A Local region-based Chan-Vese modal for image segmentation

《A Local region-based Chan-Vese modal for image segmentation》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、PatternRecognition45(2012)2769–2779ContentslistsavailableatSciVerseScienceDirectPatternRecognitionjournalhomepage:www.elsevier.com/locate/prAlocalregion-basedChan–Vesemodelforimagesegmentationa,ba,nShigangLiu,YaliPengaSchoolofComputerScience,ShaanxiNo

2、rmalUniversity,Xi’an710062,ChinabSchoolofElectronicsandInformationEngineering,Xi’anJiaotongUniversity,Xi’an710049,ChinaarticleinfoabstractArticlehistory:Inthispaper,anewregion-basedactivecontourmodel,namelylocalregion-basedChan–Vese(LRCV)Received16Mar

3、ch2011model,isproposedforimagesegmentation.Byconsideringtheimagelocalcharacteristics,theReceivedinrevisedformproposedmodelcaneffectivelyandefficientlysegmentimageswithintensityinhomogeneity.To18August2011reducethedependencyonmanualinitializationinmanya

4、ctivecontourmodelsandforanautomaticAccepted26November2011segmentation,adegradedCVmodelisproposed,whosesegmentationresultcanbetakenastheinitialAvailableonline10January2012contouroftheLRCVmodel.Inaddition,weregularizethelevelsetfunctionbyusingGaussianfil

5、teringKeywords:tokeepitsmoothintheevolutionprocess.ExperimentalresultsonsyntheticandrealimagesshowtheActivecontourmodeladvantagesofourmethodintermsofbotheffectivenessandrobustness.Comparedwiththewell-knowImagesegmentationlocalbinaryfitting(LBF)model,ou

6、rmethodismuchmorecomputationallyefficientandmuchlessLevelsetsensitivetotheinitialcontour.&2012ElsevierLtd.Allrightsreserved.1.Introductionsimultaneously,andmultipleinitialcontourscanbeplaced.ThisflexibilityandconvenienceprovideameansforanautomaticImages

7、egmentationistheprocessofdividingimagesintosegmentationbyusingapredefinedsetofinitialcontours.meaningfulsubsetsthatcorrespondtosurfacesorobjects.ItisaWithoutlossofgenerality,mostoftheACMsstudiedunderthefundamentalprobleminthefieldofcomputervision,becaus

8、elevelsetframeworkcanbecategorizedintotwotypes:edge-basedrecognitionandreconstructionoftenrelyonthisinformation[8–10]andregion-based[11–18]ones.Theedge-basedmodels[1,2].Tosolvetheproblem,manyresearchershavedonegreatutilizeimagegradienttoconstr