Ernesto Damiani Jechang Jeong Editors Multimedia Techniques for Device and Ambient Intelligence 123 Multimedia Techniques for Device and Ambient Intelligence . Ernesto Damiani Jechang Jeong Editors Multimedia Techniques for Device and Ambient Intelligence Editors Ernesto Damiani Jechang Jeong Dipartimento di. Tecnologie Department of Electronics dell’Informazione & Computer Engineering Università degli studi di Milano Hanyang University via Bramante 65 17 Haengdang-dong 26013 Crema Seoul, Seongdong-Gu Italy Korea d [email protected] j [email protected] ISBN978-0-387-88776-0 e-ISBN978-0-387-88777-7 DOI10.1007/978-0-387-88777-7 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2009926476 ©SpringerScience+ BusinessMedia,LLC2009 Allrightsreserved. Thisworkmaynotbetranslatedorcopiedinwholeorinpartwithoutthewritten permissionofthepublisher(SpringerScience+BusinessMedia,LLC,233SpringStreet,NewYork,NY 10013,USA),exceptforbriefexcerptsinconnectionwithreviewsorscholarlyanalysis.Useinconnection withanyformofinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar ordissimilarmethodologynowknownorhereafterdevelopedisforbidden. Theuseinthispublicationoftradenames,trademarks,servicemarks,andsimilarterms,eveniftheyare notidentifiedassuch,isnottobetakenasanexpressionofopinionastowhetherornottheyaresubject toproprietaryrights. Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Preface AmbientIntelligenceusedtobeavisionofourfuture.Today,itismorelikeacen- tralaspectofourlife:wealreadylivesurroundedbyanewgenerationofelectronic devices,helpingusinmostofourworkandleisureendeavors.AmbientIntelligence technologiescombineconceptsofubiquitouscomputingandintelligentsystemsto puthumansatthecenteroftechnologicaldevelopment.Manyyoungpeople,inpar- ticular,arefondofambientintelligenceandseemtohavefoundaneasyandnatural waytoexploittheinformationandknowledgeprovidedbythenetworkconnecting thesedevicesandbythedevicethemselves.Insuchascenario,givingtothedevices withthecapabilityofextractingandprocessingmultimediainformationiscrucial. Indeed, time-honored research areas such as video processing and image analysis arenowexperiencingasecondyouth,andplayacrucialroleinsupportingthede- vices’advancedmultimediacapabilities. MultimediaTechniquesforDeviceandAmbientIntelligence(MTDAI)isaedited volumewrittenbywell-recognizedinternationalresearchersincluding,butnotlim- itedto,extendedchapter-styleversionsofthepaperspresentedatthehomonymous MTDAI seminar that we started in 2008, in the unique setting of villa Braida at MoglianoVeneto,nearVenice.TheMTDAIseminarisintendedtobringtogether, without the usual formalities of a conference, a number of top researchers from academia and industry interested in multimedia issues. MTDAI is based on short presentationsandopendiscussions,fosteringinterdisciplinarycollaborationanden- couragingtheexplorationofnewfrontiersintheareaofambientanddeviceintelli- gence. Aftertheseminar,someMTDAIauthorswereaskedtoreviseandextendtheircon- tributions, taking into account the lively discussion and remarks made during the seminar.Also,acallforchapterwaspublished,attractingsomeinterestingpropos- als of additional chapters. A rigorous refereeing was then carried out; the result is thisbook,presentingthestate-of-the-artandsomerecentresearchresultsinthefield ofimageunderstandinganditsapplicationstodeviceandambientintelligence.The bookisdividedintotwoparts:thefirstpartdiscussesnewlow-leveltechniquesfor imageandvideounderstanding,whilethesecondpartpresentsaseriesofnovelap- plications,focusingonmultimedia-orientedknowledgemanagement. vi Preface Puttingtogetherabooklikethisisalwaysateameffort,andwegratefullyacknowl- edge the hard work and dedication of many people. First of all, e appreciate the fundamentalworkoftheMTDAIcommitteemembers,whoacceptedtohandlethe refereeing of the book chapters, and contributed with valuable comments and ob- servation.Wealsowouldliketoacknowledgethehelp,supportandpatienceofthe Springerpublishingteam.Butevenmoreimportantly,wewishtothanktheauthors whohavecontributedtheirbestresearchworktothisvolume.Webelievethatwhile fully attaining the rigorousness and originality one would expect from a scientific editedvolume,theircontributionsretainmuchofthelivelinessandappealtonon- specialistswhichareamajorfeatureofourMTDAIseminar. Milan,Seoul ErnestoDamiani February2009 JechangJeong Contents PartI LowLevelApproachforImageandVideoUnderstanding 1 GOPStructureConversioninTranscodingMPEG-2toH.264/AVC . 3 KangjunLee,GwanggilJeonandJechangJeong 1.1 Introduction.............................................. 