基于奇异点的字典模型指纹方向场去噪方法*

方向场估计是指纹识别过程中非常重要的步骤。传统方法如基于梯度的方法等在处理潜指纹图像时很容易受噪音干扰,而最近提出的基于字典模型的方法无法解决“真词错误”的问题。针对上述问题,本文提出一种融合了零极点模型的字典模型的指纹方向场去噪方法,即将指纹方向场看做是零极点控制的方向场和平滑的残差方向场相叠加的结果,通过首先用零极点模型生成正确的零极点控制的方向场,然后用字典模型修正残差方向场方向场,最后将零极点模型生成的方向场与去噪后的残差方向场融合形成重建方向场,通过基于奇异点的字典模型,我们解决了“真词错误”的问题。为了验证算法的有效性,在NIST SD27潜指纹图像数据库上进行了实验。实验结果表明：对于潜指纹,本文算法能获得比字典模型更精确的方向场,继而可以更好地增强潜指纹图像,并在后续的匹配实验中取得更好的结果。

A Singular Point Based Dictionary Model for Fingerprint Orientation Field Estimation

Fingerprint orientation field estimation is a very important step in fingerprint identification. While classical methods such as the gradient based can not well cope with latent fingerprint images, a recently proposed dictionary based method has some difficulty around the singular points in fingerprints. To further improve its performance, this paper incorporates the zero-pole model into the dictionary model. We consider the fingerprint field orientation is consist of point-charge orientation and residual orientation. Then we use dictionary model to denoise the residual orientation and use the zero-pole model to generate correct point-charge orientation. At last, we reconstruct the latent fingerprint field orientation by adding back the influence of singular points to the denoised residual orientation. By adopting this method, we get better fingerprint orientation and solve the "real-word" problem. Evaluation experiments have been done on the NIST SD27 database. The experiment results show that the algorithm we proposed in this paper can extract more accurate orientation field from the latent fingerprint image than the dictionary model, and help to achieve higher latent fingerprint recognition accuracy.