## A probabilistic framework for space carving (2001)

Citations: | 66 - 3 self |

### BibTeX

@INPROCEEDINGS{Broadhurst01aprobabilistic,

author = {Adrian Broadhurst},

title = {A probabilistic framework for space carving},

booktitle = {},

year = {2001},

pages = {388--393}

}

### Years of Citing Articles

### OpenURL

### Abstract

for the degree of Doctor of Philosophy. ii This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration. This dissertation is not substantially the same as any I have submitted for a degree or diploma or other qualification at any other University. I further state that no part of my dissertation/thesis has already been, or is being concurrently submitted for any such degree, diploma or other quali-fication. This dissertation contains 78 figures and approximately 46000 words. This dissertation was revised December 2001. This thesis investigates the problem of reconstructing three-dimensional objects from image sequences. There are two major contributions in this thesis. The first contribution is an extension to the Space Carving framework that elimi-

### Citations

1164 |
Error bounds for convolutional codes and an asymptotically optimum decoding algorithm
- Viterbi
- 1967
(Show Context)
Citation Context ...fferent algorithms in the past, which have attempted to use different types of image features [71, 83, 5, 8] different methods for locating the features [42, 70, 17, 29] different matching strategies =-=[5, 128, 97, 83, 98, 58, 35]-=-, and different approaches for representing the shapes of the scenes [71, 89, 10, 19, 103, 35, 123, 110]. So why is three-dimensional reconstruction an important problem to solve? and why is another a... |

385 | Bundle Adjustment - A modern synthesis
- Triggs, McLauchlan, et al.
(Show Context)
Citation Context ...ly possible when two or more views are available as it relies on being able to triangulate (see Hartley [44]) the 3-Ds168 Multiple view geometry position of each vertex. Bundle adjustment (see Triggs =-=[126]-=-) is the simplest solution to the structure and motion problem. In this approach all the points and all the camera parameters, are assumed to be be approximately known, and gradient descent is used to... |

322 | Determining the epipolar geometry and its uncertainty: A review
- Zhang
- 1998
(Show Context)
Citation Context ...fundamental matrix 1 encodes all the information to describe the geometry of two views. The derivation which follows is based on the work by Luong [67], but a more rigorous derivation can be found in =-=[132]-=-. Referring to Figure B.2 and (B.6), it can be seen that any Cartesian � � XYZ three-dimensional point ˜x = projects to point p1 = ˜x in the first image (in projective co-ordinates), and in the second... |

212 | The Absolute Quadric
- Triggs
- 1997
(Show Context)
Citation Context ..., except for final viewing of the model. Many two view calibration techniques do exist (e.g. Maybank and Faugeras [74], or Hartley [43]), but are less stable than the multiple view approaches (Triggs =-=[125]-=-, and Pollefeys [90]).s172 Multiple view geometry B.7 Trifocal tensor B.7.1 Derivation It has been shown how the fundamental matrix describes the two view constraints, and the next step is to consider... |

108 | An integrated Bayesian approach to layer extraction from image sequences
- Torr, Szeliski, et al.
- 2001
(Show Context)
Citation Context ..., 83, 5, 8] different methods for locating the features [42, 70, 17, 29] different matching strategies [5, 128, 97, 83, 98, 58, 35], and different approaches for representing the shapes of the scenes =-=[71, 89, 10, 19, 103, 35, 123, 110]-=-. So why is three-dimensional reconstruction an important problem to solve? and why is another approach necessary to solve it? These questions are answered in the remainder of this chapter, and in the... |

103 |
Using Extremal Boundaries for 3-D Object Modeling
- Vaillant, Faugeras
- 1992
(Show Context)
Citation Context ... from silhouettes is not limited to volumetric approaches. Apparent (or occluding) contours may also be used to obtain a surface-based representation (Cipolla and Blake [23] and Vaillant and Faugeras =-=[127]-=-). To give some idea of the power of occluding contours, Mendonça, Wong and Cipolla [79] are able to infer the camera pose from a turntable sequence without requiring any additional point corresponden... |

49 | Rendering real-world objects using view interpolation
- Werner, Hersch, et al.
- 1995
(Show Context)
Citation Context ...el arrays (Voxel Colouring [104]) in the latter parts of his PhD thesis [103]. Several other approaches have used image interpolation in some form: Laveau and Faugeras [62]; Werner, Hersch and Hlaváč =-=[129]-=-; Chang and Zakhor [18]; and Avidan and Shashua [3], but all these approaches rely on having some sort of underlying depth representation. McMillan and Bishop [76] represented the 3-D world as a 5-D l... |

32 | Resolving ambiguities in autocalibration
- Zisserman, Liebowitz, et al.
- 1998
(Show Context)
Citation Context ...l basis by Quan [94], elegantly wrapped in a RANSAC [36] framework.s176 Multiple view geometry B.8 Auto-calibration: Finding a Euclidean basis Quite a lot is known about the field of auto-calibration =-=[63, 74, 43, 133, 117, 125, 90, 78]-=- and there are two main approaches. The traditional approaches (e.g. [63, 74, 43, 78]) aim to solve for the calibration parameters only, i.e. the K matrices (see Section B.2). More recent approaches l... |

10 | Layer extraction with a bayesian model of shapes
- Torr, Dick, et al.
- 2000
(Show Context)
Citation Context ...e polygon representations are very general in shape, it is always useful to include prior knowledge about the scene. This is especially useful in the case of architecture (e.g. Torr, Dick and Cipolla =-=[122]-=-), where windows and doors are often regular and can be represented by parametric shape models. Polygon-based reconstructions can be made significantly more realistic by using photographic and bas-rel... |

8 |
Fast approximate energy minimisation via graph cuts
- Boykov, Veksler, et al.
- 2001
(Show Context)
Citation Context ...and to obtain dense stereo correspondences, but can only be applied on a scanline-by-scanline basis. More recently, the inter-scanline consistency problem has been solved by Boykov, Veksler and Zabih =-=[131]-=- and Roy and Cox [98], who have used graph cuts to find a global solution to the correspondence problem. This is an elegant solution, but still requires the model to be stored as a 2 1 2-D sketch. The... |

6 |
Robust parameterisation and computation of the trifocal tensor
- Tort, Zisserman
- 1997
(Show Context)
Citation Context ...e view of the author, it has been superseded by Schaffalitzkey, Zisserman, Hartley and Torr [101]. This latter work (ECCV2000) is derived from the 6 point approach of Quan [94] and Torr and Zisserman =-=[124]-=-, rather than the epipole optimisation technique of Hartley [46].s7.1 Calibration 115 Figure 7.1: The Master’s Lodge image sequence. This image sequence was taken in Great Court, Trinity College, Camb... |

5 |
Structure and motion from silhouettes
- K-Y, Cipolla
- 2001
(Show Context)
Citation Context ...Multiple scale reconstruction Hierarchical reconstruction techniques have been used extensively in the shape from silhouettes field (e.g. Chien and Aggarwal [22], Szeliski [118], and Wong and Cipolla =-=[130]-=-), but have not been exploited to any great extent in voxel algorithms. When considering a silhouette it is always possible to make a definites154 Conclusion decision about whether a voxel is complete... |