Comparison of quality-based fusion of face and iris biometrics

Abstract

Multimodal systems have been used for the increased robustness of biometric recognition tasks. A unique strength of multimodal systems can be found when presented with biometric samples of degraded quality in a subset of the modalities. This study looks at the effect of quality degradation on system performance using the Q-FIRE database. The Q-FIRE database is a multimodal database composed of face and iris biometrics captured at defined quality levels, controlled at acquisition. This database allows for assessment of biometric system performance pertaining to image quality factors. Methods for measuring image quality based on illumination conditions are explored as well as strategies for incorporating these quality metrics into a multimodal fusion algorithm. This paper provides further evidence in a unique dataset that utilizing sample quality metrics into the fusion scheme of a multimodal system improves system performance in non-ideal acquisition environments.