## Structure and Motion from Silhouettes (2001)

### Cached

### Download Links

Citations: | 50 - 12 self |

### BibTeX

@INPROCEEDINGS{Wong01structureand,

author = {Kwan-yee Kenneth Wong},

title = {Structure and Motion from Silhouettes},

booktitle = {},

year = {2001},

pages = {217--222}

}

### Years of Citing Articles

### OpenURL

### Abstract

I hereby declare that no part of this thesis has already been or is being submitted for any other degree or qualification. This dissertation is the result of my own original work carried out in the Department of Engineering at the University of Cambridge, except where explicit reference has been made to the work of others. This dissertation contains 36,194 words and 91 figures. ii “Cogito, ergo sum. ” (I think, therefore I am.)- René Descartes, Le Discours de la Méthode. iv

### Citations

2078 | E.: Marching cubes: A high resolution 3D surface construction algorithm
- LORENSEN, CLINE
- 1987
(Show Context)
Citation Context ... a volumetric model of the object can be constructed by octree carving algorithm [18] using the silhouettes. Surface triangles can then be generated from the octree using the marching cubes algorithm =-=[13]-=- and texture for each triangle is taken from the view which gives the largest projected area. Instead of using texture-mapping which sometimes produces color inconsistency between neighboring triangle... |

630 |
A computer algorithm for reconstructing a scene from two projections
- Longuet-Higgins
- 1981
(Show Context)
Citation Context ...oretically, if 7 or more epipolar tangencies are available, the epipolar geometry between 2 views can be estimated, and the camera intrinsic parameters can then be used to recover the relative motion =-=[12, 6]-=-. However, when the epipolar geometry is not known, the localization of the epipolar tangencies involves a nonlinear optimization with nontrivial initialization. The unrealistic demand for a large num... |

504 |
What can be seen in three dimensions with an uncalibrated stereo rig
- Faugeras
- 1992
(Show Context)
Citation Context ...oretically, if 7 or more epipolar tangencies are available, the epipolar geometry between 2 views can be estimated, and the camera intrinsic parameters can then be used to recover the relative motion =-=[12, 6]-=-. However, when the epipolar geometry is not known, the localization of the epipolar tangencies involves a nonlinear optimization with nontrivial initialization. The unrealistic demand for a large num... |

322 | Determining the epipolar geometry and its uncertainty: A review
- Zhang
- 1998
(Show Context)
Citation Context ...eflectance, and can only handle objects of limited size. By allowing 3D models to be reconstructed automatically from image sequences, the structure from motion 1s2 CHAPTER 1. INTRODUCTION techniques =-=[132, 41, 81, 146, 36]-=- in computer vision provide a cost-efficient solution to the above problem. In addition, vision-based systems can also handle objects with various size and reflectance. 1.1.1 Structure from Motion In ... |

233 | The fundamental matrix: Theory, algorithms, and stability analysis
- Luong, Faugeras
- 1996
(Show Context)
Citation Context ...polar tangent points are located in viewsand¡ respectively (see fig. 6). The reprojection errors are then given by the geometric distances between the epipolar tangent points and their epipolar lines=-= [14], ,©¤§¢�� and©¤§¢�� ,©¢�¤�� ©¢�¤�� � ¢�¤�� � �¤§¢�� � e j2 � � � © ���� � � ��¢¥¤©¢¥¤�� � © ��¢-=-¥¤©¢¥¤�� ��¢�¤©¢¥¤�� ¤§¢�� ��� � ¢¥¤©¢¥¤�� ¤§¢�� � � (5) ��� � � �¢�¤©¤§¢�� �� � For a sequence of � imag... |

223 | Shape from shading : a survey
- Zhang, Tsai, et al.
- 1999
(Show Context)
Citation Context ...ss, under certain appropriate assumptions, it is possible to infer scene structure, like surface orientation and curvature, from a single image. Examples of such techniques include shape from shading =-=[58, 61, 142, 59, 145]-=- under the assumptions of point light source and Lambertian surface, shape from line drawings [112, 50, 60, 28, 136, 83, 131, 122, 84, 102] under the assumption of trihedral-vertex polyhedral scene st... |

