Accurate Computation of Fingerprint Intrinsic Images with PDE-Based Regularization

摘要

The intrinsic images offingerprint, such as orientation field and frequency map, represent the particular and basic characteristics of fingerprint ridge/valley patterns, and play a key role in feature extraction and matching of the fingerprint recognition system. In this paper, a novel algorithm is presented for accurate computation of fingerprint intrinsic images with PDE-based regularization technique. First, the coarse orientation field is estimated using general gradient-based method, and the frequency map is estimated by Fourier spectrum analysis of the projected curve of local window. Then, to measure the reliability of the intrinsic images, the quality map is computed based on the gray-scale intensity and local structural information. Finally, accurate orientation field and frequency map are reconstructed with PDE-based regularization of nonlinear diffusion filtering which is controlled by the quality-based diffusivity. Experimental results illustrated the efficiency of the proposed approach.