FoD Enroll Image Quality Classification Method for Fingerprint Authentication System

Abstract

Typical fingerprint authentication system flow including preprocessing, feature extraction, and feature matching. To improve the user experience of it, more intelligent process for such system is needed. Fingerprint on display (FoD) is a popular kind of sensing technique in recent years, in this research, we proposed an enroll image quality classification method for the preprocessing step of the system, which is able to reject the invalid input, in order to shorten the response time, especially for FoD applications. We had evaluated our proposed method through a self-collected FoD sensing image dataset, including 50,130 fingerprint images, and proved that our method is able to reach 95.83% accuracy, which is really helpful for the improvement of the system’s user experience.