Fingerprint image quality assessment and scoring

Abstract

Fingerprint quality estimation is an essential step for eliminating poor quality fingerprint images which can degrade the recognition performance of automatic fingerprint identification system (AFIS). A quality assessment technique along with fingerprint quality score will enable AFIS system to make appropriate decision regarding rejecting the low quality image and recapture a better quality fingerprint image. In this paper, we propose an effective method for evaluating fingerprint image quality (dry, normal dry, good, normal wet and wet) on a local level (block-wise). Feature vector for evaluating fingerprint quality covers moisture, mean, variance, ridge valley area uniformity and ridge line count. Block-wise quality label is assigned through pattern classification based on these features. In addition to quality labels, our proposed method also provides a quality score for a fingerprint image. Manually labeled dry, normal dry, good, normal wet and wet quality blocks of FVC 2004 DB1 _a dataset is used to create a classification model using decision tree classifier. Block classification accuracy of 95.20% is achieved. Further, the same classification model is utilized to compute overall quality score of a fingerprint image. It has been observed that the overall quality score is accurate according to the manually labeled fingerprint image and also through visual inspection.