A survey of learning-based no-reference image quality assessment
Botao An, Hongwei Zhou, Peiran Peng, Lei Zhang, Shubo Ren, Jianan Li, Tingfa Xu · 2025
Digital images are captured by various fixed and mobile cameras, compressed with traditional and novel techniques, transmitted through different communication channels, and stored in various storage devices. Distortions can occur at each stage of the image acquisition, processing, transmission and storage pipeline, resulting in loss of perceptual information and degradation of quality. Therefore, image quality assessment is becoming increasingly important in monitoring image quality and ensuring the reliability of image processing systems. And as the most widely applicable and usable of the image quality assessment fields, a large number of learning-based no-reference quality assessment studies have been conducted in recent years. In this survey, we provide an up-to-date and comprehensive review of these studies. Specifically, this paper presents recent advances in the field of deep learning-based no-reference quality assessment and provides an overview of benchmark databases for deep learning-based no-reference quality assessment tasks as well as assessment metrics and the backbone networks commonly used in quality assessment tasks.