一种扭曲指纹图像的细节点修正方法  

A Method to Adjust Minutiae Location and Direction in Nonlinear Distorted Fingerprint Image

作　　者：陈晖[1] 殷建平[1] 祝恩[1] 

机构地区：[1]国防科学技术大学计算机学院,长沙410073

出　　处：《计算机研究与发展》2010年第12期2141-2148,共8页Journal of Computer Research and Development

基　　金：国家自然科学基金项目(60603015);高等学校全国优秀博士学位论文作者资助项目(2007B4);湖南省教育厅资助科研项目(湖南省优秀博士学位论文获得者资助项目)

摘　　要：非线性形变在指纹图像中非常普遍,严重时会导致指纹识别系统的识别性能显著降低,是指纹图像识别领域重要的研究课题之一.尤其对基于细节点的指纹匹配方法而言,由于相当一部分细节点的位置和方向都在非线性形变的作用下发生了改变,从而导致算法匹配准确率下降.在对非线性形变图像反复观察和实验的基础上,发现了细节点偏离的一些有用的规律,并用于细节点的位置和方向调整.提出了一种基于多参考细节点对齐修正指纹图像中部分细节点的方法,首先获取一组匹配细节点对作为基准点对,然后再以此为基础对邻域其他细节点进行修正,在FVC2004 DB2数据集上进行了一系列实验,实验结果表明该算法能够明显改善匹配性能.Nonlinear distortion is common in sensor-acquired fingerprint images. It degrades the performance of fingerprint verification systems severely. Especially for minutiae-based matching algorithm, which is most widely adopted, both the location and direction of minutiae are changed because of distortion. Based on the result of careful observation and examination on kinds of distorted fingerprint images,some useful characteristics of minutiae＇ deviation on nonlinear distorted fingerprint image are found, and further used for minutiae adjustment. In this paper, a method to adjust the minutiae position and direction on the basis of multiple reference minutiae alignment method is proposed. However, other effective alignment methods can also be taken instead. Query fingerprint image and template fingerprint image are aligned first using multiple reference minutiae alignment method, and some landmark minutiae pairs from two images are computed. Then minutiae around each paired landmark minutia are adjusted, and comparison between adjusted query minutiae set and template minutiae set is made during the matching process. Compared with other typical methods, this method is easier to be added to existing minutiae-based matching algorithms while no additional feature but minutia is needed. Experimental results on FVC2004 DB2 show that this method can improve the matching performance clearly.

关 键 词：非线性形变 细节点调整 多参考细节点 FVC2004 ROC曲线 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]