关键词: 光场图像, 质量评价, 光场图像的表现形式, 人眼视觉系统

Abstract: Distortions often occur during the acquisition, compression, transmission, reconstruction, and display of light field images, degrading the visual experience of users. To address different types of distortions, researchers have proposed various quality assessment methods to accurately evaluate light field image quality. Existing reviews of light field image quality assessment methods typically classify them based on reference information or mapping techniques. However, these classification criteria fail to fully capture the unique spatial and angular characteristics of light field images. This paper provides a review of recent advancements in light field image quality assessment. Firstly, based on the representation forms of light field images, the quality assessment methods are categorized into six types: sub-aperture image-based methods, epipolar plane image-based methods, micro-lens image-based methods, pseudo video sequence-based methods, refocused image-based methods, and mixed extraction methods. Representative methods from recent years are introduced, and the strengths and limitations of each category are summarized. Then, seven commonly used light field image datasets and three performance evaluation metrics are listed, and the performance of different methods is compared and analyzed from the perspectives of representation characteristics, datasets, and model structures. Finally, future research directions are discussed, including multimodal fusion, model lightweighting, large-scale models, human visual system modeling, and high-dimensional extensions.

Key words: light field image, quality assessment, representation of light field image, human visual system