基于图像灰度的三维重构递推算法的研究

《基于图像灰度的三维重构递推算法的研究》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、大连理工大学硕士学位论文基于图像灰度的三维重构递推算法的研究姓名：满臣申请学位级别：硕士专业：机械电子工程指导教师：李剑忠20051201满臣：基于图像灰度的三维重构递推算法的研究ResearchontheCumulationAlgorithmof3DReconstructionFromGray—levelImageAbstractAsakeyresearchdirectionofcomputervision-threedimensionpasswereconstructionbecomesmoreandmoreimpoaantsoasnotto

2、beignoredinthevariousareasofoursocietyandhasbeenwidelyusedinindustry，agriculture，countrydefense，medicine，spacetechnology，etc，especially．Duetoitsspecialfeature，whichcanrestore3dshapeinformationjustfromonegray—levelimage，recently，shapefromshading(SFS)getsgreatimprovedanddevelop

3、srapidlyinmethodologyandapplication，andbecomesanewresearchfocusof3dpassivereconstmction．OnthebaseofcomprehensivesummarizationandanalysisofcurrentSFSresearchwork，a3dreconstructiontechnologyfromgray—levelimagehasbeenproposedinthispaper．Withthequadranglecomposedoffournearestpixe

4、lsastheleastelement，asetofnonlinearequationscanbeobtainedfromtherelationshipbetweenthenormalandgrayoffourtriangularplanes；then，therelativelocationoffourcorrespondingpointofthesefourpixelscallbeacquiredfromthesolutionofthesenonlinearequationswiththenonlinearleastsquaremethod．B

5、ythisway，theprecisionoftheresultof3dreconstructionismainlydeterminedbytheresolutionofthisimageafteralltherelativepixels’locationsofthisobjectsurfacearefiguredoutthrou【ghacumulationalgorithm．Becausethisnewalgorithmdon’tadoptrigorassumptionsthatdon’tcorrespondtotherealnaturalco

6、nditionandunsuitableassumptionsthatsomepriorconditionhasknownasmosttraditionalSFSalgorithmsdo，itcangetmorepreciseanduniversal3Dreconstructionresult．Meanwhile，anewrenderingalgorithm-reversecalculationalgorithm-hasalsobeenpresentedbythecontrastbetweenthecurrentpopularrenderingm

7、ethod—GBR、IBR、PBRinthispaper，throulghwhichthecolorofeachpixelcanbefiguredoutunderthelightingmodelafterthecoordinatesandnormal，ofeverycorrespondingpointonthe3dobjectsurfacehavebeencalculatedfromthepixelsintherangewherethe3dobjectisprojectedontotheimagingplane，inordertoprovidee

8、achpixelarealcolorandgetanimagewiththestrongestfeelingofreality．Byth