预览加载中，请您耐心等待几秒...

基于深度学习的图像去噪研究.docx

预览
1/3
2/3
3/3

在线预览结束，喜欢就下载吧，查找使用更方便

下载提示 文本预览

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

基于深度学习的图像去噪研究 Abstract Imagedenoisingplaysanessentialroleinmanycomputervisionapplications.Inrecentyears,deeplearning-basedmethodshaveshowngreatpotentialinthefieldofimagedenoisingduetotheirsuperiorperformance.Inthispaper,wereviewthestate-of-the-artdeeplearning-basedimagedenoisingmethodsandanalyzetheirstrengthsandweaknesses.Wealsodiscusssomepromisingfutureresearchdirectionsinthisfield. Introduction Imagedenoisingistheprocessofremovingnoisefromimageswithoutlosingtoomuchimagedetail.Thepurposeofimagedenoisingistoimprovethevisualqualityoftheimagesandmakethemmoresuitableforfurtheranalysis.Inthepastfewdecades,alargenumberoftechniqueshavebeenproposedforimagedenoising,includingtraditionalmethodssuchasfilteringandwavelet-basedmethods. Withtherapiddevelopmentofdeeplearning,manyresearchershaveexploredtheuseofdeepneuralnetworksforimagedenoising.Comparedwithtraditionalmethods,deeplearning-basedmethodscanlearnmorecomplexandabstractimagefeatures,leadingtobetterimagedenoisingperformance.Inaddition,deeplearning-basedmethodscaneasilyintegratewithothercomputervisiontasks,suchasimageclassificationandsegmentation. Inthispaper,wereviewthestate-of-the-artdeeplearning-basedimagedenoisingmethodsandanalyzetheiradvantagesandlimitations.Wealsodiscusssomepromisingfutureresearchdirectionsinthisfield. DeepLearning-basedImageDenoisingMethods Deeplearning-basedimagedenoisingmethodscanbebroadlyclassifiedintotwocategories:supervisedlearningandunsupervisedlearning.Insupervisedlearning,thedenoisingmodelistrainedwithpairsofnoisyandcleanimages,whileinunsupervisedlearning,themodelistrainedwithonlynoisyimages. SupervisedLearning-basedImageDenoisingMethods Supervisedlearning-basedimagedenoisingmethodshaveachievedexcellentperformanceinrecentyears.OneofthemostsuccessfulmodelsistheDeepResidualNetwork(DnCNN)proposedbyZhangetal.in2017.DnCNNisafullyconvolutionalneuralnetworkthatcaneffectivelyremovenoisewhilepreservingimagedetails.Themodelconsistsofmultipleconvolutionallayerswithresidualconnections,anditistrainedona