Mathematics and Visualization SeriesEditors GeraldFarin Hans-ChristianHege DavidHoffman ChristopherR.Johnson KonradPolthier MartinRumpf Forfurthervolumes: http://www.springer.com/series/4562 • David H. Laidlaw Anna Vilanova (cid:2) Editors New Developments in the Visualization and Processing of Tensor Fields With133Figures, 129in color 123 Editors DavidH.Laidlaw AnnaVilanova DepartmentofComputerScience DepartmentofBiomedicalEngineering BrownUniversity,Providence EinhovenUniversityofTechnology RhodeIsland,USA Einhoven,Netherlands ISBN978-3-642-27342-1 ISBN978-3-642-27343-8(eBook) DOI10.1007/978-3-642-27343-8 SpringerHeidelbergNewYorkDordrechtLondon LibraryofCongressControlNumber:2012947651 MathematicalSubjectClassification(2010):68-XX,70-XX,15-XX,76-XX,92-XX (cid:2)c Springer-VerlagBerlinHeidelberg2012 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof thematerialisconcerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation, broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionorinformation storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodology nowknownorhereafterdeveloped.Exemptedfromthislegalreservationarebriefexcerptsinconnection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’slocation,initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer. PermissionsforusemaybeobtainedthroughRightsLinkattheCopyrightClearanceCenter.Violations areliabletoprosecutionundertherespectiveCopyrightLaw. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. While the advice and information in this book are believed to be true and accurate at the date of publication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityfor anyerrorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,with respecttothematerialcontainedherein. Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Wededicatethisbooktoourfamilies. • Preface In this book, we attempt to capture the excitement and inspiration that has been generated during a series of Dagstuhl Seminars that have looked at visualization andprocessingoftensorfields.Thisbookincludescontributionsfromattendeesof thethirdmeetingheldinJuly2009.Asinthetwoearliervolumes,theauthorsreport onrecentresearchresultsaswellasopiningonfuturedirectionsfortheanalysisand visualizationoftensorfields.Topicsrangefromapplicationsoftheanalysisoftensor fieldstopurerresearchintotheirmathematicalandanalyticalproperties.Oneofthe goalsofthisseminarwastobringtogetherresearchersfromalongtheaxisbetween pure and applied research, identifying new multidisciplinary research challenges. Thisbook,wehope,willcontinuetofurtherthatgoalinabroadercontext. Themanuscriptisorganizedintosevenparts.Theparts,tosomedegree,increase indimensionalityandinmathematicalsophistication.Onechangeweimplemented in this third in a series of Dagstuhl seminars is that it no longer restricts the underlyingdata modelsto second-ordertensor fields. Research on representations thatgobeyondsecond-ordertensorsareclearlynecessary,asbecamemoreevident duringtheseminars. Part I, “Structure-TensorComputation,”focuses primarily on the generationof the so-called structure tensor to characterize image structure information. This information is often needed for feature preserving image processing techniques. Thechapterspresentgeneralizationsofthemethodologyofclassicstructuretensor estimationanditsapplicationbeyondscalarfields. PartII,“Tensor-FieldVisualization,”presentsthreeresultsthatextendourability togenerate,utilize,andvisualizetensorfields.Onelooksatfabriclikevisualizations on surfaces. The second uses a Lagrangian metaphor to show the structure of 3D tensor fields. The third presents methods for designing tensor fields that have differentapplicationsinvisualizationsuchasrealizingsampledtexturefields. Part III, “Applicationsof Tensor-Field Analysis and Visualization,” offers con- tributions from researchers in disciplines where the application of analysis and visualization of second-order tensor fields is relevant. These applied domains include the simulation of combustion and the measurement of material elasticity usingmagneticresonanceimaging. vii viii Preface PartIV,“DiffusionWeightedMRIVisualization,”breaksoutonescientificarea where second-ordertensor fields are often applied. The chaptersin this partshow techniquesforthespecificstudyofdiffusiontensorfieldsproducedfrommagnetic resonance imaging, typically of the brain. These tensor-valued images capture propertiesofthebrainwhitematterand,therefore,itsstructuralconnectivity.Brain connectivity is complex and difficult to understand. These chapters take a step in makingthatunderstandingeasier.Thelastchapterespeciallyfocusesonuncertainty visualizationoffeaturesderivedfromthetensorfield,whichisarecentlyemerging researchtopic. PartV,“BeyondSecond-OrderDiffusionTensorMRI,”expandsthecomplexity oftheunderlyingdatamodelusedintheearlierchapters.Despitethechallengesof working with second-ordertensor valued images, the use of second-ordertensors haslimitations.Indiffusionweightedimaging,itimpliesamodeloftheunderlying anatomy and physics that is not sufficiently accurate in some situations. In this part, models that are more accurate or provide more information are presented. In diffusion weighted imaging, they also require more complex acquisition and processing.Asthechaptersinthispartexemplify,someoftheresearchinthisfield focusesonlimitingtheextracomplexity. In Part VI, “Tensor Metrics,” some of the challengesof workingwith sampled tensorfieldsareaddressed.Tensorfieldsareintrinsicallydefinedoveracontinuous Euclideanspace, and yet they must be representedfor computationin some finite way. Typically, this is done by sampling them spatially, keeping only one tensor for each small region of an image. Furthermore, distance measures are needed forseveralcommonimageprocessingtechniques,e.g.,interpolation,segmentation, and registration. However, the distance or metric between tensors is not uniquely defined.Thethreechaptersinthispartdiscusssomeoftheconsequencesofvarious choicesforthesedistances. Finally, in Part VII, “Tensor Analysis,” two mathematical areas are presented that were judged to have relevance to researchers at this seminar, but which were not widely known to them in advance. These two chapters offer moderately deep introductions to the areas with the hope of stimulating cross-disciplinary collaborations. Wehopethattheseofferingswill,overallorindividually,inspireinreaderssome oftheenthusiasmandenergythattheseminarbroughttoitsparticipants. Providence,RI,USA DavidH.Laidlaw Eindhoven,TheNetherlands AnnaVilanova Contents PartI Structure-TensorComputation Structure Tensor Estimation: Introducing Monomial QuadratureFilterSets........................................................... 