Modeling and Rendering Architecture from Photographs
Table of Contents
List of Figures vii
List of Tables ix
1 Introduction 1
2 Background and Related Work 8
2.1 Camera calibration ............................................... 8
2.2 Structure from motion ............................................ 9
2.3 Shape from silhouette contours .................................. 10
2.4 Stereo correspondence ........................................... 15
2.5 Range scanning .................................................. 18
2.6 Image-based modeling and rendering .............................. 18
3 Overview 22
4 Camera Calibration 25
4.1 The perspective model ........................................... 25
4.2 How real cameras deviate from the pinhole model ................. 28
4.3 Our calibration method .......................................... 31
4.4 Determining the radial distortion coefficients .................. 31
4.5 Determining the intrinsic parameters ............................ 38
4.6 Working with uncalibrated images ................................ 39
5 Photogrammetric Modeling 42
5.1 Overview of the Facade photogrammetric modeling system .......... 45
5.2 The model representation ........................................ 47
5.2.1 Parameter reduction ........................................... 47
5.2.2 Blocks ........................................................ 48
5.2.3 Relations (the model hierarchy) ............................... 51
5.2.4 Symbol references ............................................. 52
5.2.5 Computing edge positions using the hierarchical structure ..... 52
5.2.6 Discussion .................................................... 54
5.3 Facade's user interface ......................................... 55
5.3.1 Overview ...................................................... 55
5.3.2 A Facade project .............................................. 57
5.3.3 The windows and what they do .................................. 57
5.3.4 The Camera Parameters Form .................................... 60
5.3.5 The Block Parameters form ..................................... 61
5.3.6 Reconstruction options ........................................ 66
5.3.7 Other tools and features ...................................... 67
5.4 The reconstruction algorithm .................................... 67
5.4.1 The objective function ........................................ 67
5.4.2 Minimizing the Objective Function ............................. 69
5.4.3 Obtaining an initial estimate ................................. 70
5.5 Results ......................................................... 72
5.5.1 The Campanile ................................................. 72
5.5.2 University High School ........................................ 77
5.5.3 Hoover Tower .................................................. 77
5.5.4 The Taj Mahal and the Arc de Trioumphe ........................ 79
6 View-Dependent Texture Mapping 81
6.1 Motivation ...................................................... 81
6.2 Overview ........................................................ 82
6.3 Projecting a single image onto the model ........................ 83
6.3.1 Computing shadow regions ...................................... 84
6.4 View-dependent composition of multiple images ................... 84
6.4.1 Determining the fitness of a particular view .................. 85
6.4.2 Blending images ............................................... 88
6.5 Improving rendering quality ..................................... 89
6.5.1 Reducing seams in renderings .................................. 89
6.5.2 Removal of obstructions ....................................... 89
6.5.3 Filling in holes .............................................. 91
6.6 Results: the University High School fly-around .................. 92
6.7 Possible performance enhancements ............................... 92
6.7.1 Approximating the fitness functions ........................... 95
6.7.2 Visibility preprocessing ...................................... 95
7 Model-Based Stereo 96
7.1 Motivation ...................................................... 96
7.2 Differences from traditional stereo ............................. 97
7.3 Epipolar geometry in model-based stereo ........................ 100
7.4 The matching algorithm ......................................... 102
7.5 Results ........................................................ 103
8 Rouen Revisited 105
8.1 Overview ....................................................... 105
8.2 Artistic description ........................................... 105
8.3 The making of Rouen Revisited .................................. 109
8.3.1 Taking the pictures .......................................... 109
8.3.2 Mosaicing the Beta photographs ............................... 112
8.3.3 Constructing the basic model ................................. 112
8.4 Recovering additional detail with model-based stereo ........... 115
8.4.1 Generating surface meshes .................................... 115
8.4.2 Rectifying the series of images .............................. 116
8.5 Recovering a model from the old photographs .................... 120
8.5.1 Calibrating the old photographs .............................. 120
8.5.2 Generating the historic geometry ............................. 121
8.6 Registering the Monet Paintings ................................ 122
8.6.1 Cataloging the paintings by point of view .................... 122
8.6.2 Solving for Monet's position and intrinsic parameters ........ 122
8.6.3 Rendering with view-dependent texture mapping ................ 123
8.6.4 Signing the work ............................................. 125
Bibliography 132
Appendix: Obtaining color images and animations 139
Back to Modeling and Rendering Architecture from Photographs Thesis
Paul E. Debevec / paul@debevec.org