图像分割的活动轮廓模型研究-计算机应用技术专业论文.docx

Research onActiveContourModelBased onImageSegmentationAbstractImagesegmentationisthetechniquethatdividesimageintosimilarfeatureregions, whichisthebasisforadvancedimageprocessing,suchasimageunderstandingand recognition.ImagesegmentationmethodbasedontheC-Vactivecontourmodelcannot onlyrelyonadequateandreliablelow-levelimagefeatures,butalsorelyonhigh-level knowledge.Comparedwiththetraditionalimagesegmentationmethod,theC-Vactive contourmodelhasstrongadvantages.However,themodelmakeswelluseoftheregion information,soimagescontainingtheheterogeneousandcomplexbackgroundcannotbe segmentedwell.Inaddition,inpracticalapplication,forthecomputingcomplexityofthe C-Vmodel,thereal-timeproblemisalsoneededtotakeintoaccount.Hence,further researchesontheC-Vactivecontourmodelaremuchnecessaryinboththeoreticaland applicationaspects.IntensivestudiesarecarriedoutinthisdissertationonthekeytechniquesoftheC-V activecontourmodel,whichincludemoreimageinformationconsideredinthemodeland regiondivision.Thecontributions ofthis dissertationareas follows:Firstly,theclassicalgeodesicactivecontour(GAC)modeltendstoproducefake edgesinsmoothregion,whiletheC-Vmodelcannoteffectivelydetectimageswithholes andobtainthepreciseboundary.Toaddresstheaboveissues,thisdissertationproposesan adaptivemixturemodelsynthesizingtheGACmodelandtheC-Vmodelbyaweight function.Accordingtoimagecharacteristics,theproposedmodelcanadaptivelyadjustthe weightfunction.Bythisway,themodeldevelopstheadvantageoftheGACmodelinthe regionwithrichtexturesoredges,whileexploitstheadvantageoftheC-Vmodelin smoothlocalregion.Moreover,theproposedmodel isextendedtothevector-valued images.Byexperiments,itisverifiedthattheproposedmodelobtainsbetterresultsthan thetraditionalmodels.AbstractResearchonActiveContourModelBasedonImageSegmentationSecondly,animprovedvariationallevelsetforimagesegmentationframeworkthat combinestheedgeinformationandregioninformationisproposed.Theframeworkis implementedontheC-Vactivecontour,whichisatypicalregion-basedactivecontour mode