An Approach to Improve Mouth-State Detection to Support the ICAO Biometric Standard for Face Image Validation

Face image analysis continues as an ongoing challenge in biometrics and image processing due to the state variations of facial elements. In this context, the mouth-state plays a fundamental role because its impact on the perception of facial gestures. Current work on mouth-state detection is mainly focused on the creation of classifiers derived from large training datasets. This technique requires extensive training sessions and its results entirely rely on the quality of datasets and learning methods. This paper describes an original approach for detecting mouth-state to support the ICAO standard for face image validation. The proposed approach reduces the error margins by considering face proportions for image segmentation and estimates the magnitude of mouth aperture by conducting an analysis of skin color. Experimentation demonstrates improvements by 21% on the correct detection of mouth-state by slightly affecting the processing time in comparison to the classifiers approach.