Evaluating Image Quality Estimators for Face Matching
Praneet Singh, Haoyu Chen, Edward J. Delp, Amy R. Reibman · 2022
Understanding the quality of a face image can be useful for improving the performance of automated face matching systems. With the increasing number of face quality estimators (QEs) being proposed recently, it is important to have systematic methods to evaluate and compare the performance of these QEs. In this paper, we describe two existing strategies for evaluating face QEs, and propose several new approaches. Our new approaches focus on targeted QE evaluation using carefully constructed image datasets. We show that these strategies lead to important insights about the effectiveness of existing face QEs.