Image-Based Modeling

Foreword

Preface

Acknowledgements

Notation

Contents

Introduction

Part I Geometry: fundamentals of multi-view geometry

Geometry prerequisite

2.1 Introduction

2.2 Projective geometry

2.2.1 The basic concepts

2.2.2 Projective spaces and transformations

2.2.3 Affine and Euclidean specialization

2.3 Algebraic geometry

2.3.1 The simple methods

2.3.2 Ideals, varieties, and Gr¨obner bases

2.3.3 Solving polynomial equations with Gr¨obner bases

Multi-view geometry

3.1 Introduction

3.2 The single-view geometry

3.2.1 What is a camera?

3.2.2 Where is the camera?

3.2.3 The DLT calibration

3.2.4 The three-point pose algorithm

3.3 The uncalibrated two-view geometry

3.3.1 The fundamental matrix

3.3.2 The seven-point algorithm

3.3.3 The eight-point linear algorithm

3.4 The calibrated two-view geometry

3.4.1 The essential matrix

3.4.2 The five-point algorithm

3.5 The three-view geometry

3.5.1 The trifocal tensor

3.5.2 The six-point algorithm

3.5.3 The calibrated three views

3.6 The N-view geometry

3.6.1 The multi-linearities

3.6.2 Auto-calibration

3.7 Discussions

3.8 Bibliographic notes

Part II Computation: from pixels to 3D points

Feature point

4.1 Introduction

4.2 Points of interest

4.2.1 Tracking features

4.2.2 Matching corners

4.2.3 Discussions

4.3 Scale invariance

4.3.1 Invariance and stability

4.3.2 Scale, blob and Laplacian

4.3.3 Recognizing SIFT

4.4 Bibliographic notes

Structure from Motion

5.1 Introduction

5.1.1 Least squares and bundle adjustment

5.1.2 Robust statistics and RANSAC

5.2 The standard sparse approach

5.2.1 A sequence of images

5.2.2 A collection of images

5.3 The match propagation

5.3.1 The best-first match propagation

5.3.2 The properties of match propagation

5.3.3 Discussions

5.4 The quasi-dense approach

5.4.1 The quasi-dense resampling

5.4.2 The quasi-dense SFM

5.4.3 Results and discussions

5.5 Bibliographic notes

Part III Modeling: from 3D points to objects

Surface modeling

6.1 Introduction

6.2 Minimal surface functionals

6.3 A unified functional

6.4 Level-set method

6.5 A bounded regularization method

6.6 Implementation

6.7 Results and discussions

6.8 Bibliographic notes

Hair modeling

7.1 Introduction

7.2 Hair volume determination

7.3 Hair fiber recovery

7.3.1 Visibility determination

7.3.2 Orientation consistency

7.3.3 Orientation triangulation

7.4 Implementation

7.5 Results and discussions

7.6 Bibliographic notes

Tree modeling

8.1 Introduction

8.2 Branche recovery

8.2.1 Reconstruction of visible branches

8.2.2 Synthesis of occluded branches

8.2.3 Interactive editing

8.3 Leaf extraction and reconstruction

8.3.1 Leaf texture segmentation

8.3.2 Graph-based leaf extraction

8.3.3 Model-based leaf reconstruction

8.4 Results and discussions

8.5 Bibliographic notes

Fac¸ade modeling

9.1 Introduction

9.2 Fac¸ade initialization

9.2.1 Initial flat rectangle

9.2.2 Texture composition

9.2.3 Interactive refinement

9.3 Fac¸ade decomposition

9.3.1 Hidden structure discovery

9.3.2 Recursive subdivision

9.3.3 Repetitive pattern representation

9.3.4 Interactive subdivision refinement

9.4 Fac¸ade augmentation

9.4.1 Depth optimization

9.4.2 Cost definition

9.4.3 Interactive depth assignment

9.5 Fac¸ade completion

9.6 Results and discussions

9.7 Bibliographic notes

Building modeling

10.1 Introduction

10.2 Pre-processing

10.3 Building segmentation

10.3.1 Supervised class recognition

10.3.2 Multi-view semantic segmentation

10.4 Building partition

10.4.1 Global vertical alignment

10.4.2 Block separator

10.4.3 Local horizontal alignment

10.5 Fac¸ade modeling

10.5.1 Inverse orthographic composition

10.5.2 Structure analysis and regularization

10.5.3 Repetitive pattern rediscovery

10.5.4 Boundary regularization

10.6 Post-processing

10.7 Results and discussions

10.8 Bibliographic notes

List of Algorithms

List of Figures

References

Index