## Robust Shape Recovery from Occluding Contours Using a Linear Smoother (1993)

### Cached

### Download Links

- [ftp.digital.fr]
- [vis-ftp.cs.umass.edu]
- [www.hpl.hp.com]
- DBLP

### Other Repositories/Bibliography

Citations: | 46 - 9 self |

### BibTeX

@MISC{Szeliski93robustshape,

author = {Richard Szeliski and Richard Szeliski and Richard Weiss and Richard Weiss},

title = {Robust Shape Recovery from Occluding Contours Using a Linear Smoother},

year = {1993}

}

### Years of Citing Articles

### OpenURL

### Abstract

Recovering the shape of an object from two views fails at occluding contours of smooth objects because the extremal contours are view dependent. For three or more views, shape recovery is possible, and several algorithms have recently been developed for this purpose. We present a new approach to the multiframe stereo problem which does not depend on differential measurements in the image, which may be noise sensitive. Instead, we use a linear smoother to optimally combine all of the measurements available at the contours (and other edges) in all of the images. This allows us to extract a robust and dense estimate of surface shape, and to integrate shape information from both surface markings and occluding contours. Keywords: Computer vision, image sequence analysis, motion analysis and multiframe stereo, shape and object representation, occluding contours (profiles). c flDigital Equipment Corporation 1993. All rights reserved. 1 Computer and Information Science Department, University...

### Citations

3260 | A computational approach to edge detection - Canny - 1986 |

1978 |
Numerical Recipes: The Art of Scientific Computing
- Press
- 1986
(Show Context)
Citation Context ... a larger camera displacement, but we can greatly increase the reliability of our shape estimate by 3 Alternative techniques for solving the least squares problem include singular value decomposition =-=[PFTV86]-=- and Householder transforms [Bie77]. directly solving for the surface point (x s ; y s ), where x s = x c + r and y s = y c . 4 The new set of equations is thus c i x s + s i y s + (1 \Gamma c i )r = ... |

1918 |
Robust Statistics
- Huber
- 1981
(Show Context)
Citation Context ...ty and reliability of our estimates, we can apply robust statistics to reduce the effects of outliers which are due to grossly erroneous measurements as well as large changes in the surface curvature =-=[Hub81]-=-. Many robust techniques are based on first computing residuals, r i = d i \Gamma a i \Delta x, and then re-weighting the data by a monotonic function (oe 0 i ) \Gamma2 = oe \Gamma2 i g(jr i j) or thr... |

899 | Theory of edge detection - Marr, Hildreth - 1980 |

876 | The design and use of steerable filters
- Freeman, Adelson
- 1991
(Show Context)
Citation Context ...nce we are interested mostly in detecting strong edges such as occluding contours and visible surface markings. 1 For our system, we have chosen the steerable filters developed by Freeman and Adelson =-=[FA91]-=-, since they provide good angular resolution at moderate computation cost, and since they can find both step and peak edges. We have used both the G 1 and (G 2 ; H 2 ) sets of filters, with the defaul... |

748 |
Tracking and Data Association
- Bar-Shalom, Fortmann
- 1988
(Show Context)
Citation Context ...ation gate approach used by Blake et al. for their Kalman-filter snake tracking [BCZ93]. Even more sophisticated data association techniques could be used to disambiguate multiple intersecting tracks =-=[BSF88]-=-. 6 Kalman filter The Kalman filter is a powerful technique for efficiently computing statistically optimal estimates of time-varying processes from series of noisy measurements [Gel74, Bie77, May79].... |

358 | Dynamic 3D Models with Local and Global Deformations : Deformable Superquadrics - Metaxas - 1991 |

337 | A computational theory of human stereo vision - Marr, Poggio - 1979 |

319 | Uniqueness and Estimation of Three-Dimensional Motion Parameters of Rigid Objects with Curved Surfaces - Tsai, Huang - 1984 |

302 | Fitting parameterized three-dimensional models to images
- Lowe
- 1991
(Show Context)
Citation Context ...s of noisy measurements [Gel74, Bie77, May79]. In computer vision, the Kalman filter has been applied to diverse problems such as motion recovery [RBE86], multiframe stereo [MKS89], and pose recovery =-=[Low91]-=-. In this section, we develop a Kalman filter for contour-based shape recovery in two parts: first, we show how to perform the batch fitting of the previous section incrementally; second, we show how ... |

249 | Stochastic Models, Estimation and Control, volume 1 - Maybeck - 1979 |

227 | Rapid octree construction from image sequences - Szeliski - 1993 |

216 | Kalman Filter-Based Algorithms for Estimating Depth from Image Sequences
- Matthies, Kanade, et al.
- 1989
(Show Context)
Citation Context ...arying processes from series of noisy measurements [Gel74, Bie77, May79]. In computer vision, the Kalman filter has been applied to diverse problems such as motion recovery [RBE86], multiframe stereo =-=[MKS89]-=-, and pose recovery [Low91]. In this section, we develop a Kalman filter for contour-based shape recovery in two parts: first, we show how to perform the batch fitting of the previous section incremen... |

215 | Epipolar image analysis: An approach to determine structure from motion - Bolles, Baker, et al. - 1987 |

206 |
A framework for spatio-temporal control in the tracking of visual contours
- Blake, Curwen, et al.
- 1993
(Show Context)
Citation Context ... onto the ith frame, and oe 2 x is the variance in the local x direction. Our approach is similar in spirit to the validation gate approach used by Blake et al. for their Kalman-filter snake tracking =-=[BCZ93]-=-. Even more sophisticated data association techniques could be used to disambiguate multiple intersecting tracks [BSF88]. 6 Kalman filter The Kalman filter is a powerful technique for efficiently comp... |

193 | R.:Bayesian modeling of uncertainty in low-level vision - Szeliski - 1990 |

146 | Data Fusion for Sensory Information Processing System - Clark, Yulle - 1990 |

140 |
Factorization Methods for Discrete Sequential Estimation
- Bierman
- 1977
(Show Context)
Citation Context ...e can greatly increase the reliability of our shape estimate by 3 Alternative techniques for solving the least squares problem include singular value decomposition [PFTV86] and Householder transforms =-=[Bie77]-=-. directly solving for the surface point (x s ; y s ), where x s = x c + r and y s = y c . 4 The new set of equations is thus c i x s + s i y s + (1 \Gamma c i )r = d i : (6) While there is still some... |

134 | Relative Orientation - Horn - 1990 |

127 | Regression and the Moore-Penrose pseudoinverse - Albert - 1972 |

126 |
Surface shape from the deformation of apparent contours
- Cipolla, Blake
- 1992
(Show Context)
Citation Context ...ating circle, and the plane of this circle is called the osculating plane. It is easy to see that if the epipolar curve is non-singular, then the epipolar plane is an estimate of its osculating plane =-=[CB92]-=-, and the lines defined by the view directions can be projected onto this plane. The relationship between the curvature of a curve such as the epipolar curve and the curvature of the surface is determ... |

105 | Using extremal boundaries for 3-D object modeling - Vaillant, Faugeras - 1992 |

73 |
Modeling surfaces of arbitrary topology with dynamic particles
- Szeliski, Tonnesen, et al.
- 1993
(Show Context)
Citation Context ...-D data. An alternative, which does not suffer from the restrictions on topology imposed by previous techniques, is to use a selftriangulating system of particles to model and interpolate the surface =-=[STT93]-=-. We plan to investigate the intergration of this system with our multiframe stereo algorithm in future work. 9 Experimental results To determine the performance of our shape reconstruction algorithm,... |

65 |
Reconstruction of surfaces from profiles
- Giblin, Weiss
- 1987
(Show Context)
Citation Context ...termined, i.e., one obtains a surface patch rather than a set of points. The problem of reconstructing a smooth surface from its profiles has been explored for known planar motion by Giblin and Weiss =-=[GW87]-=- and subsequently for more general known motion by Vaillant and Faugeras [Vai90, VF92] and Cipolla and Blake [BC90, CB90, CB92]. These approaches are either based on a differential formulation and ana... |

58 | The dynamic analysis of apparent contours - Cipolla, Blake - 1990 |

53 | Generalizing Epipolar-Plane Image Analysis on the Spatiotemporal Surface - Baker, Bolles |

44 | Motion and structure from motion from point and line matching - Faugeras, Lustman, et al. - 1987 |

34 | Closed-form solutions for physically-based shape modeling and recognition - Pentland, Sclaroff - 1991 |

27 | Shape from rotation - Szeliski - 1991 |

20 |
Building surfaces of evolution: The weaving wall
- Baker
- 1989
(Show Context)
Citation Context ... on the contour. While we do not operate directly on the spatiotemporal volume, our tracking and contour linking processes form a virtual surface similar to the weaving wall technique of Harlyn Baker =-=[Bak89]-=-. Unlike Baker's technique, however, we do not assume a regular and dense sampling in time. 3 Reconstructing surface patches The surface being reconstructed from a moving camera can be parametrized in... |

18 | Robust Estimation of Surface Curvature from Deformation of Apparent Contours - Blake, Cipolla - 1989 |

13 | Automated Stereo Perception - Arnold - 1983 |

13 | Using Occluding Contours for 3D Object Modeling - Vaillant, Faugeras - 1992 |

13 | Epipolar fields on surfaces - Giblin, Weiss - 1994 |

7 | An egomotion algorithm based on the tracking of arbitrary curves - Arbogast, Mohr - 1992 |

7 | The computation of binocular edges - Mayhew, Frisby - 1980 |

6 |
Real-time visual tracking for surveillance and path planning
- Curwen, Blake, et al.
- 1992
(Show Context)
Citation Context ...nications. The most traditional application of visually based shape recovery is in the reconstruction of a mobile robot's environment, which allows it to perform obstacle avoidance and planning tasks =-=[CBZ92]. Visually-=- based shape recovery can also be used to develop strategies for robotics grasping and manipulation tasks, or as an off-line technique to automatically "learn" object descriptions for object... |

3 |
Recursive estimation of 3d features using optical flow and camera motion
- Rives, Breuil, et al.
- 1986
(Show Context)
Citation Context ...optimal estimates of time-varying processes from series of noisy measurements [Gel74, Bie77, May79]. In computer vision, the Kalman filter has been applied to diverse problems such as motion recovery =-=[RBE86]-=-, multiframe stereo [MKS89], and pose recovery [Low91]. In this section, we develop a Kalman filter for contour-based shape recovery in two parts: first, we show how to perform the batch fitting of th... |

2 | Moving surfaces - Giblin, Rycroft, et al. - 1993 |

1 |
Reconstructing surfaces from known camera motion. (in preparation
- Giblin, Weiss
- 1993
(Show Context)
Citation Context ...rity of the profile. In addition, the epipolar curve can end where the normal to the profile is parallel to the instantaneous axis of rotation or where the critical set is occluded as at a T-junction =-=[GW93]-=-. Given three or more edgels tracked with our technique, we would like to compute the location of the surface and its curvature by fitting a circular arc to the lines defined by the view directions at... |

1 | Robust estimation of surface curvature from 10 Discussion and Conclusion 25 deformation of apparent contours - Blake, Cipolla - 1990 |