A div-curl regularization model for fingerprint orientation extraction

Extracting reliable fingerprint orientation over low quality fingerprints is still a challenging problem. In this paper, we analyze the distribution of divergence and curl of fingerprint orientation vector field. Then a div-curl regularization model is proposed to minimize the combination of weighted divergence, curl and data fidelity term. This model yields a new nonlinear partial differential equation (PDE) for orientation regularization. The equation has an adaptive diffusivity depending on the position relative to the reference point, providing orientation vector fields which have desired distribution of divergence and curl. The proposed approach has been evaluated on a web-based automated evaluation system FVC-onGoing, and it ranks the first in published six algorithms.