论文标题
面部图像质量评估的艺术状态
State of the Art of Quality Assessment of Facial Images
论文作者
论文摘要
项目“ EES面部指标”的目标是开发和发布一种开源算法，用于质量评估面部识别面部图像（ofiq）的质量评估，特别是用于边境控制场景。1为了刺激对QA的要求和实践的协调，用于QA的质量图像和实践是针对面部图像的20种杂物，该项目的兴趣是2的。 ISO/IEC 29794-5标准。此外，实施的质量指标和算法将考虑其他相关标准的建议和要求，特别是ISO/IEC 19794-5：2011，ISO/IEC 29794-5：2010，ISO/IEC/IEC 39794-5：39794-5：2019和2019年版本和BSI技术指南的5.2版，以供BSI TRE-03121。调查了相应质量评估算法，方法和算法的现状（定义度量），用于面部图像的质量评估的实现和数据集。对于所有相关质量方面，本文档总结了上述标准的要求，已知的结果对其对面部识别性能的影响，公开可用的数据集，建议的方法和算法以及开源软件实现。
The goal of the project "Facial Metrics for EES" is to develop, implement and publish an open source algorithm for the quality assessment of facial images (OFIQ) for face recognition, in particular for border control scenarios.1 In order to stimulate the harmonization of the requirements and practices applied for QA for facial images, the insights gained and algorithms developed in the project will be contributed to the current (2022) revision of the ISO/IEC 29794-5 standard. Furthermore, the implemented quality metrics and algorithms will consider the recommendations and requirements from other relevant standards, in particular ISO/IEC 19794-5:2011, ISO/IEC 29794-5:2010, ISO/IEC 39794-5:2019 and Version 5.2 of the BSI Technical Guideline TR-03121 Part 3 Volume 1. In order to establish an informed basis for the selection of quality metrics and the development of corresponding quality assessment algorithms, the state of the art of methods and algorithms (defining a metric), implementations and datasets for quality assessment for facial images is surveyed. For all relevant quality aspects, this document summarizes the requirements of the aforementioned standards, known results on their impact on face recognition performance, publicly available datasets, proposed methods and algorithms and open source software implementations.
