结合特征增强注意力的混合卷积去雾网络

打开文本图片集

中图分类号：TN911.73-34；TP391 文献标识码：A 文章编号：1004-373X（2026）01-0027-07

DOI：10.16652/j.issn.1004-373x.2026.01.005引用格式：，等.结合特征增强注意力的混合卷积去雾网络[J].现代电子技术，2026，49（1）：：27-33.

Mixed convolutional dehazing network combining feature enhancement attention

FUChengcheng，WEIWeimin，YANG Tong，YANG Tiancheng （CollegeofComputer ScienceandTechnology，ShanghaiUniversityofElectricPower，Shanghai 2O13o6，China）

Abstract：Fogcancause severe visual degradation toimages，andafect theirdetailsand contrast.Furthermore，it wil impactthereadabilityoftheimagesandtheperformanceofsubsequentprocessingtasks.Inviewoftheincompletefeature extraction，lossofimagedetails，andpoordehazingeffectonnon-uniformhazyimagesfoundinexisting imagedehazing algorithms，amixedconvolutionaldehazingnetworkintegratingfeatureenhancementatentionisproposed．Diferential convolutioniscombinedwithriginalconvolutiontoformamixedconvolutionlayer，expandingthefeatureinformationextraction range.Thefeatureenhancementattention moduleformed bypixelatention mechanismandconvolutionalblockatentionmodule isusedtoimprovethedetailprocessngabilityofthenetwork.Thefeatureinformationofchannel，spaceandpixelisfusedto makethenetwork focusonthediferencesoffogdistribution.Experimentalresultsshowthattheproposednetworkcanextract featurescomprehensively，producedetailedandcleardehazedimages，andachievethoroughdehazing.Itperformswellonboth objectiveindicatorsandsubjectivevisualassessments，andhasgooddehazingefectwhilemaintainingstrongrobustnessand generalization ability.

Keywords：imagedehazing;image procesing;atention mechanism;featureenhancement;mixedconvolution;feature fusion

0 引言

图像去雾技术可以改善受雾气影响的图片质量。（剩余11891字）