Assessing Facial Image Accordance to ISO/ICAO Requirements

摘要

Face has been adopted as default biometric validation method and the International Standard Organization (ISO) proposed a standard which states constraints for facial image. This paper presents methods for evaluating face image conformance to the following ISO/ICAO requirements: pixelation, hair across eyes, veil over face and mouth opened. Each requirement is individually evaluated. The algorithm for analyzing pixelation achieved an equal error rate (EER) equals to 1.7%, result very close to state-of-the-art (0.0% EER). The "Hair Across Eyes" analysis method achieved an EER equals to 11.9% which surpass the best state-of-art result (12.4%). The algorithm for evaluating "Veil Over Face" requirement achieved EER equals to 1.2% which also surpass the best state-of-art result (2.5%). The "Mouth Opened" requirement achieved an EER equals to 4.20%, a result compatible with state-of-art rates for this requirement (3.3%).