资源描述:
《A NOVEL GRAPH CUTS BASED LIVER SEGMENTATION METHOD》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、2010InternationalConferenceofMedicalImageAnalysisandClinicalApplication(MIACA)ANOVELGRAPHCUTSBASEDLIVERSEGMENTATIONMETHODHongzheYang,YongtianWang,JianYang,YueLiuSchoolofOptoelectronics,BeijingInstituteofTechnology,Beijing100081,Chinasegmentation.ItattractsmoreattentionforitsAB
2、STRACTuser-interactiveandmoreaccuratethansimplyautomaticsegmentation.Blakeetc.(Grab-cuts)[12]mixedGaussianThispaperpresentsanovelliversegmentationmethodmodelbasedonclueswithGraphcutstoimprovethebasedonthefastmarchingandgraphcutsmethods.Theinteractionregion.Lietc.(Lazysnapping)
3、viewregionasaalgorithmiscomposedofthreemainsteps:first,roughedgenodeusingwatershed,andapplythesuper-pixelmethodofoftheliverisextractedfromtheCTimagebyfastmarchingcuttingandsoon.method.Second,hardconstrainoftheforegroundandAlthoughthesemethodsarepowerfulandeffectiveinbackground
4、whichisusedforinitialcalculationofgraphsegmentation,theystillhavesomeshortcomings,andfailincutisobtainedbymathematicalmorphologymethod.Third,somecases:(1)sincethestoppingtermofthedeformationbasedontheformercalculation,thegraphcutsareutilizedevolutiondependsontheimagegradientfl
5、owbeingtorefinethesegmentationboundaryoftheliver.Theapproximatelyzero,thisoftenforcesthecontourstostopdevelopedmethodgreatlyreducesthecomplexityoftheseveralpixelsawayfromdesiredboundary.Thus,theactivecommonlyusedgraphcutsmethods,whichcanobtainthecontoursometimesdoesnotmatchthe
6、boundaryofthehardconstrainsautomatically.Also,thedevelopedmethodstructureaccurately;especiallyinregionssteepcurvaturereducesthedependenceofempiricalparametersofthefastandlowgradientvalues.(2)Thechoiceofelasticparametersmarchingbasedmethod.Experimentalresultsshowthattheandenerg
7、yfunctionminimizationwithnumbersofminimumdevelopedmethodisveryeffectiveforthesegmentationof(regionallocalminimum)areseriouslyaffectingtheliverfromCTimages.accuracyofsegmentationresults.(3)Theregion-basedsegmentationdetectsedgesbyprocessingeachpixelvalue.Soresultsareusuallyaccu
8、rate,butcomputationalcostisgreatatthesametime.(4)Currently,mo