基于深度学习的图像超分辨率综述

Review of image super-resolution based on deep learning

随着时代的发展,图像逐渐取代文字和音频,成为人们认识世界的主要手段。同时,卫星遥感与医学影像分析等领域对于图像质量的要求越来越高,使得图像超分辨率成为研究的重点发展方向。首先,文章通过数学建模的方式建立了图像退化模型,根据模型分析了图像退化的原因和图像超分辨率的原理。随后,介绍了基于深度学习的单幅图像超分辨率的研究现状与优缺点,并阐述了多种主流深度学习模型在该任务上的应用。最后,基于深度学习应用现状,介绍了其未来的发展方向和应用前景。

With the development of the times,images have gradually replaced text and audio as the main means for people to understand the world.At the same time,fields such as satellite remote sensing and medical image analysis have increasingly high requirements for image quality,making image super-resolution a key development direction in research.Firstly,the article establishes an image degradation model through mathematical modeling,and analyzes the causes of image degradation and the principle of image super-resolution based on the model.Subsequently,the research status and advantages and disadvantages of single image super-resolution based on deep learning were introduced,and the applications of various mainstream deep learning models in this task were elaborated.Finally,based on the current status of deep learning applications,the future development direction and application prospects are introduced.