The Fingerprint Quality Evaluation And Its Application

Fingerprint quality control is one of the most important parts of automated fingerprint identification system. Considering the differences of the collection instrument types, the extent of the finger's wet and dry states, the injuries of the fingers and so on, approximate 10% of the fingerprint images are belonged to low-quality fingerprints. Low-quality image has significant impact on feature extraction, fingerprint classification, and feature matching. Fingerprint quality evaluation can be divided into three aspects as following: 1) overall quality evaluation; 2) local quality evaluation; 3) minutiae quality evaluation. Overall quality evaluation is mainly used for the control of the fingerprint registration. Local quality evaluation can be directly used on fingerprint segmentation. The minutiae quality evaluation can be added to minutiae matching algorithm.This paper focuses on the local quality and the minutiae quality evaluation and theirs applications. First of all, the paper used a local quality evaluation with entropy feature. The higher local entropy and the quality, the better and the clearer the texture were, and could be used to the local quality evaluation more effectively. As a direct application of the local quality evaluation, we developed a fingerprint image segmentation algorithm based on entropy feature and AdaBoost classifier. The algorithm was tested and compared with other common segmentation algorithm on FVC2000, FVC2002, FVC2004, which were the databases of fingerprint verification competition. Experiment results showed that our algorithm was optimal to be applied in all acquisition instruments, and better than other segmentation algorithms both in terms of error rate and stability. Second, the paper also studied a minutiae quality evaluation algorithm. The minutiae local structure had a broad application in fingerprint matching algorithm. The quality of the fingerprint minutiae structure could exerted great influence on the performance of matching directly. In this paper, we proposed a minutiae structure quality factor, and embed the quality factor into the existing matching algorithm with minutiae structure. Experimental results of this algorithm tested on FVC2002 DB2 showed that the quality of minutiae structural could improve the accuracies of fingerprint matching. The EER was increase by 0.54% after the minutiae quality was added to matching algorithm.