Minutiae Quality Scoring and Filtering Using a Neighboring Ridge Structural Analysis on a Thinned Fingerprint Image

摘要

This paper introduces a novel minutiae quality scoring method that relies on analyzing neighboring ridge structures around a minutia on thinned Fingerprint image. Normal ridges with neighboring a minutia have regular structures in the parts of inter-ridge distance, connectivity, and symmetry etc. This is important features of measuring minutiae quality score. It is meaning of a possibility that the present minutia is a true minutia. For making the score, Two test DB sets is firstly made for TM(True Minutiae sets) and FM(False Minutiae sets) by manually filtering minutiae found from automatic extraction. Then the score function is made by statistical method with Bayesian rule for TM and FM. I should evaluate for its discrimination power to these sets and apply to false minutiae filtering in extraction. Experimental results showed that minutiae for TM class is not nearly filtered, but ones for FM class is filtered about 30%. Therefore, I should confirm that it is useful and compatible for minutiae filtering and have an expectation in some fields.