222 |
Rapid Octree Construction from Image Sequences
- SZELISKI
- 1993
(Show Context)
Citation Context ...��� ¢¥¤ ����� � � the baseline formed with viewspasses through the object, otherwise it is � . After motion estimation, a volumetric model of the object can be constructed by =-=octree carving algorithm [18]-=- using the silhouettes. Surface triangles can then be generated from the octree using the marching cubes algorithm [13] and texture for each triangle is taken from the view which gives the largest pro... |

206 | Inverse global illumination: Recovering reflectance models of real scenes from photographs - Yu, Debevec, et al. - 1999 |

140 |
What does the occluding contour tell us about solid shape
- Koenderink
- 1984
(Show Context)
Citation Context ...motion of an object, and are indeed the only information available in the case of smooth textureless surfaces. Nonetheless, structure and motion from silhouettes has always been a challenging problem =-=[11, 3, 2, 17, 1, 5, 10, 15]-=-. Unlike corners, silhouettes are projections of contour generators [3] which are viewpoint dependent, and hence do not readily provide point correspondences. A traditional approach to the problem is ... |

123 |
Surface shape from deformation of apparent contours
- Cipolla, Blake
- 1992
(Show Context)
Citation Context ...motion of an object, and are indeed the only information available in the case of smooth textureless surfaces. Nonetheless, structure and motion from silhouettes has always been a challenging problem =-=[11, 3, 2, 17, 1, 5, 10, 15]-=-. Unlike corners, silhouettes are projections of contour generators [3] which are viewpoint dependent, and hence do not readily provide point correspondences. A traditional approach to the problem is ... |

119 | Object-centered surface reconstruction: Combining multiimage stereo and shading - Fua, Leclerc - 1995 |

102 | Visual Motion of Curves and Surfaces
- Cipolla, Giblin
- 1999
(Show Context)
Citation Context ...mental results, and conclusions are given in Section 5. 2. Theoretical Background Silhouettes are projections of contour generators, and do not readily provide point correspondences. A frontier point =-=[9, 2, 4]-=- is the intersection of two contour generators and lies on an epipolar plane which is tangent to the surface (see fig. 1). It follows that a frontier point will project onto a point in the silhouette ... |

93 | Automatic 3D model construction for turn-table sequences
- Fitzgibbon, Cross, et al.
- 1998
(Show Context)
Citation Context ...es due to selfocclusion, and greatly simplifies the matching problem. where � is ����� the camera calibration matrix. The fundamental matrix can be parameterized explicitly in terms of the=-=se features [19, 7, 15], and is given by where � ��� � ¦����������-=-������ ¥ �s¡s��£ ��� � £s��¡�� �s(3) is the angle of rotation, and � is a constant which can be determined from the camera intrinsic parameters. � is the only ... |

70 | Automatic Model Construction, Pose Estimation, and Object Recognition from Photographs using Triangular Splines
- Sullivan, Ponce
- 1998
(Show Context)
Citation Context ...motion of an object, and are indeed the only information available in the case of smooth textureless surfaces. Nonetheless, structure and motion from silhouettes has always been a challenging problem =-=[11, 3, 2, 17, 1, 5, 10, 15]-=-. Unlike corners, silhouettes are projections of contour generators [3] which are viewpoint dependent, and hence do not readily provide point correspondences. A traditional approach to the problem is ... |

50 | Motion from the Frontier of Curved Surfaces
- Cipolla, Astrom, et al.
- 1995
(Show Context)
Citation Context |

50 | Curve Matching and Stereo Calibration
- Porrill, Pollard
- 1991
(Show Context)
Citation Context ...projections of contour generators [3] which are viewpoint dependent, and hence do not readily provide point correspondences. A traditional approach to the problem is to search for epipolar tangencies =-=[16, 2]-=-. However, such a search involves a nonlinear optimization with nontrivial initialization, and requires the presence of 7 or more epipolar tangencies. Recently, a practical solution has been found in ... |