3 1.2 GOPStructureConversion.................................. 5 1.2.1 MVScalingintheTemporalDirection................ 5 1.2.2 CorrelationBetweentheCurrentMBModeandthe ReferenceRegion ................................. 6 1.3 ProposedAlgorithms ...................................... 8 1.3.1 AdaptiveSearchRangeSelectionthroughtheMV LinearityTestintheTemporalDirection .............. 8 1.3.2 AdaptiveModeDecisionMethodBasedonRegion Information ...................................... 9 1.4 SimulationResults ........................................ 13 1.5 Conclusion............................................... 16 References..................................................... 16 2 SimpleLowLevelFeaturesforImageAnalysis................... 17 PaoloFalcoz 2.1 Introduction.............................................. 17 2.2 TheRoleofColor......................................... 19 2.2.1 ColorSpaces ..................................... 20 2.2.2 HSLandHSV .................................... 23 2.2.3 CIE-Lab ......................................... 25 2.2.4 ColorFlattening................................... 26 2.3 BlobDetection ........................................... 27 2.4 EdgeDetection ........................................... 30 2.5 SimpleShapes............................................ 33 2.5.1 ScaleandPositionInvariants:ProcrustesAnalysis ...... 33 2.5.2 ShapeAlignment:IterativeClosestPoint.............. 34 viii Contents 2.5.3 ShapeEncodingandMatching:CurvatureSpaceScale .. 35 2.6 Combinationofsimplefeatures ............................. 37 2.7 Conclusions.............................................. 38 References..................................................... 39 3 FastandrobustFaceDetection ................................ 43 MarcoAnisetti 3.1 Relatedwork............................................. 43 3.1.1 Feature-based..................................... 44 3.1.1.1 Lowlevel ............................... 44 3.1.1.2 Skin-map ............................... 45 3.1.1.3 Featureanalysis.......................... 48 3.1.1.4 Templatebased .......................... 49 3.1.2 Appearance-based................................. 50 3.2 Introduction.............................................. 52 3.3 Facedetectiononvideostream .............................. 52 3.3.1 Efficientobjectdetection ........................... 53 3.3.2 Appearance-basedfacedetection..................... 56 3.3.3 Features-basedfacedetection ....................... 58 3.3.3.1 AdaptiveSkindetection ................... 59 3.3.3.2 Eyesdetectionandvalidation .............. 61 3.3.3.3 Facenormalization ....................... 64 3.4 Experimentalresults....................................... 64 3.4.1 DiscussionandConclusions......................... 68 References..................................................... 69 4 Automatic3DFacialFittingforTrackinginVideoSequence ....... 73 ValerioBellandi 4.1 The3DFaceModel ....................................... 73 4.2 3Dmorphingbasis ........................................ 75 4.2.1 3Dmorphingbasisforshapeandexpression........... 77 4.2.1.1 ShapeUnit .............................. 77 4.2.1.2 ExpressionUnit.......................... 79 4.3 Appearancebasis ......................................... 81 4.3.0.3 PCA ................................... 81 4.3.0.4 Image-basedPCA ........................ 82 4.4 3DIlluminationbasis...................................... 87 4.5 TheGeneralPurposes3DTrackingAlgorithm ................. 90 4.5.1 FeatureLocation .................................. 92 4.6 Modeladaptation ......................................... 94 4.6.1 Feature-basedposeestimation....................... 95 4.6.2 Shapeandexpressioninference...................... 98 4.6.3 3DTracking-basedModelrefinement.................101 4.6.4 Initialrefinement..................................101 4.6.5 Deeprefinement ..................................103 4.7 Experimentalresults.......................................105 Contents ix References.....................................................110 PartII MultimediaKnowledge-BasedApproachesandApplications 5 Input Devices and Interaction Techniques for VR-Enhanced Medicine................................................... 