An accurate fingerprint orientation modeling algorithm

• The paper estimates fingerprint orientation field (OF) by regularization and weighted OF modeling.
• The regularization preserves OF in uniform flow areas while reconstruct OF in bad quality areas.
• To reconstruct OF accurately near singular points areas, weighted OF modeling is proposed.
• The proposed algorithm does not require prior knowledge of singular points.
• It performs better than the existing state of the art algorithms.

Orientation field (OF) plays a crucial role in singular point detection, fingerprint image enhancement, fingerprint matching, fingerprint indexing, fingerprint synthesis, fingerprint registration and fingerprint classification. Thus, an accurate OF estimation algorithm is proposed in this paper. It presents a novel OF regularization algorithm based on variational approach, which aims to preserve the genuine OF in the areas containing uniform flow and to accurately reconstruct the OF in bad quality areas. Estimated OF may be spurious in singular points close-by areas thus, it is further refined by weighted OF modeling based on Fourier basis. Experimental results are conducted on a publicly available database, FVC2004DB1A in terms of singular point detection and deviation in OF. It demonstrated that the proposed algorithm exhibit better performance that the state of the art algorithms.