数字图像处理英文文献翻译参考.docx
- 文档编号:29321924
- 上传时间:2023-07-22
- 格式:DOCX
- 页数:31
- 大小:688.75KB
数字图像处理英文文献翻译参考.docx
《数字图像处理英文文献翻译参考.docx》由会员分享,可在线阅读,更多相关《数字图像处理英文文献翻译参考.docx(31页珍藏版)》请在冰豆网上搜索。
数字图像处理英文文献翻译参考
HybridGeneticAlgorithmBasedImageEnhancement
Technology
Abstract:
Inimageenhancement,TubbsproposedanormalizedincompleteBetafunctiontorepresentseveralkindsofcommonlyusednon-lineartransformfunctionstodotheresearchonimageenhancement.ButhowtodefinethecoefficientsoftheBetafunctionisstillaproblem.WeproposedaHybridGeneticAlgorithmwhichcombinestheDifferentialEvolutiontotheGeneticAlgorithmintheimageenhancementprocessandutilizethequicklysearchingabilityofthealgorithmtocarryouttheadaptivemutationandsearches.FinallyweusetheSimulationexperimenttoprovetheeffectivenessofthemethod.
Keywords:
Imageenhancement;HybridGeneticAlgorithm;adaptiveenhancement
I.INTRODUCTION
Intheimageformation,transferorconversionprocess,duetootherobjectivefactorssuchassystemnoise,inadequateorexcessiveexposure,relativemotionandsotheimpactwillgettheimageoftenadifferencebetweentheoriginalimage(referredtoasdegradedordegraded)Degradedimageisusuallyblurredoraftertheextractionofinformationthroughthemachinetoreduceorevenwrong,itmusttakesomemeasuresforitsimprovement.
Imageenhancementtechnologyisproposedinthissense,andthepurposeistoimprovetheimagequality.FuzzyImageEnhancementsituationaccordingtotheimageusingavarietyofspecialtechnicalhighlightssomeoftheinformationintheimage,reduceoreliminatetheirrelevantinformation,toemphasizetheimageofthewholeorthepurposeoflocalfeatures.Imageenhancementmethodisstillnounifiedtheory,imageenhancementtechniquescanbedividedintothreecategories:
pointoperations,andspatialfrequencyenhancementmethodsEnhancementAct.Thispaperpresentsanautomaticadjustmentaccordingtotheimagecharacteristicsofadaptiveimageenhancementmethodthatcalledhybridgeneticalgorithm.Itcombinesthedifferentialevolutionalgorithmofadaptivesearchcapabilities,automaticallydeterminesthetransformationfunctionoftheparametervaluesinordertoachieveadaptiveimageenhancement.
II.IMAGEENHANCEMENTTECHNOLOGY
Imageenhancementreferstosomefeaturesoftheimage,suchascontour,contrast,emphasisorhighlightedges,etc.,inordertofacilitatedetectionorfurtheranalysisandprocessing.Enhancementswillnotincreasetheinformationintheimagedata,butwillchoosetheappropriatefeaturesoftheexpansionofdynamicrange,makingthesefeaturesmoreeasilydetectedoridentified,forthedetectionandtreatmentfollow-upanalysisandlayagoodfoundation.
Imageenhancementmethodconsistsofpointoperations,spatialfiltering,andfrequencydomainfilteringcategories.Pointoperations,includingcontraststretching,histogrammodeling,andlimitingnoiseandimagesubtractiontechniques.Spatialfilterincludinglow-passfiltering,medianfiltering,highpassfilter(imagesharpening).Frequencyfilterincludinghomomorphismfiltering,multi-scalemulti-resolutionimageenhancementapplied[1].
III.DIFFERENTIALEVOLUTIONALGORITHM
DifferentialEvolution(DE)wasfirstproposedbyPriceandStorn,andwithotherevolutionaryalgorithmsarecompared,DEalgorithmhasastrongspatialsearchcapability,andeasytoimplement,easytounderstand.DEalgorithmisanovelsearchalgorithm,itisfirstinthesearchspacerandomlygeneratestheinitialpopulationandthencalculatethedifferencebetweenanytwomembersofthevector,andthedifferenceisaddedtothethirdmemberofthevector,bywhichMethodtoformanewindividual.Ifyoufindthatthefitnessofnewindividualmembersbetterthantheoriginal,thenreplacetheoriginalwiththeformationofindividualself.
TheoperationofDEisthesameasgeneticalgorithm,anditconcludemutation,crossoverandselection,butthemethodsaredifferent.WesupposethatthegroupsizeisP,thevectordimensionisD,andwecanexpresstheobjectvectoras
(1):
xi=[xi1,xi2,…,xiD](i=1,…,P)
(1)
Andthemutationvectorcanbeexpressedas
(2):
i=1,...,P
(2)
arethreerandomlyselectedindividualsfromgroup,andr1
r2
r3
i.Fisarangeof[0,2]betweentheactualtypeconstantfactordifferencevectorisusedtocontroltheinfluence,commonlyreferredtoasscalingfactor.Clearlythedifferencebetweenthevectorandthesmallerthedisturbancealsosmaller,whichmeansthatifgroupsclosetotheoptimumvalue,thedisturbancewillbeautomaticallyreduced.
DEalgorithmselectionoperationisa"greedy"selectionmode,ifandonlyifthenewvectoruithefitnessoftheindividualthanthetargetvectorisbetterwhentheindividualxi,uiwillberetainedtothenextgroup.Otherwise,thetargetvectorxiindividualsremainintheoriginalgroup,onceagainasthenextgenerationoftheparentvector.
