欢迎来到天天文库
浏览记录
ID:37350501
大小:5.32 MB
页数:89页
时间:2019-05-22
《真实感三维人脸合成方法研究》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、中国科学院计算技术研究所博士学位论文真实感三维人脸合成方法研究姓名:姜大龙申请学位级别:博士专业:计算机应用技术指导教师:高文20050124AbstractRealistic3DFacialSynthesisJIANGDalong(ComputerApplicationTechnology)Directedby:GAOWen(Professor)Withthequickdevelopmentofmultimediaandnetwork,thehigherqualityhumanlikehuman—computerinterfacesarerequired.A
2、soneofthemostimportantpartsofvirtualhumansynthesistechnology,3Dhumanfacesynthesistechniquehasbeengreatlydevelopedintherecent30years.ItcanbeusedwidelyintheareaofvirtualrealityenvironmentsuchaSvirtualpresenter,netmeeting,videophone,remoteeducation,remotemedicaldiagnosis,movieeffectsan
3、dgamesetc.Researchon3DfacesynthesishaSthegreatvalueinboththeoryandapplication.Thisdissertationresearchedsomekeyproblemsinhumanfacesynthesis,suchas3Dfacemodeling,MPEG-4basedfacialanimation,facialanimationretargeting,animationwithsubtletexture,animationdrivingetc.Thefollowingarethemai
4、nachievementsinthisdissertation:1)Inthefieldof3Dhumanfacemodeling,themajorproblemishowtOconstruct3Dfacemodelscorrectly,eaSilyandquickly.Inthisdissertation,anefficient2D—to一3Dintegratedfacereconstructionapproachisintroducedtoreconstructapersonalized3Dfacemodelfromonlyonesinglefrontal
5、faceimage.Thisapproachisfullyautomaticandthespeedisfast.Itisaboutfifteentimesfasterthanthesimilarmethods.ThistechniqueCanalsobeusedforthefacerecognitiontoexpandthetrainingdata.Theaccuracyoffacerecognitionissignificantlyimproved,whilethealgorithmsoffacerecognitionarethesanle.2)MPEG一4
6、based3Dfacialanimationframeworkhastheadvantagesofcurrencyandfewdata.ThisdissertationpresentsamethodtoconstructFATs(facialanimationtables)forspecialhumanface.TheFATsCanbeusedtodrivethestandardfacemodeldefinedbythisdissertationandthemodels,whichhavethesametopologywiththestandardmodel.
7、3)Animationretargetingbecomesahotresearchtopicofcomputeranimationintherecentseveralyears.UsingtheFATsforthestandardmodel,thisdissertationpresentsanapproachtoconstructtheFATsforanytopology3Dfacemodels.Themethodsolvestheproblemofanimationdatareusingforanyfacemodels.Furthermore,itishig
8、hlyautomaticandthes
此文档下载收益归作者所有