Qualifying Fingerprint Samples Captured by Smartphone Cameras in Real-Life Scenarios
Abstract
While biometrics has been extensively adopted by industry and governments for identification and forensics purposes relying on dedicated biometric sensors and systems, the consumer market driven by innovations in consumer electronics (smartphones, tablets, etc.) is believed to be the next sector that biometric technologies can find wider applications. Compared to dedicated biometric sensors, the sensors embedded in such general-purposed devices may suffer from sample quality instability, which has significant impact on biometric performance. The concern on sample quality may jeopardize the market confidence in consumer devices for biometric applications. In this paper, we propose an approach to assessing the quality of fingerprint samples captured by smartphone cameras under real-life uncontrolled environments. Our approach consists of a sample processing pipeline during which a sample is divided into blocks and a set of local quality features are extracted from each block, including 3 pixel-based features, 4 autocorrelation based features, and 5 frequency features from the autocorrelation result. Afterwards, a global sample quality score is calculated by fusing all image blocks’ qualification status. Thanks to the extracted features’ capability in discriminating high-quality foreground (fingerprint area) blocks from low-quality foreground ones and background ones, the proposed approach does not require foreground segmentation in advance and thus we call it a one-stop-shop approach. Experiments compare the proposed approach with NFIQ and the proposed pipeline using standardized quality features, and demonstrate our approach’s better performance in qualifying smartphone-camera fingerprint samples.