关键词: 非线性形变, 细节点局部描述子, 指纹图像匹配, 指纹识别竞赛, 等错误率

Abstract:

Nonlinear distortion is common in fingerprint images. It degrades the performance of fingerprint verification systems severely. To overcome the global distortion problem in fingerprint images, this paper presents a principle for distorted fingerprint image matching as “align globally while match locally, align coarsely while match accurately”. Based on such a principle, this paper defines a local minutia descriptor, and further uses it to align and match two fingerprint images. The similarity computing method is especially studied. Experimental results on FVC 2000 DB2, FVC2002 DB1 and FVC 2004 DB2 prove the superiority of the proposed method. It can restrict the effect of nonlinear distortion on fingerprint image matching.

Key words: nonlinear distortion;local minutia descriptor;fingerprint matching;FVC;EER