基于跨层协同交互与空频联合感知的遥感图像去雾

打开文本图片集

关键词：遥感图像去雾；跨层协同交互；空频联合感知；特征融合；U型网络；Mamba-CNN中图分类号：TN911.73-34；TP751 文献标识码：A 文章编号：1004-373X（2026）09-0051-09

Remote sensing image dehazing based on cross-layer collaborative interaction andspatial-frequency jointperception

Yu Mei1，2，3，LiuShaobin1，²，LuLin1，2

（1.Hubei 2.CollegeofComputerandInformationTechnology，ChinaThreeGorgesUnversity，Yichang443oo2，China; 3.ThreeGorgesDigital IntelligenceInstitute，ChinaThreeGorgesUniversity，Yichang 443OO2，China）

Abstract：Inviewof theimage degradation，texture loss，andcolordistortion inthedehazingprocessof theexisting U-shapeddehazing networks，thispaper proposesaremotesensing image dehazing network basedoncross-layercolaborative interactionandspatialfrquencyjtprcetionSpecifcallyo-aerollbratieintectionmouleisitruceds modulecanefectivelycapture long-range dependenciesamong cross-layer features，expandthelocalreeptivefield，and facilitatethecolaborativeinteractionof globalandlocalinformationduringbothencodinganddecodingstages，soatoenhance thedehazing performance.Aditionall，aspatial-frequencyjointperceptionmoduleisproposed.Thismodulecanrealizethe complementofthespatialdomainandfrequencydomainbyfusingspatialandfrequencyfeatures，soastoimprovetherecovery of imagetextureandcolor.Experimentsshowthattheproposedalgorithmoutperformstheexistingmethodsonthepublicremote sensing haze datasets HRSD and RRSHID-Thick.To sum up，this method has a certain applicability.

Keywords：remotesensingimagedehazing;cross-layercolaborativeinteraction;spatial-frequencyjointperception;feature fusion;U-shaped network;Mamba-CNN

0 引言

随着环境污染加剧和极端天气增多，雾霾对光学传感器捕获的图像质量造成了严重影响，导致图像出现模糊、对比度不足和空间信息丢失等问题。（剩余12610字）