基于流体力学的图像修复

《基于流体力学的图像修复》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、安徽大学硕士毕业论文基于流体力学的图像修复何信息，所以也不能满足“连接性准则”，因此我们对该模型进行了改进，在修复方程中添加了曲率驱动的扩散项，它使改进后的修复方法能够满足“连接性准则”，最后通过模拟实验证明该改进方法的有效性。关键词：图像修复、偏微分方程、流体力学、Navier-Stokes、TV-StokesIIAbstractAbstractImageinpaintingistheprocessoffillinginmissingdatainadesignatedregionofastillorvideoimage.Thegoalistoproducearevise

2、dimageinwhichtheinpaintedregionisseamlesslymergedintotheimageinawaythatisnotdetectablebyatypicalviewer.Imageinpaintingisnotonlyanimportantpartofimagerestorationresearch,butalsoahotspotinthefieldofImageProcessingandComputerVisionresearch.Thepaperistostudythedigitalimageinpainting,thatis,we

3、havetousethecomputertoinpaintthedamagedimage.Imagecanbesimplydividedintothestructureimageandthetextureimagetwocategories,soimageinpaintingcanbeaccordinglydividedintothestructureimageinpaintingandthetextureimageinpaintingtwocategories.Thebasicmethodoftextureimageinpaintingisthetexturesynth

4、esis,whichisaspecializedresearch.ThestructureimagemainlyusesPDE-basedimageinpaintingmodels.PDE-basedimageinpaintingmodel'smainidealistoestablishsetofpartialdifferentialequationsaccordingtoenergyminimizationmode1s,andconsideritasaboundaryproblemofpartialdifferentialequations,andthensolveit

5、bytimematchmethod.Becauseitcanaccuratelybuildmodelforimages,soithassolvedwellmanycomplexproblemsofimageinpainting.Inthepaper,wemainlystudyseveraltypesofPDE-basedimageinpaintingmodels:TotalVariation(TV)inpaintingmodel,CurvatureDrivenDiffusion(CDD)model，whichimprovesonTVmodel.Then,wediscuss

6、indetailabouttwonewPDE-basedimageinpaintingmodels,thatis,theimageinpaintingmodelusingaNavier-StokesequationandtheimageinpaintingmodelusingaTV-Stokesequation.Theycomparetheimageintensitytoafluid.TheimageinpaintingmodelusingaNavier-StokesequationimitatestheNavier-Stokesequationoffluiddynami

7、cs,andsetupanimageinpaintingalgorithm,whichsmoothlypropagatestheisophotelines.Theexperimentsprovethatithasgoodinpaintingeffectforsmall-scaledamagessuchasimagenicks,butthereissomedifficultyinusingthemethodandthemethodcannotrealizetheConnectivityIII安徽大学硕士毕业论文基于流体力学的图像