一种基于扩散方程的指纹方向场提取方法

A fingerprint orientation extraction method based on diffusion equation

方向场估计是指纹识别过程中非常重要的步骤。传统方向场提取算法存在着3个较为严重的问题:指纹质量较低时,无法获得满意的结果;无法在背景区域获得可靠的方向场;基于模型的方法,如Legendre模型或FOMFE模型,在奇异区域会产生较大偏差。针对这些问题,该文改进了基于复数滤波的质量评价方向,并提出了一种融入指纹质量的扩散方程来提取指纹方向场。为了验证算法的有效性,在各种质量的指纹图像上进行了实验。实验结果表明:无论是低质量还是高质量指纹图像,基于扩散方程模型的方向场提取算法都能获得准确的指纹方向场。

Fingerprint orientation field estimation is a very important step in the process of fingerprint identification. Traditional algorithms have three serious drawbacks that when the quality of the fingerprints is bad, the performance is unsatisfied, that good performance can not be obtained in the background area, and that the model-based methods, such as Legendre model and FOMFE model, can cause serious error in singular areas. The complex filter based fingerprint quality estimation method was modified, with a diffusion equation based method then developed for low quality fingerprints orientation extraction. Experiments were made to evaluate the performance on various quality fingerprints to show that all the fingerprints benefit from the developed method.