改进暗通道先验的大幅面倾斜影像快速去雾  

Fast dehazing for large format oblique images based on improved dark channel prior

作　　者：黄凰 欧阳欢 董杨[2] 贺欣 赵星皓 HUANG Huang;OUYANG Huan;DONG Yang;HE Xin;ZHAO Xinghao(93920 Troop,Hanzhong 723213,China;Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]93920部队,陕西汉中723213 [2]信息工程大学,河南郑州450001

出　　处：《光学精密工程》2025年第3期476-485,共10页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.41971427);河南省自然科学基金资助项目(No.242300421665)。

摘　　要：针对经典暗通道去雾算法针对大幅面倾斜影像存在远地端去雾不彻底,且处理效率不高的问题,提出一种改进暗通道先验的倾斜影像快速去雾算法。采用分级窗口加权融合的方法求解暗通道进而细化透射率,然后结合辐射传输方程,根据倾斜影像的固定成像角度构建雾气分布模型,实现透射率变换;最后进行大气光值估计并完成去雾。在此基础上设计了并行计算方案以提升去雾速度。实验结果表明,在浓雾条件下处理后影像信息熵提升约22.7%,平均梯度提升约30.0%;批处理去雾相较经典暗通道算法速度提升约11倍,显著提升了处理效率。本文算法能够克服倾斜影像雾气分布不均的影响,显著改善影像质量。The traditional dark channel dehazing algorithm is limited by its inability to adequately remove haze from the distal regions of large-format oblique images and suffers from low processing efficiency.To address these challenges,an improved dark channel prior-based dehazing algorithm for oblique images was proposed in this paper.Initially,a rapid compensation method for the dark channel was employed,utiliz⁃ing block-level and pixel-level window weighted fusion to enhance transmittance estimation.Subsequent⁃ly,the dark channel prior was integrated with the radiative transfer equation to establish a haze distribution model and calculate transmittance.The atmospheric light value was then estimated to complete the dehaz⁃ing process.Furthermore,a parallel computing scheme was designed to enhance dehazing speed.Experi⁃mental results indicate that the proposed algorithm effectively mitigates the effects of uneven haze distribu⁃tion,thus significantly improving the quality of hazy images.In scenes with dense haze,image entropy and average gradient values increase by approximately 22.7%and 30.0%,respectively.Additionally,the dehazing speed during batch processing is approximately 11 times faster than that of traditional dark channel algorithms,demonstrating a substantial enhancement in processing efficiency.

关 键 词：影像去雾 大幅面倾斜影像 辐射传输方程 暗通道先验 并行运算 

分 类 号：TP394.1[自动化与计算机技术—计算机应用技术]