1、浙江大學(xué)理學(xué)院碩士學(xué)位論文基于局部相似性的輪廓提取算法姓名：施紅星申請學(xué)位級別：碩士專業(yè)：應(yīng)用數(shù)學(xué)指導(dǎo)教師：尹永成劉利剛20080501浙江大學(xué)碩l：學(xué)位論文AbstractImagesegmentationisanimportantresearchtopicindigitalimageprocessingObjectcutoutistoextractthesemanticobjectfiomtheimage，whichnotonlyd

2、ependsonlowlevelinformationofimage，butalsoneedshigh—levelinformationofuserinterfaceGoodalgorithmsofimagecutoutshouldextractthecontouroftheobjectswhatusersdesiretocutout，throughasmallnumberofuserinteractionaslessaspossibl

3、eInmostofthephotoeditingsoftware，usersarerequiredtofindthecontourbytrackingtheentirecontouroftheobjectItisverytedious，andneedslotsofuserinterface，buttheresultisnotalwayssatisfiedTherehavebeensomebetteralgorithmssuchasSna

4、kemodelandLazysnappingSnakemodelneedsacoarsecontourandLazysnappingonlyneedsomescribblestodistinguishbackgroundandforegroundHoweverSnakemodelcannothandlemanytypesofimagebecauseitneedsmanyparameterswhichishardtodedde，andhi

5、ghlydependsoninitialcontouLlazysnappingmightfailwhentheforegroundandbackgroundsharethesimilartextureWeproposeanovelimagecutoutalgorithmbasedonlocalsimilarityTheuserinterfaceisdifferentfromSnakeandLazysnappingTheusersonly

6、marksomecriticalpointsalongtheboundaryoftheobjectThenthewholecontouroftheobjectisextractedfromthesecriticalpointsbasedonthelocalsimilaritythroughanoptimizationprocessFirst，wewillintroducethealgorithmbasedonsimilarityofno

7、rmalvectorneighbor，andthenweimproveonsimilarityoflocalneighborWealsoimproveontheuserinterfaceofmarkingpointsalongboundarywhichletusersnothavetoaccuratelymarkpointsalongtheboundaryThatmeanstoallowsametinyerrorSOitmakesthe