|
|
自动化学报 2009
基于主动轮廓模型的肺纹理自动提取新方法DOI: 10.3724/SP.J.1004.2009.00345, PP. 345-349 Keywords: Activecontourmodel,levelset,localbinaryfittingenergy,dynamicparametermodulation,3Dsurfacereconstruction Abstract: ？Computedtomography(CT)istheprimaryimagingmodalityforinvestigationoflungfunctionandlungdiseases.HighresolutionCTsliceimagesofchestcontainlotsoftextureinformation,whichprovidespowerfuldatasetstoresearchcomputeraid-diagnosis(CAD)system.Buttheextractionoflungtissuetexturesisachallengetask.Inthispaper,weintroduceanovelmethodbasedonlevelsettoextractlungtissuetexturetree,whichisautomaticandeffectual.Firstly,weproposeanimprovedimplicitactivecontourmodeldrivenbylocalbinaryfittingenergy,andtheparametersaredynamicandmodulatedbyimagegradientinformation.Secondly,anewtechniqueofpaintingbackgroundbasedonintensitynonlinearmappingisbroughtforwardtoremovetheinfluenceofbackgroundduringtheevolutionofsinglelevelsetfunction.Atlast,anumberofcontrastexperimentsareperformed,andtheresultsof3Dsurfacereconstructionshowourmethodisefficientandpowerfulforthesegmentationoffinelungtreetexturestructures.
|
