da/sec scientific talk on Biometrics

Topic: Face Image Quality Assessment Research Retrospective

by Torsten Schlett
D19/2.03a (also online via the corresponding BBB room), November 19, 2026 (Thursday), 12.00 noon

Keywords — Research Overview, Biometric Utility, Face Recognition, Evaluation Methodology

Abstract

„This talk provides an overview of the speaker’s Ph.D. project research work on various aspects of Face Image Quality Assessment (FIQA). FIQA commonly is about the automatic estimation of the utility of a face image for face recognition or other biometric purposes, which can involve an overall unified quality assessment, or more specific analysis of certain defect classes, such as motion blur. The presentation mostly concentrates on broader summaries, including research on duplicates in face image datasets, lossy face image compression, FIQA-assisted Differential Morphing Attack Detection (D-MAD), and FIQA for motion blur defects. Special attention in the form of some additional details is given to the topic of FIQA evaluation methodologies, which are likely to remain relevant to a wide variety of future FIQA research, and possibly quality assessment research for other biometric modalities. These in particular are Error versus Discard Characteristic (EDC) plots for evaluations corresponding to an image discard scenario, and cumulative distribution plots to study the demographic variability of FIQA algorithms.“