文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > IT/计算机 > Multi-level Partition of Unity Implicits图形学
侵权投诉

Multi-level Partition of Unity Implicits图形学

multi-level 思想在图形分割的应用的应用

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Figure6:Left:EyefromStanford’sreconstructionofMichalangelo’sDavid(scannedat1mmresolution).Right:TheeyeisreconstructedasanMPUimplicitwithrelativeaccuracyε0=10 4.

Wealsotreatseparatelycornersofdegreefour:test(14)isap-pliedtothenormalsofthethirdclusterandif(14)issatis ed,theclusterissubdividedintotwopieces.Iftheresultingfourclustersofnormalscorrespondtoeitheraconvexorconcavecorner,itisreconstructedviaBooleanoperations.Otherwise,wegoto(a)andageneralquadric(8)is ttedtoP .

Ifthenumberofpointsusedtoestimatethecoef cientsofbi-variatequadraticpolynomial(12)islessthensix,wesetalltheunknowncoef cientsin(12)equaltozero.

Giventheaboveapproach,morecomplextypesofsharpfea-tures(forexample,asaddlecornerofdegree4)areapproximatedbysmoothfunctions.Noticehoweverthat“generic”sharpfeaturesareobtainedfromtheintersectionsoftwoorthreesurfaces,andthereforeconsistofedgesandcornersofatmostdegree3.Itisnotagenericeventforfoursmoothsurfacestointersectatonepoint.

Figure7:VisualizationofthefandiskmodelimplicitizedwithMPU.Topleft:Bloo-menthalpolygonizationwasused;inspiteofsuf cientlyhighpolygonizationresolu-tion(200Ktriangles)onecannoticesmallaliasingdefectsalongsharpfeatures.Topmiddle:alowresolutionpolygonizationisapplied.Topright:anoptimizedmesh(17Ktriangles)isobtainedfromthelowresolutionmesh,thesharpfeaturesareaccuratelyreconstructed.Bottomleft:amagni edpartofthehighresolutionmesh.Bottommid-dle:thesamepartoftheoptimizedmesh.Bottomright:apartofthemodelrenderedusingHartspheretracing.

6Results&Applications

Inthissectionwediscussresultsandapplicationsofapproximat-ingorinterpolatingMPUimplicitsforsurfacereconstructionfromrangescansandincompletepointdata,andfunction-basedopera-tions.

ApproximationandInterpolation.MostoftheillustrationsinthisworkhavebeengeneratedusingapproximatingMPUimplicitsasdescribedinSection4.However,ourMPUapproachcanalsobeadaptedtoexactdatapointinterpolationifweusealocalinter-polationmethodsuchasFrankeandNielson’ssingularweights(4)insteadof(3).

SinceMPUwith(4)requiresadeeperoctree-basedsubdivision(everynonemptysubdivisioncellcontainsonlyonepointofP),ourinterpolationprocedurerequiresmorememoryresourcesthantheapproximationone.Forinterpolationwithsingularweights(4)thesubdivisionstopsonlywhenallofpointsofPhavebeenplacedintheirowncells.Theballaroundanonemptyoctreecelliscen-teredattheinterpolatedpointinsidethecell.Wesetα=1.25in(5)inordertoensurethatwecovertheboundingboxbytheballsaroundtheoctreecells.

ForeachcellcontainingoneinterpolatedpointpofP,itslocalshapefunctionQ(x)isde nedby(12),wheretheoriginofcoor-dinatesoflocalcoordinatesystem(u,v,w)islocatedatp(henceF=0in(12))andthepositivedirectionofwcoincideswiththedi-rectionoftheaveragednormalatp.Wedon’tusethenormalofNassignedtopbecauseofstabilityproblemscommonforHermite-likeinterpolationschemes.Theunknowncoef cientsaredeter-minedbyminimizingquadraticenergy(13)withc =p. Now(1)interpolatesPbecausepartitionofunityfunctions i(·)de nedby(2)satisfy

1ifi=j

and i(pj)=0. i(pj)=δij=

0ifi=jTherightimageofFigure8showsresultsofapplyingtheMPUinterpolationandshapemodelingoperations(Booleansubtraction,twisting)totheStanfordBuddhamodel.

Reconstructionfromincompletedata.ReconstructionfromscatteredpointdatawithMPUimplicitsisrobustwithrespecttovariationsofpointdensity,asdemonstratedinFigure9.

Reconstructionfromrangescans.MPUimplicitscanbeusedtoreconstruct3Dmodelsfromacollectionofrangescans.Wehave

5Visualization

Conventionaltechniquesforvisualizingimplicitmodelsincludepolygonization(isosurfaceextraction),raytracing,andvolumeren-dering.Fromamongthesevariousvisualizationmethods,weuseBloomenthal’spolygonizer[Bloomenthal1994]becauseofitsnicecontinuationpropertiesandtheHartspheretracingmethod[Hart1996].ThesetwomethodsbothworkwellusingtheapproximatedistancefunctionsofMPUimplicits.

Ifourgoalistocreateapolygonalmesh,wecansavetimeandmemorybycomputingMPUapproximationsonthe yduringthepolygonizationprocess.Wehavefoundthatanapproximationac-curacyofε0=10 4(thatis,0.01%ofthelengthofthemaindiag-onaloftheboundingbox)isquitesuf cientforthereconstructionof nefeatures,asdemonstratedbyFigure6.

Ifanon-adaptivesurfaceextractionroutineisusedwithanim-plicitmodelthathassharpfeatures,a nesamplingdensityisre-quiredtocapturethesefeatures.AnexampleofthiscanbeseeninthetopandbottomleftimagesofFigure7.Analternativeistouseadaptivesamplingandremeshingstrategiessuchasthosein[Kobbeltetal.2001;OhtakeandBelyaev2002;Juetal.2002].We nditattractivetocombinealowresolutionBloomenthalpolygo-nizationwithapostprocessingmeshoptimizationtechniquedevel-opedin[OhtakeandBelyaev2002],asshowninthetopmiddle,topright,andbottommiddleimagesofFigure7.

Evenhigherqualityrenderingcanbeachievedusingraytrac-ingtechniques.ThespheretracingmethodofHart[1996]workswelltogetherwithMPUimplicitssinceitusesthedistancefunctionrepresentationanditisquitecapableofrenderingimplicitmodelswithsharpfeatures.ThebottomrightimageofFigure7showsanMPUimplicitmodelofthefandiskmodelthatwasrenderedwithspheretracing.TheleftimageofFigure8showsspheretracingofamorecompleximplicitmodel.Thismodelwasgeneratedfromafunction-basedoperation(subtraction)appliedtothedragonandtheDavidmodel,bothrepresentedasMPUimplicits.Noticehowwellthesharpfeaturesarereconstructedand

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

rendered.

上一页第5页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email