A Deep Insight into Measuring Face Image Utility with General and Face-specific Image Quality Metrics

  title={A Deep Insight into Measuring Face Image Utility with General and Face-specific Image Quality Metrics},
  author={Biying Fu and Congmin Chen and Olaf Henniger and Naser Damer},
  journal={2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  • Biying Fu, Congmin Chen, N. Damer
  • Published 21 October 2021
  • Computer Science
  • 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
Quality scores provide a measure to evaluate the utility of biometric samples for biometric recognition. Biometric recognition systems require high-quality samples to achieve optimal performance. This paper focuses on face images and the measurement of face image utility with general and face-specific image quality metrics. While face-specific metrics rely on features of aligned face images, general image quality metrics can be used on the global image and relate to human perceptions. In this… 

Figures and Tables from this paper

CR-FIQA: Face Image Quality Assessment by Learning Sample Relative Classifiability
Through extensive evaluation experiments on eight benchmarks and four face recognition models, this work demonstrates the superiority of the proposed CRFIQA over state-of-the-art (SOTA) FIQA algorithms.
The Effect of Wearing a Face Mask on Face Image Quality
This work studies, for the first time, the effect of wearing a face mask on face image quality by utilising state-of-the-art face imagequality assessment methods of different natures, aimed at providing better understanding on the effects of face masks on the operation of face recognition as a whole system.
Explainability of the Implications of Supervised and Unsupervised Face Image Quality Estimations Through Activation Map Variation Analyses in Face Recognition Models
  • Biying Fu, N. Damer
  • Computer Science
    2022 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)
  • 2022
A novel set of explainability tools to derive reasoning for different FIQA decisions and their face recognition (FR) performance implications and points out that high-quality images typically cause consistent low activation on the areas outside of the central face region, while low- quality images, despite general low activation, have high variations of activation in such areas.
Evaluation Gaps in Machine Learning Practice
The evaluation gaps between the idealized breadth of evaluation concerns and the observed narrow focus of actual evaluations are examined, pointing the way towards more contextualized evaluation methodologies for robustly examining the trustworthiness of ML models.
Face Image Quality Assessment: A Literature Survey
This survey provides an overview of the face image quality assessment literature, which predominantly focuses on visible wavelength face image input and a trend towards deep learning based methods is observed, including notable conceptual differences among the recent approaches.


The relative contributions of facial parts qualities to the face image utility
This work investigates the effects of face subregions (semantic contexts) and link the general image quality of faceSubregions with face image utility and applies four face image quality assessment methods on four face sub Regions and the aligned faces.
Learning Face Image Quality From Human Assessments
This is the first work to utilize human assessments of face image quality in designing a predictor of unconstrained face quality that is shown to be effective in cross-database evaluation.
SER-FIQ: Unsupervised Estimation of Face Image Quality Based on Stochastic Embedding Robustness
This work proposes a novel concept to measure face quality based on an arbitrary face recognition model that avoids the training phase completely and further outperforms all baseline approaches by a large margin.
The effect of face morphing on face image quality
It is shown that MagFace (face utility metric) shows a clear difference between the bona fide and the morph images, regardless if they were digital or re-digitized, and most quality and utility metrics do not capture the artifacts introduced by the morphing process.
Predicting Face Recognition Performance Using Image Quality
The results show that the proposed model can accurately predict performance using an accurate and unbiased Image Quality Assessor (IQA) and highlight the impact of the unaccounted quality space -- the image quality features not considered by IQA -- in contributing to performance prediction errors.
SDD-FIQA: Unsupervised Face Image Quality Assessment with Similarity Distribution Distance
A novel unsupervised FIQA method that incorporates Similarity Distribution Distance for Face Image Quality Assessment (SDD-FIQA), which generates quality pseudo-labels by calculating the Wasserstein Distance between the intra-class and inter-class similarity distributions and is capable of training a regression network for quality prediction.
Face Image Quality Assessment Based on Learning to Rank
A learning to rank based framework for assessing the face image quality is proposed and Experimental result demonstrates its effectiveness in improving the robustness of face detection and recognition.
A Bayesian model for predicting face recognition performance using image quality
A Bayesian approach to model the relation between image quality and corresponding face recognition performance and it is shown that this model can accurately aggregate verification samples into groups for which the verification performance varies fairly consistently.
FaceQnet: Quality Assessment for Face Recognition based on Deep Learning
A Quality Assessment approach for face recognition based on deep learning that employs an existing ICAO compliance framework and a pretrained CNN to automatically label data with quality information and shows that the predictions from FaceQnet are highly correlated with the face recognition accuracy of a state-of-the-art commercial system not used during development.
Reduced-Reference Image Quality Assessment by Structural Similarity Estimation
This paper extracts statistical features from a multiscale multiorientation divisive normalization transform and develops a distortion measure by following the philosophy in the construction of SSIM, a widely used full-reference image quality measure shown to be a good indicator of perceptual image quality.