基于遞歸最短生成樹方法的壓縮域運(yùn)動(dòng)對象分割技術(shù)研究.pdf

1、上海大學(xué)碩士學(xué)位論文基于遞歸最短生成樹方法的壓縮域運(yùn)動(dòng)對象分割技術(shù)研究姓名：趙錕申請學(xué)位級別：碩士專業(yè)：通信與信息系統(tǒng)指導(dǎo)教師：張兆揚(yáng)20080101上海大學(xué)碩士學(xué)位論文基于遞歸最短生成樹方法的壓縮域運(yùn)動(dòng)對象分割技術(shù)研究ABSTRACTVideoobjectsegmentationhasshownitsbrightfutureofapplicationalongwiththedevelopmentofmultimediatechnolo

2、gyandtheappearanceofthesecondgenerationvideoencodingtechnologyDuetothefactthatmostofvideodataarememoriedincompresseddomainvideoobjectshavetobedecodedatfirstforsegmentation，anditisverytimeconsumingToavoidthisproblem，morea

3、ndmoreresearchersbegintofocusonsegmentatingvideoobjectsfromcompresseddomaindirectlyH264／AVC，issuedbyITUTinternationallyisthenewestvideocodingstandardItsexcellentcompressionperformanceandnetworkaffinitymakesitattractsmany

4、attentionsTheresearchonH264compresseddomainvideoobjectsegmentationlookssignificativeTherearetwochallangesinvideoobjectsegmentationFirstlyscenesintherealworldaretoocomplextobuildauniformtranscendentmodelScondlythesemantic

5、videoobject，akindofhighlevelfeatures，ishardtobedescribedaccuratelywithlow—levelfeaturessuchasmovingandtextureDuetothesechallangesabove，videoobjectsegmentationisoftenrestrictedintheboundsofmovingobjectsThethesisresearches

6、onH264compresseddomainmovingobjectsegmentationandtheMaincontributionsofthispaperareasfollows：FirstlythemotionvectorisunifiedandcumulatedwithabackwarditerativeprojectionprocesstobedenserSecondlyeJobalmotioncompensatationi

7、susedtoeliminatethe#obalmotionThirdlyamodifiedRSSTarithmeticisusedtomakeblocksclusteredExperimentalresultsforseveralvideosequencesdemonstratetheeffectivenessoftheproposedapproachKeywords：H264，CompressedDomain，MovingObjec