細(xì)胞圖像主動(dòng)輪廓分割模型研究.pdf

1、浙江大學(xué)博士學(xué)位論文細(xì)胞圖像主動(dòng)輪廓分割模型研究姓名：馬竟鋒申請學(xué)位級(jí)別：博士專業(yè)：計(jì)算機(jī)科學(xué)與技術(shù)指導(dǎo)教師：陳純；包尚聯(lián)20110420浙江大學(xué)博士學(xué)位論文Abs仃actAbstractCellimageprocessingisplayingamoreandmoreimportantroleintheclinicaldiagnosisandtherapywiththedevelopmentofmedicalimagingdevices

2、觴wellascomputerscienceImagesegmentationasabasicdementincellimageprocessingisthekeytothewholeapplication，andthequalityofitdirectlyinfluencestheafterwardsActivecontourmodelsandtheirlevelsetmethodimplementationshavecaughtth

3、eresearchers’muchattentioninpastdecadesbecauseoftheiroutstandingperformanceinimagesegmentationTheyarethemostactivemodelsamongallsegmentationmethodssincetheyhavemanymeritssuchas：theedgeoftheresultissubpixellevel，吐leycallb

4、eeasilyimplementedinanenergyminimizationfiameworkandinthisframeworkmanypriorknowledgesaboutshapesandintensifiescanbeeasilyincludedInthisdissertation，areabasedactivecontourmodelsaleconstructedforcellimagesegmentationaccor

5、dingtothecharacteristicsofthethemAndthemoddsutilizednotonlytheconstructioninformationbutalsotheedgegradientinformationoftheimagesTherealemainlythreepartswithnewdevelopmentinthisdissertation：1Formulatedanewenergyfunctionc

6、ombiningthecharacteristicsofcellimageandtheevolutionprocessoflevelsetfunction，whichenablesusingonlyonelevelsetfunctiontosegmentthreepartsofthecellimage：nuelcolus，cytoplasmandbackground2Insomecases，thedifferencesofintensi

7、tiesbetweenobjectsandbackgroundsalesmall，whichmakesthesegmentationverydifficultInthisdissertation，anewenergyfunctionwasformulatedutilizingnotonlytheinformationofintensitybutalsothegradientinformationofthesameimage，whichm

8、adethesegmentationmoreeasily3IntroducedbilateralfilteringtechniqueintothecellimagesegmentationmodeltomakethesegmentationresultsmoreaccuratebecausebilateralfilteringCanachievekeepingedgeinformationandsmoothingtheimageatth