中值滤波器脉冲噪声毕业论文中英文资料外文翻译文献.docx

中值滤波器脉冲噪声毕业论文中英文资料外文翻译文献.docx

《中值滤波器脉冲噪声毕业论文中英文资料外文翻译文献.docx》由会员分享，可在线阅读，更多相关《中值滤波器脉冲噪声毕业论文中英文资料外文翻译文献.docx（9页珍藏版）》请在冰豆网上搜索。

中值滤波器脉冲噪声毕业论文中英文资料外文翻译文献

中值滤波器脉冲噪声

中英文资料外文翻译文献

Improved2-DMedianFilterforOn-LineImpulseNoiseSuppressiom

Abstract-Aninproved2-Dmedianfilteremployingmultishellconcepttosuppressimpulsenoise,ispresented.Theperformanceofproposedfilterisevaluatedoverimage‘LENA’,TheimpulsivenoiseisaddedusingMATLAButility.Themodifiedstrategyreducesthemnuberofreplacementandresultsinbetterperformanceandsimplehardwarerealizationthatissuitableforon-lineimplementation.

Indexterms-MedianFilter,Multi-shellMedianFilter,ImpulseNoise

I.INTRODUCTION

InTVandotherimagingsystems,impulsenoiseisacommonimpairment.ThestandardT.V.Broadcastsignalisoftencontaminatedwithimpulsivenoisearisingfromvarioussourcessuchashouseholdelectricalapplianceandatmosphericdisturbances.Broadbandingofthesignalfurtherincreasesthelevelofimpulsivenoise.Variousfiltersareproposedtosuppresssuchimpairments[1].Themedianfilter（MF）[1-2]iswidelyusedforimpulsenoisesuppressionandthemultishellmedianfilter（MMF）[3]introducestheconceptofmissinglinerecovery.Althoughthesefiltershavesatisfactoryperformance,MMFfailstofiltertwoimpulsenoisesinthesameprossingwindow.Moveover,thesefilterstendtoblurtheimagesduetotoomanyreplacements.C.J.Juanproposedamodifiedmultishellmedianfilter（MMMF）[4],whichremovesmostoftheshortcomingsassociatedwiththeMFandtheMMF.However,itisobservedthatundercertaincondions,tobediscussedinthefollowsections,MMMFfailstoperformthedesiredfilteringoperation.Moreover,thenumberofcalculations/replacementsinvovedonthebasisofMIN/MAXconditionsisstilltoolargeandmakesthefilterdifficulttorealize,particulariyforrealtimeapplications.

Inthispaper,thethresholdstrtegyofMMMFismodifiedsothat:

（a）effectivenoisefilteringoperationsareperformedunderallconditions,and

（b）numberofcalculations/replacementsisreducedandsimplified.

Thisresultsinasimplehardwarerealizationofthefilter.

II.PROPOSEDMODIFICATION

Considera3x3-processingwindow,withP5asthecentralpixel,asshowninFigure1.

P1

P2

P3

P4

P5

P6

P7

P8

P9

Fig.1.A3x3processingwindow

TheoutputofMMMFasproposedin[4]is

Output（X,Y）=Max（P2,P8）ifP5﹥Max[S]

P5ifMin[s]﹤Max[S]

Min（P2,P8）ifP5﹤Max[S]

（1）

WhereSisthesetofsamplessurroundingcentralpixelsexcept（P4.P6）i.e.

S=｛P1,P2,P3,P7,P8,P9｝

（2）

TheprincipleinvovedinthereplacementstrategyofEquation

（1）isthatifP5iscorruptedbynoise,itisbettertoreplaceitsgraylevelbyP2orP8thanbyusingMin[S]orMax[S].also,duetomissinglineserror,sinceP4andP6maybelost,theyarenotconsideredinEquation

（2）.

ThelimitationofEquation

（1）isthatwhenMin[S]orMax[S]arealsocorruptedbyimpulsenoise,i.e.eitherMin[S]orMax[S]isequaltoP5,Equation

（1）failstoperformthedesiredfilteringoperation.ToovercomethislimitationfollowingmodificationsinthereplacementstrategyofEquation

（1）,areproposed.

Output（X,Y）=Max（P2,P8）ifP5≥Max[S]

P5ifMin[s]＜P5＜Max[S]

Min（P2,P8）ifP5≤Max[S]（3）

Ithasbeenobservedthatmorethan70-80%pointsinanimage,thegrayleveldiatancesofP5from（P2orP8）andfromMax[S]arebelow16.ThisisshowninFig.2fortheimage‘LENA’.Thisfactisusedtofurtherreduceunnessaryreplacements,therebyreducingthebluringoftheimages.ThustakingintoconsidertionofFigure（3）canbefurthermodifiedas

Output（X,Y）=Max（P2,P8）ifP5-Max[S]≥16

Max（P2,P8）ifMin[s]-P5≥16

P5otherwise（4）

Equation4indicatesthatreplacingactiontakesplaceonlywhenthedistancebetweenP5andMin[S]orMax[S]isnosmallerthan16.Thisstrtegythusavoidsthenecessaryreplacementsandreducesblurringoftheimages.Moreover，itcanbeimplementedusingsimplecomparatorsandsubtractors.

Grayleveldistances

Fig.2.Grayleveldistancesbetweencentralpointanditsneighboringpointsfortheimage‘LENNA’

Ⅲ.RESULTS

Figure3showstheoriginalimage‘LENNA’andFigure4showsthesameimagewhencorruptedwithimpulsenoise.ResultsofmedianfilterandtheproposedfilteraregiveninFigures5and6,respectively.ComparingFigures5and6,itisobservedthattheresultoftheproposedfilterismuchbetterthanthoseobtainedusingthemedianfilter.Aithough,themedianfilterremovetheimpulsivemoiseeffectively,however,theimagegetsblurred.Theproposedfilterremovestheimpulsivenoiseandalsopreservesthedetailsoftheimage.

Amultishellfilteremployingthemodifiedreplacementstrategyispresentdeinthispaper.Themodifiedfiltereffectivelysuppressestheinpulsemoise.Itusesthresholdconditionsthatrequirefewercomparisonsandreplacementsandisfasterascomparedtotheothermultishellmedianfilters.moreover,itcanberealizedusingsimplecomparators