Cet ouvrage et des milliers d'autres font partie de la bibliothèque YouScribe
Obtenez un accès à la bibliothèque pour les lire en ligne
En savoir plus

Partagez cette publication

Video Coding with Superimposed
Motion-Compensated Signals
Videocodierung mit ub¨ erlagerten
bewegungskompensierten Signalen
Der Technischen Fakult¨ at der
Friedrich-Alexander-Universit¨ at Erlangen-Nurn¨ berg
zur Erlangung des Grades
Doktor-Ingenieur
vorgelegt von
Markus Helmut Flierl
Erlangen, 2003Als Dissertation genehmigt von
der Technischen Fakult¨ at der
Friedrich-Alexander-Universit¨ at
Erlangen-Nurn¨ berg
Tag der Einreichung: 13.03.2003
Tag der Promotion: 27.06.2003
Dekan: Prof. Dr. rer. Nat. A. Winnacker
Berichterstatter: Prof. Dr.-Ing. B. Girod
Prof. A. KaupAcknowledgments
This work required a substantial effort and I am very grateful to many people who made
this endeavor possible. I would like to thank my supervisor Prof. Bernd Girod for the
opportunity to join his group in Erlangen and Stanford as well as to benefit from the
inspiring environment. I would also like to thank Prof. Andr´e Kaup for his interest in my
work and reviewing my thesis as well as Prof. Heinrich Niemann and Prof. Wolfgang Koch
for joining the committee. I am thankful to my fellow students Anne Aaron, Chuo-Ling
Chang, Jacob Charareski, Joachim Eggers, Peter Eisert, Niko F¨ arber, Sang-Eun Han,
Frank Hartung, Mark Kalman, Yi Liang, Marcus Magnor, James Mammen, Prashant
Ramanathan, Shantanu Rane, Marion Schabert, Eric Setton, Wolfgang Sorgel,¨ Eckehard
Steinbach, Klaus Stuhlmuller,¨ Jonathan Su, Matthias Teschner, Lutz Trautmann, Thomas
Wiegand, and Rui Zhang for many stimulating discussions, joint work, and proofreading.
In addition, I am thankful for helpful discussions with Dr. Nick Kingsbury. I am also
grateful to the people at the Telecommunications Laboratory and Information Systems
Laboratory, especially to Ursula Arnold and Kelly Yilmaz for their invaluable adminis-
trative support. My special thanks belong to my family and friends for all the support
they gave.Erkenne die Musik dessen vollendete T¨ one verklingen.
F¨ uhle die Stille die zum Traumen¨ verfuhrt.¨ Erfahre den
unendlichen Raum der hinschwindenden Ansammlung
sich vollendender T¨ one.
— Deinem Odysseusvii
Contents
1 Introduction 1
2 Background and Related Work 7
2.1 CodingofVideoSignals............................ 7
2.2 MotionCompensation............................. 8
2.2.1 BidirectionalMotionCompensation.................. 9
2.2.2 Overlapped Block Motion Compensation . . . . . . . . . . . . . . . 10
2.2.3 Variable Block Size Motion Compensation . . . . . . . . . . . . . . 11
2.2.4 Multiframe Motion Compensation . . . . . . . . . . . . . . . . . . . 12
2.2.5 Superimposed Motion Compensation . . . . . . . . . . . . . . . . . 13
2.3 MotionEstimation............................... 16
2.3.1 Rate-Constrained Motion Estimation . . . . . . . . . . . . . . . . . 16
2.3.2 R Estimation of Superimposed Motion . . . . . . . 18
2.3.3 QuantizerSelectionattheResidualEncoder............. 20
2.3.4 EfficientMotionEstimation...................... 2
2.4 Theory of Motion-Compensated Prediction . . . . . . . . . . . . . . . . . . 23
2.4.1 FrameSignalModel .......................... 24
2.4.2 Signal Model for Motion-Compensated Prediction . . . . . . . . . . 25
2.4.3 Signal Model for Multihypothesis Prediction . . . . . . . . . . . . . 26
2.4.4 PerformanceMeasures......................... 28
2.4.5 Conclusions............................... 29
2.5 Three-Dimensional Subband Coding of Video . . . . . . . . . . . . . . . . 29
2.5.1 Motion Compensation and Subband Coding . . . . . . . . . . . . . 30
2.5.2 Motion-CompensatedLiftedWavelets................. 31viii Contents
3 Motion-Compensated Prediction with Complementary Hypotheses 33
3.1 Introduction................................... 3
3.2 Extended Model for Superimposed Motion-Compensated Prediction . . . . 34
3.2.1 Superimposed Prediction and Correlated Displacement Error . . . . 34
3.2.2 ComplementaryHypotheses...................... 36
3.2.3 GradientofthePredictionErorVariance.............. 41
3.3 HypotheseswithAdditiveNoise........................ 42
3.3.1 AveragingFilter............................. 43
3.3.2 WienerFilter.............................. 45
3.4 Forward-AdaptiveHypothesisSwitching................... 52
3.4.1 SignalModelforHypothesisSwitching................ 53
3.4.2 MinimizingtheRadialDisplacementEror.............. 53
3.4.3 EquivalentPredictor.......................... 56
3.4.4 Motion Compensation with Complementary Hypotheses and For-
ward-AdaptiveHypothesisSwitching................. 57
3.5 PictureswithVaryingNumberofHypotheses ................ 59
3.6 Conclusions................................... 61
4 ITU-T Rec. H.263 and Superimposed Prediction 63
4.1 Introduction 63