IV.HYBRIDGAFORIMAGEENHANCEMENTIMAGE
enhancementisthefoundationtogetthefastobjectdetection,soitisnecessarytofindreal-timeandgoodperformancealgorithm.Forthepracticalrequirementsofdifferentsystems,manyalgorithmsneedtodeterminetheparametersandartificialthresholds.Canuseanon-completeBetafunction,itcancompletelycoverthetypicalimageenhancementtransformtype,buttodeterminetheBetafunctionparametersarestillmanyproblemstobesolved.ThissectionpresentsaBetafunction,sinceaccordingtotheapplicablemethodforimageenhancement,adaptiveHybridgeneticalgorithmsearchcapabilities,automaticallydeterminesthetransformationfunctionoftheparametervaluesinordertoachieveadaptiveimageenhancement.
Thepurposeofimageenhancementistoimproveimagequality,whicharemoreprominentfeaturesofthespecifiedrestorethedegradedimagedetailsandsoon.Inthedegradedimageinacommonfeatureisthecontrastlowersideusuallypresentsbright,dimorgrayconcentrated.Low-contrastdegradedimagecanbestretchedtoachieveadynamichistogramenhancement,suchasgraylevelchange.WeuseIxytoillustratethegraylevelofpoint(x,y)whichcanbeexpressedby(3).
Ixy=f(x,y)(3)
where:
“f”isalinearornonlinearfunction.Ingeneral,grayimagehavefournonlineartranslations[6][7]thatcanbeshownasFigure1.WeuseanormalizedincompleteBetafunctiontoautomaticallyfitthe4categoriesofimageenhancementtransformationcurve.Itdefinesin(4):
(4)where:
(5)
Fordifferentvalueofαandβ,wecangetresponsecurvefrom(4)and(5).
ThehybridGAcanmakeuseoftheprevioussectionadaptivedifferentialevolutionalgorithmtosearchforthebestfunctiontodetermineavalueofBeta,andtheneachpixelgrayscalevaluesintotheBetafunction,thecorrespondingtransformationofFigure1,resultinginidealimageenhancement.Thedetaildescriptionisfollows:
Assumingtheoriginalimagepixel(x,y)ofthepixelgraylevelbytheformula(4),denotedby
hereΩistheimagedomain.EnhancedimageisdenotedbyIxy.Firstly,theimagegrayvaluenormalizedinto[0,1]by(6).
(6)
where:
and
expressthemaximumandminimumofimagegrayrelatively.
Definethenonlineartransformationfunctionf(u)(0≤u≤1)totransformsourceimagetoGxy=f(
),wherethe0≤Gxy≤1.
Finally,weusethehybridgeneticalgorithmtodeterminetheappropriateBetafunctionf(u)theoptimalparametersαandβ.WillenhancetheimageGxytransformedantinormalized.
V.EXPERIMENTANDANALYSIS
Inthesimulation,weusedtwodifferenttypesofgray-scaleimagesdegraded;theprogramperformed50times,populationsizesof30,evolved600times.Theresultsshowthattheproposedmethodcanveryeffectivelyenhancethedifferenttypesofdegradedimage.
Figure2,thesizeoftheoriginalimagea320×320,it'sthecontrasttolow,andsomedetailsofthemoreobscure,inparticular,scarvesandotherdetailsofthetextureisnotobvious,visualeffects,poor,usingthemethodproposedinthissection,toovercometheabovesomeoftheissuesandgetsatisfactoryimageresults,asshowninFigure5(b)shows,thevisualeffectshavebeenwellimproved.Fromthehistogramview,thescopeofthedistributionofimageintensityismoreuniform,andthedistributionoflightanddarkgrayareaismorereasonable.Hybridgeneticalgorithmtoautomaticallyidentifythenonlineartransformationofthefunctioncurve,andthevaluesobtainedbefore9.837,5.7912,fromthecurvecanbedrawn,itisconsistentwithFigure3,c-class,thatstretchacrossthemiddleregioncompressiontransformtheregion,whichwereconsistentwiththehistogram,theoveralloriginalimagelowcontrast,compressionatbothendsofthemiddleregionstretchingregionisconsistentwithhumanvisualsense,enhancedtheeffectofsignificantlyimproved.
Figure3,thesizeoftheoriginalimagea320×256,theoverallintensityislow,theuseofthemethodproposedinthissectionaretheimagesb,wecanseetheground,chairsandclothesandotherdetailsoftheresolutionandcontrastthantheoriginalimagehasImprovedsignificantly,theoriginalimagegraydistributionconcentratedinthelowerregion,andtheenhancedimageofthegrayuniform,graybeforeandaftertransformationandnonlineartransformationofbasicgraph3(a)thesameclass,namely,theimageDimregionstretching,andthevalueswere5.9409,9.5704,nonlineartransformationofimagesdegradedtypeinferenceiscorrect,theenhancedvisualeffectandgoodrobustnessenhancement.
Difficulttoassessthequalityofimageenhancement,imageisstillnocommonevaluationcriteria,commonpeaksignaltonoiseratio(PSNR)evaluationintermsofline,butthepeaksignaltonoiseratiodoesnotreflectthehumanvisualsystemerror.Therefore,weusemarginalprotectionindexandcontrastincreaseindextoevaluate
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 数字图像 处理 英文 文献 翻译 参考