資源描述：
《基于素描語(yǔ)義信息和超像素合并的圖像分割》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、萬(wàn)方數(shù)據(jù)AbstractImagesegmentationisafundamentalandchallengingresearchprobleminimageprocessingfield.Researchonsegmentationmethodshouldcombineboththecharacteristicsoftheimagedataitselfandthesubsequentapplicationofdividingresult.Inimagescontainingtargetwhichhasthefeatur
2、ethatthebodymarkingshavetwodifferentcolorsalternatelyrepeated,likezebrasandtigers,tosplitupthetargetasawholeisdifficult.Basedonthevisualcomputingtheory,thispapergetsthesketchmapusingtheinitialsketchmodel.Sketchsegmentscharacterizethepositionanddirectionofthesingu
3、larityintheimage.Inimageswearedealingwith,singularityinformationfallsintotwocategories,namely,traditionalboundariesandthebordersbetweenbandsandstripes.Inimagesegmentationfiled,thetraditionalboundariesshouldbereservedasthefinalsegmentboundaries,whiletheboundarybet
4、weenthebandsandstripesisduetodifferenceincolorofadjacentbands.Becauseoftheregularityinbandsandstripesofzebraandtiger,inthefinalsegmentationresult,differentwithtraditionalboundaries,boundarybetweenthebandscannotbethefinalsegmentationboundaries,soastosegmentthetarg
5、etasawhole.Therefore,thispaperbuildsgeometricblockswiththesegmentsthatcomposethesketchmapandthenmapsthegeometricblockstothecorrespondingpositionsoftheoriginalimageandextractstheco-occurrencematrixbasedonthegeometricblocks.Wetreattheco-occurrencematrixasthefeature
6、softhecorrespondinglinesandthendividethesketchsegmentsintobandandstripescategoryandgeneralboundarymarkingscategorybasedonthefeatures.Forthesuperpixelsgotfromtheover-segmentationmethod,wemergethemundertheguidanceofthesketch-classification-basedsemanticinformation.
7、Forthesuperpixelswhichbeguidedbythebandandstripescategory,wecountupthegrayvaluesofthemandthenmakeafurthersegmentationbasedonthegrayscalestatisticssymbioticrelationshipbetweeneachsuperpixelanditsneighbors.Thus,wegetthefinalsegmentationresult.Simulationresultsshowt
8、hattheproposedmethodcangetbettersegmentationresults.Thispaperalsoappliestheproposedmethodtothecolorimagesegmentation.Comparedwiththegray-scaleimage,thecolorima