4.2 VideoCodingwithSuperimposedMotion................... 64
4.2.1 SyntaxExtensions ........................... 64
4.2.2 CoderControl.............................. 64
4.3 ExperimentalResults.............................. 66
4.3.1 Multiple Hypotheses for Constant Block Size . . . . . . . . . . . . . 66
4.3.2 Multiple Hyp for Variable Block Size . . . . . . . . . . . . . 67
4.3.3 Multiple Hypotheses and Multiple Reference Pictures . . . . . . . . 69
4.4 Conclusions................................... 76
5 ITU-T Rec. H.264 and Generalized B-Pictures 77
5.1 Introduction 7
5.2 PredictionModesforB-Pictures........................ 78
5.2.1 Overview ................................ 78
5.2.2 DirectMode 79
5.2.3 SuperpositionMode .......................... 80
5.2.4 Rate-Distortion Performance of Individual Modes . . . . . . . . . . 80
5.3 SuperimposedPrediction............................ 83
5.3.1 Bidirectionalvs.SuperpositionMode................. 83Contents ix
5.3.2 Two Combined Forward Prediction Signals . . . . . . . . . . . . . . 84
5.3.3 EntropyCoding............................. 89
5.4 EncoderIsues ................................. 90
5.4.1 RateConstrainedModeDecision................... 90
5.4.2 Rate Constrained Motion Estimation . . . . . . . . . . . . . . . . . 91
5.4.3 Rate Superimposed Motion Estimation . . . . . . . . . 92
5.4.4 Improving Overall Rate-Distortion Performance . . . . . . . . . . . 93
5.5 Conclusions................................... 97
6 Motion Compensation for Groups of Pictures 99
6.1 Introduction 9
6.2 CodingScheme................................. 9
6.2.1 Motion-Compensated Lifted Haar Wavelet . . . . . . . . . . . . . . 100
6.2.2 Mensated Lifted 5/3 Wavelet . . . . . . . . . . . . . . . 101
6.2.3 ExperimentalResults..........................101
6.3 A Mathematical Model of Motion-Compensated Subband Coding . . . . . 108
6.3.1 Motion-Compensated Lifted Haar Wavelet . . . . . . . . . . . . . . 108
6.3.2 Mensated Lifted 5/3 Wavelet . . . . . . . . . . . . . . . 109
6.3.3 SignalModel ..............................109
6.3.4 TransformCodingGain ........................13
6.4 Conclusions...................................15
7 Summary 117
A Mathematical Results 123
A.1 Singularities of the Displacement Error Covariance Matrix . . . . . . . . . 123
A.2AClasofMatricesandtheirEigenvalues ..................123
A.3 Inverse of the Power Spectral Density Matrix . . . . . . . . . . . . . . . . . 124
A.4PowerSpectralDensityofaFrame......................126
Glossary 127
Bibliography 131x Contents
1Inhaltsverzeichnis
1Einfuhrung¨ 1
2 Grundlagen und Stand der Forschung 7
2.1 CodierungvonVideosignalen ......................... 7
2.2 Bewegungskompensation............................ 8
2.2.1 BidirektionaleBewegungskompensation................ 9
2.2.2 Bewegungskompensation mit ub¨ erlappenden Bl¨ocken........ 10
2.2.3 Bewompe mit Bl¨ ocken variabler Gr¨oße........ 1
2.2.4 Bewegungskompensation mit mehreren Referenzbildern . . . . . . . 12
¨2.2.5 UberlagerteBewegungskompensation................. 13
2.3 Bewegungssch¨atzung.............................. 16
2.3.1 Bewegungssch¨atzungmitRatennebenbedingung........... 16
2.3.2 Schatzung¨ ub¨ erlagerter Bewegung mit Ratennebenbedingung . . . . 18
2.3.3 Wahl des Quantisierers beim Pr¨adiktionsfehler-Encoder....... 20
2.3.4 Effiziente Bewegungssch¨atzung..................... 2
2.4 Theorie der bewegungskompensierten Pr¨adiktion............... 23
2.4.1 Bildsignalmodell ............................ 24
2.4.2 Signalmodell fur¨ bewegungskompensierte Pr¨adiktion......... 25
2.4.3 Sdell fur¨ Multihypothesen-Pr¨adiktion ............ 26
2.4.4 Bewertungsmaße 28
2.4.5 Schlusfolgerungen........................... 29
2.5 Drei-DimensionaleTeilbandcodierungvonVideo............... 29
2.5.1 Bewegungskompensation und Teilbandcodierung . . . . . . . . . . . 30
2.5.2 BewegungskompensierteWavelets................... 31
3 Bewegungskompensierte Pr¨adiktion mit komplement¨aren Hypothesen 33
3.1 Einleitung.................................... 33
3.2 Erweitertes Modell fur¨ die ub¨ erlagerte bewegungskompensierte Pr¨ adiktion . 34
¨3.2.1 Uberlagerte Pr¨ adiktion und korrelierte Versatzfehler . . . . . . . . . 34
3.2.2 Komplement¨areHypothesen...................... 36
3.2.3 Gradient der Pr¨adiktionsfehlervarianz................. 41
3.3 HypothesenmitadditivemRauschen..................... 42
3.3.1 Mitelungsfilter............................. 43
3.3.2 WienerFilter.............................. 45
1This translation of the table of contents is a requirement according to the ”Prufungsordn¨ ung fur¨ die
Technischen Fakult¨ at der Friedrich-Alexander-Universit¨ at Erlangen-Nurn¨ berg”.

Un pour Un
Permettre à tous d'accéder à la lecture
Pour chaque accès à la bibliothèque, YouScribe donne un accès à une personne dans le besoin