We address the problem of the existence of quality measures for face recognition. We introduce the concept of an oracle quality measure, which is an optimal quality measure. We approximate oracle quality measures by greedy pruned ordering (GPO). GPO analysis provides a best case performance upper bound for traditional quality measures. We assess the performance of 12 commonly proposed face image quality measures. In addition, we investigate the potential for learning quality measures using supervised learning. Finally, we show that GPO analysis is applicable to other biometrics.
Conference Dates: September 30-October 2, 2013
Conference Location: Arlington, VA
Conference Title: The IEEE Sixth International Conference on Biometrics: Theory, Applications and Systems (BTAS 2013)
Pub Type: Conferences