一种基于GAN的政务数据中模糊图像复原算法研究

打开文本图片集

引用格式：，.一种基于GAN的政务数据中模糊图像复原算法研究[J].现代电子技术，2025，48（18）：134-138.

关键词：政务数据处理；模糊图像复原；生成对抗网络；卷积神经网络；AutoEncoder网络；U-Net网络中图分类号：TN919-34；TP391 文献标识码：A 文章编号：1004-373X（2025）18-0134-05

Research on GAN-based blurry image restoration algorithm in government data

XUANWenjing，ZHANGXinyue （CollegeofTechnology，HubeiEngineeringUniversity，Xiaogan432ooo，China）

Abstract：Governmentdatacontainsalarge amountof imagedata，which playsacrucialroleinrecordingkeyinformation. Thevaguenessoftenappears ingovernmentimages，whichbringsgreat troubletotheextractionandutilizationof information.On thisbasis，agenerativeadversarialnetwork（GAN）basedblurryimagerestorationalgorithm（GovRGAN）isproposed.Inthis algorithm，theGANisusedfortheimagerestoration，whichcanefectivelylearnandrecoverdetailedimageinformationIis composedof generatoranddiscriminator.Thegeneratorof theGANispre-trainedbymeansofthetrainedweightparametersof U-Netnetwork.Theconvolutionalneuralnetworkisusedasthediscriminatortodistinguishbetweenrealimagesandthose generatedbythegenerator.Inodertovalidatethealgorit'sfectiveness，agovermentdatasetwith5OOinvoicevouchersis constructedimingtoprovidesufficientanddiversetrainingsamplesforthemodel.Themotionbluranddefocusblurareused forthedegradationprocesing，making thedataclosertoblurryimagesinreality.Thecomparativeexperimentswereconducted onthisdatasetbetwee GovRGAN，AutoEncodernetwork，andU-Net，verifying thatGovRGANexhibitsexcelentperforancein restoringblurredgovernmentiages，andthequalityoftherestoredimageshasbeenimprovedsignificantlyOnthemotiofuzzy dataset，incomparisonwiththeU-Netnetwork，thePSNRandSSIMvaluesoftheproposedalgorithmareimprovedby9.664dB and 0.157，respectively.

Keywords：government data processing;；blurry image restoration；generativeadversarial network；convolutional neural network;AutoEncodernetwork;U-Net network

0 引言

信息化时代，政务数据对政府决策、服务及监管至关重要。（剩余6760字）