32 | Resolving ambiguities in autocalibration
- Zisserman, Liebowitz, et al.
- 1998
(Show Context)
Citation Context ...tion (3.15) gives the pole-polar relationship, with respect to the image of the absolute conic, between the vanishing point ÚÜ of the normal direction of the plane¥ × and the vanishing line Ð × of¥ × =-=[148]-=-. By assuming the skew of È to be zero (i.e.� ), substituting (2.3) into (3.15) gives ÚÜ� � � � Ù Ù Ú Ù Ù Ú � Ú Ú Ù Ú � Ð ×� (3.16) where �, � and (Ù ,Ú ) are the intrinsic parameters of È, as defined... |

27 | Structure and motion estimation from dynamic silhouettes under perspective projection
- Joshi, Ahuja, et al.
- 1999
(Show Context)
Citation Context |

25 | S.: Class-based grouping in perspective images
- Zisserman, Mundy, et al.
- 1995
(Show Context)
Citation Context ...ion, then a single view alone allows shape recovery. In this dissertation, the surface geometry of surfaces of revolution is studied. Through the use of differential geometry and projective invariant =-=[147, 77, 101, 149, 33]-=-, it is shown that the 3D shape of a surface of revolution can be recovered from its silhouette in a single view, up to an 1-parameter ambiguity. An alternative approach for depth recovery is to intro... |

24 |
Recognizing general curved objects efficiently
- Zisserman, Forsyth, et al.
- 1992
(Show Context)
Citation Context ...ion, then a single view alone allows shape recovery. In this dissertation, the surface geometry of surfaces of revolution is studied. Through the use of differential geometry and projective invariant =-=[147, 77, 101, 149, 33]-=-, it is shown that the 3D shape of a surface of revolution can be recovered from its silhouette in a single view, up to an 1-parameter ambiguity. An alternative approach for depth recovery is to intro... |

24 | Motion estimation in image sequences using the deformation of apparent contours
- Kahl, ˚Aström
- 1999
(Show Context)
Citation Context |

24 | Using singular displacements for uncalibrated monocular vision systems
- Vieville, Lingrand
- 1995
(Show Context)
Citation Context ...es due to selfocclusion, and greatly simplifies the matching problem. where � is ����� the camera calibration matrix. The fundamental matrix can be parameterized explicitly in terms of the=-=se features [19, 7, 15], and is given by where � ��� � ¦����������-=-������ ¥ �s¡s��£ ��� � £s��¡�� �s(3) is the angle of rotation, and � is a constant which can be determined from the camera intrinsic parameters. � is the only ... |

19 |
Volumetric Descriptions from a Single Intensity Image
- Zerroug, Nevatia
- 1996
(Show Context)
Citation Context ... of straight homogeneous generalized cylinders (SHGCs) and their silhouettes had been studied by various researchers [57, 106, 77], and exploited for object recognition and object pose estimation. In =-=[114, 49, 144]-=-, algorithms for segmentation and 3D recovery of SHGCs under orthographic projection were presented. In [133], Ulupinar and Nevatia addressed the recovery of curved-axis planar right constant generali... |

19 | Parallax geometry of smooth surfaces in multiple views
- Cross, Fitzgibbon, et al.
- 1999
(Show Context)
Citation Context |

19 |
Recovery of an unknown axis of rotation from the profiles of a rotating surface
- Giblin, Pollick, et al.
- 1994
(Show Context)
Citation Context ...mental results, and conclusions are given in Section 5. 2. Theoretical Background Silhouettes are projections of contour generators, and do not readily provide point correspondences. A frontier point =-=[9, 2, 4]-=- is the intersection of two contour generators and lies on an epipolar plane which is tangent to the surface (see fig. 1). It follows that a frontier point will project onto a point in the silhouette ... |

19 | Camera pose estimation and reconstruction from image profiles under circular motion
- Mendonça, Wong, et al.
- 2000
(Show Context)
Citation Context |