115 LuigiGalloandGiuseppeDePietro 5.1 Introduction..............................................116 5.2 RelatedWorks............................................117 5.3 RequirementsAnalysis.....................................119 5.4 InteractionMetaphorsandTechniques........................121 5.4.1 RealisticMetaphors................................122 5.4.1.1 ARealisticMetaphor:VirtualHand..........122 5.4.2 MagicMetaphors..................................122 5.4.2.1 AMagicMetaphor:VirtualPointer. .........123 5.4.3 Pros and Cons of Realistic vs. Magic Interaction Metaphors .......................................123 5.5 TheProposedInputDevice:theWiimote .....................124 5.5.1 Communication...................................125 5.5.2 Inputs ...........................................125 5.5.3 Outputs..........................................125 5.5.4 Classification .....................................125 5.6 TheProposedInteractionTechniques.........................126 5.6.1 TheManipulationState. ............................126 5.6.1.1 Pointing.................................127 5.6.1.2 TranslationandZooming. .................127 5.6.1.3 Rotation.................................128 5.6.2 TheCroppingState. ...............................130 5.7 Discussion ...............................................132 References.....................................................132 6 BridgingSensingandDecisionMakinginAmbientIntelligence Environments .............................................. 135 ElieRaad,BecharaAlBounaandRichardChbeir 6.1 Introduction..............................................136 6.2 RelatedWorks............................................137 6.3 Preliminaries .............................................139 6.4 Templates................................................141 6.5 UncertaintyResolverviaAggregationFunctions ...............142 6.5.1 Average-basedFunction ............................143 6.5.2 BayesianNetwork-BasedFunction...................143 6.5.3 ”DempsterandShafer”-BasedFunction...............146 6.5.4 DecisionTree-BasedFunction.......................148 6.6 Experimentation ..........................................150 6.6.1 AggregationFunctionAccuracyandTimeProcessing ...151 6.6.2 ValueDistribution .................................153 x Contents 6.6.2.1 Test1:Valueshigherthan0.5 ..............153 6.6.2.2 Test2:Valueslessthan0.5 ................154 6.6.2.3 Test3:RandomValues....................154 6.6.2.4 Test5:75%ofthevaluesarelessthan0.5....156 6.6.2.5 Test6:Equallydistributedvalues ...........157 6.6.2.6 Test7:Distributionchange ................158 6.6.2.7 Test8:Influenceofthenumberofreturned values0and1ontheaggregatedresult ......158 6.6.2.8 Discussion ..............................159 6.6.3 TemplateTuning ..................................160 6.6.3.1 Case1:usingthemultimediafunction f .....161 1 6.6.3.2 Case2:usingthemultimediafunction f .....162 2 6.6.3.3 Uncertaintythresholdtuning ...............162 6.7 Conclusion...............................................163 References.....................................................163 7 Ambient Intelligence in Multimedia and Virtual Reality Environmentsfortherehabilitation ............................ 165 AttilaBenkoandSikLanyiCecilia 7.1 Introduction..............................................166 7.2 UsingAIbyspecialneedsusers .............................167 7.2.1 VisualImpairmentandPartiallySightedPeople ........167 7.2.2 DeafandHard-of-HearingPeople....................168 7.2.3 PhysicallyDisabledPersons.........................168 7.2.4 MentallyDisabledPeople...........................169 7.2.5 SmartHome......................................169 7.3 AdetailedexampleofusingAIinvirtualrealityforrehabilitation.170 7.4 Futurevision .............................................174 7.5 Conclusion...............................................175 7.6 Acknowledgement ........................................175 References.....................................................175 8 ArtificialNeuralNetworksforProcessingGraphswithApplication toImageUnderstanding:ASurvey............................. 179 MonicaBianchiniandFrancoScarselli 8.1 FromflattostructuralPatternRecognition ....................179 8.2 Graphprocessingbyneuralnetworks.........................183 8.2.1 Notation .........................................183 8.2.2 Ageneralframeworkforgraphprocessing.............184 8.2.3 RecursiveNeuralNetworks .........................186 8.2.4 GraphNeuralNetworks ............................187 8.2.5 Othermodels .....................................188 8.3 Graph–basedrepresentationofimages........................189 8.3.1 Imagesegmentation ...............................189 8.3.2 RegionAdjacencyGraphs ..........................191 8.3.3 Multi–resolutiontrees..............................194
Description: