文档介绍：Published online：2022-07-20 09:00:53
URL：.
ropermethod.
Keywords
:singleimagesuper-resolution (SISR);verydeepsuper-resolutionconvolutionalneuralnetwork (VDSRCNN);
motionblurredimage;imagequalityindex
0 Introduction neuralnetwork(VDSRCNN),theLRimagescanbe
improvedto
Imagesuper-resolution (ISR)isaprocesswhich trainingaverydeep CNN wasdifficultbecauseof
cancreateahigh-resolution(HR)imagefromalow- slow convergence rate. however, the residual-
resolution(LR) meansthat ixeldensit learningandextremelyhighlearningratescansolve
g p y [4-5]
withinanimageishigh,therefore,anHRimagecan thisproblem .
offer details that may be critical in various basedonresiduallearning,whichmeansthenetwork
[1]
applications .ForISR,themainmethodissuper- canlearntoestimatetheresidualimage.
resolution image reconstruction (SRIR ). The In recent years,the cnn has been extensively
learnin-basedmethod,oneoftheSRIR methods,is popularowingtoitssuccessinimageclassification,