. Our proposed model does not require priorknowledge of singular points. and can be fitted in different wavelet scales.Solve theproblem about smoothing orientation data while preserving details in high curvatureareas, especially around singular points. Another characteristic of the model isCombines the advantages of parametric and non-parametric model of the fingerprintorientation. Add the regularization term to the optimization algorithm, forming aWavelet based smooth extension method on the orientation field reconstruction.Starting from the view point of flexible templates models, we develop a methodwhich constraints the incomplete fingerprint orientation to reconstruct only in ways asthey occur in nature.Extensive experiments using a fingerprint recognition competition databases forcomparative experiment. We can report statistically significant improvements in theaccuracy of fingerprint reconstruction orientation field. Our method show thatsingular points, which result in a discontinuous orientation field, can be reconstructed better by the wavelet translation. |
PDF Full Text Request