Fingerprint Image Quality and Prediction of Matching Performance

Due to their high reliability, fingerprints have been extensively used as a biometric identifier. The performance of an automatic fingerprint authentication system relies heavily on the fingerprint image quality as seen in several studies. In this work, we present a new method to quantify fingerprint image quality which is relevant to matcher performance. The ultimate goal of this research is to determine and overcome the underlying causes of the poor match. Newly developed wavelet features and previously developed spatial features, as inputs to a fuzzy c-means classifier, are used predict matcher performance. Results are obtained for two different matchers, namely NFIS bozorth3 and Verifinger (Neurotechnologija), for two different optical sensors, Crossmatch Verifier 300 and Secugen Hamster III.