3 HansKnutsson,Carl-FredrikWestin,andMatsAndersson 1 Introduction.................................................................... 4 1.1 ThisChapterPresents................................................... 4 2 MonomialFilters.............................................................. 5 2.1 RadialPart............................................................... 5 2.2 DirectionalMatrix ...................................................... 6 2.3 MonomialFilterMatrices............................................... 7 3 MonomialsLinkOrder,ScaleandGradients ................................ 9 3.1 TheGradientOperatorIncreasesOrderandShiftsScale ............. 11 4 MonomialFilterResponseMatrices ......................................... 12 4.1 SignalClasses........................................................... 13 5 MonomialStructureTensors.................................................. 15 5.1 TwoSimpleExamples.................................................. 15 5.2 GeneralStructureTensorConstruction ................................ 16 5.3 MonomialQuadrature .................................................. 17 5.4 PhaseInvariance ........................................................ 17 5.5 TensorPositivity ........................................................ 17 6 StructureTensorVariations................................................... 18 6.1 TheStructureTensor,T ................................................ 18 q 6.2 TheGradientTensor,T ................................................ 19 G 6.3 TheBoundaryTensor,T ............................................... 19 B 6.4 TheEnergyTensor,T .................................................. 20 E 6.5 GradientEnergyTensor,T ........................................... 20 GE 6.6 Spatial2ndOrderPolynomialTensor.................................. 20 6.7 SphericalHarmonics.................................................... 21 6.8 SumofMonomialTensors.............................................. 21 7 TableofStructureTensorRelatedAlgorithms............................... 21 ix x Contents 8 HigherOrderStructureTensors .............................................. 22 8.1 Non-simpleSignals..................................................... 23 9 Conclusion..................................................................... 24 References......................................................................... 24 AdaptationofTensorVotingtoImageStructureEstimation ............... 29 Rodrigo Moreno,Luis Pizarro, Bernhard Burgeth,Joachim Weickert,MiguelAngelGarcia,andDomenecPuig 1 Introduction.................................................................... 30 2 TensorVoting.................................................................. 32 2.1 StickTensorVoting...................................................... 33 2.2 ThePlateTensorVoting................................................ 34 2.3 TheBallTensorVoting ................................................. 35 3 RelationshipsBetweentheStructureTensorandTensorVoting............ 35 3.1 Similarities .............................................................. 35 3.2 Differences .............................................................. 36 4 TensorVotingforStructureEstimation ...................................... 37 4.1 Gray-ScaleImages...................................................... 37 4.2 ColorandVector-ValuedImages....................................... 38 4.3 Tensor-ValuedImages .................................................. 39 5 ExperimentalResults.......................................................... 42 6 ConcludingRemarks.......................................................... 47 References......................................................................... 48 Edge-EnhancingDiffusionFilteringforMatrixFields ...................... 51 BernhardBurgeth,LuisPizarro,andStephanDidas 1 Introduction.................................................................... 52 2 Edge-EnhancingDiffusion.................................................... 54 3 BasicDifferentialCalculusforMatrixFields................................ 55 4 TheGeneralisedStructureTensorS forMatrixFields.................... 57 G 4.1 ADiffusionTensorDforMatrixFields............................... 58 5 Edge-EnhancingDiffusionFilteringforMatrixFields...................... 58 6 NumericalIssues.............................................................. 59 7 Experiments ................................................................... 60 8 ConcludingRemarks.......................................................... 66 References......................................................................... 66 PartII Tensor-FieldVisualization Fabric-LikeVisualizationofTensorFieldDataonArbitrary SurfacesinImageSpace......................................................... 71 Sebastian Eichelbaum,Mario Hlawitschka, Bernd Hamann, andGerikScheuermann 1 MotivationandRelatedWork................................................. 72 2 Method......................................................................... 73 2.1 InitialNoiseTextureGeneration ....................................... 74
Description: