图像去雾算法研究现状与展望  

Research status and prospect of image dehazing algorithm

作　　者：张登银[1] 鞠铭烨[1] 钱雯 ZHANG Dengyin;JU Mingye;QIAN Wen(School of Internet of Things,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学物联网学院,江苏南京210003

出　　处：《南京邮电大学学报（自然科学版）》2020年第5期101-111,共11页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61872423);江苏省高等学校自然科学研究重大项目(19KJA180006);江苏省重点研发计划(产业前瞻)(BE2017111)资助项目。

摘　　要：雾天环境所摄户外图像往往存在视觉模糊、饱和度过低,以及颜色灰白等负面视觉特征,这不仅影响了用户的主观体验,且对后续智能系统的研判也带来了巨大隐患。文中首先总结了图像去雾的发展历史及其研究现状;随后,重点围绕传统图像增强、物理模型去雾以及深度学习模型去雾3个不同处理策略,对当前主流或经典图像去雾算法进行了深入分析,并归纳了其所面临的共性问题,如去雾不彻底、天空区域过增强、近景的过饱和等;最后,针对这些共有缺点或局限,提出了若干具有科学性的研究思路与可行的建议。Outdoor images taken in a hazy environment usually have negative features,such as blurred vi­sion,low saturation and gray color,thus affecting the user's subjective experience,and brings a certain degree of the hidden error for subsequent intelligent system research and judgment.Firstly,the develop­ment of image defogging and its research status are summarized.Then,focusing on three different dehazing strategies,i.e.,traditional image enhancement,physical model dehazing,and deep learning model dehaz-ing,the mainstream algorithms are analyzed and some common problems faced by current methods are summarized,such as incomplete haze removal,over-enhancement of the sky area,and over-saturation of the close-range,etc..Finally,aimed at these shortcomings or limitations,scientific research ideas and fea­sible suggestions are proposed.

关 键 词：可视性恢复 图像去雾 机器视觉 图像清晰化 图像增强 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]