基于小波变换的最优滤波的图像去噪

《基于小波变换的最优滤波的图像去噪》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、32InternationalJournalofInformationProcessingSystemsVol.1,No.1,2005Wavelet-basedImageDenoisingwithOptimalFilterYong-HwanLee*,andSang-BurmRhee*Abstract:Imagedenoisingisbasicworkforimageprocessing,analysisandcomputervision.Thispaperproposesanovelalgorithmbasedo

2、nwaveletthresholdforimagedenoising,whichiscombinedwiththelinearCLS(ConstrainedLeastSquares)filteringandthresholdingmethodsinthetransformdomain.WedemonstratedthroughsimulationswithimagescontaminatedbywhiteGaussiannoisethatourschemeexhibitsbetterperformanceinbo

3、thPSNR(PeakSignal-to-NoiseRatio)andvisualeffect.Keywords:ImageDenoising,NoiseReductionWavelet1.Introductionthisworkswellonlyiftheunderlyingsignalissmooth.Toovercometheweaknessofthespatialfiltering,awaveletDigitalimageshaveapplicationsindailylife,suchasbasedde

4、noisingschemeisintroduced[2].Waveletsgiveadigitalcameras,HDTV(HighDefinitionTelevision)andinsuperiorperformanceinimagedenoisingduetopropertiesareasofresearchandtechnologyincludingGIS(Geo-suchassparsityandmultiresolutionstructure.graphicalInformationSystem).Da

5、tasetscollectedbySimpledenoisingalgorithmsthatusedthewaveletimagesensorsaregenerallycontaminatedbynoiseandtransformconsistofthethreesteps[3].noisecanbeintroducedbytransmissionerrorsandStep1.Calculatethewavelettransformofthenoisycompression.Theproblemofimagede

6、noisingistorecoversignal;animagethatiscleanerthanitsnoisyobservation.Thus,Step2.Modifythenoisywaveletcoefficientsaccordingnoisereductionisanimportanttechnologyinimagetoarule;analysisandthefirststeptobetakenbeforeimagesareStep3.Computetheinversetransformusingt

7、heanalyzed[1].modifiedcoefficients;Althoughwaveletshaveefficientnoisereductionability,Oneofthemostwell-knownrulesforstep2issoftwaveletsstillhaveproblemsonaheavynoisynetwork.Wethresholdinganalyzedby[4].Duetoitseffectivenessandinvestigatetheproblemofimagedenois

8、ingwhenthesimplicity,itisfrequentlyusedintheliterature.ThemainsourceimageiscorruptedbyadditivewhiteGaussiannoise,ideaistosubtractthethresholdvalueTfromallwhichisavalidassumptionforimageso