基于深度学习的图像去雾研究综述  

A Review of Research on Image Defogging Based on Deep Learning

作　　者：吉波涛 陆利坤 

机构地区：[1]北京印刷学院信息工程学院,北京

出　　处：《图像与信号处理》2025年第1期21-33,共13页Journal of Image and Signal Processing

基　　金：北京市教育委员会出版学新兴交叉学科平台建设–数字喷墨印刷技术及多功能轮转胶印机关键技术研发平台(项目编号:04190123001/003);北京市数字教育研究重点课题(BDEC2022619027);北京市高等教育学会2023年立项面上课题(课题编号:MS2023168);北京印刷学院校级科研项目(20190122019、Ec202303、Ea202301、E6202405);北京印刷学院学科建设和研究生教育专项(21090122012、21090323009);北京市自然科学基金资助项目(1212010)。

摘　　要：恶劣天气环境下拍摄的图像会受到雾或霾的影响,从而导致图像饱和度过低模糊、以及颜色灰白等负面效果,这不仅会使图像中的重要信息丢失,还会对后续计算机视觉任务(如目标检测、图像分割、人员再识别)的研究造成负面影响。为了解决上述问题,文章首先对图像去雾的发展历程进行分析和梳理,接下来重点论述了深度学习在图像去雾领域的研究进展,主要包含有监督去雾、无监督去雾和半监督去雾技术,并对各自的代表性算法进行深入对比分析。最后,介绍了图像去雾领域主流的数据集和评估指标。Images captured in harsh weather environments are often affected by fog or haze, which can lead to negative effects such as low saturation, blurring, and grayish-white colors. This not only results in the loss of important information in the image, but also has a negative impact on subsequent computer vision tasks such as object detection, image segmentation, and personnel re-identification. This article first provides a comprehensive analysis and sorting of image defogging and then reviews the research progress of deep learning in the field of image defogging, mainly including supervised defogging, unsupervised defogging, and semi-supervised defogging. We compared and analyzed representative algorithms among these methods. Finally, the commonly used datasets and evaluation metrics for image defogging were introduced.

关 键 词：深度学习 单幅图像去雾 图像处理 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]