## Manhattan World: Compass Direction from a Single Image by Bayesian Inference (1999)

### Cached

### Download Links

- [www.ski.org]
- [www.stat.ucla.edu]
- [www.stat.ucla.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | In Internation Conference on Computer Vision |

Citations: | 65 - 4 self |

### BibTeX

@INPROCEEDINGS{Coughlan99manhattanworld:,

author = {James M. Coughlan and A. L. Yuille},

title = {Manhattan World: Compass Direction from a Single Image by Bayesian Inference},

booktitle = {In Internation Conference on Computer Vision},

year = {1999},

pages = {941--947}

}

### Years of Citing Articles

### OpenURL

### Abstract

When designing computer vision systems for the blind and visually impaired it is important to determine the orientation of the user relative to the scene. We observe that most indoor and outdoor (city) scenes are designed on a Manhattan three-dimensional grid. This Manhattan grid structure puts strong constraints on the intensity gradients in the image. We demonstrate an algorithm for detecting the orientation of the user in such scenes based on Bayesian inference using statistics which we have learnt in this domain. Our algorithm requires a single input image and does not involve pre-processing stages such as edge detection and Hough grouping. We demonstrate strong experimental results on a range of indoor and outdoor images. We also show that estimating the grid structure makes it significantly easier to detect target objects which are not aligned with the grid. 1 Introduction Recently there has been growing interest in building computer vision navigational systems for the blind [9...

### Citations

8565 |
Elements of information theory
- Cover, Thomas
- 1991
(Show Context)
Citation Context ...which contain one hundred presegmented images. The more different P on is from P off then the easier edge detection becomes, see Figure 2. A suitable measure of difference is the Chernoff Information =-=[2]-=- C(P on ; P off ) = \Gamma min 01 log P y P on (y)P 1\Gamma off (y). Konishi et al tested a variety of different edge filters and ranked them by their effectiveness based on their Chernoff information... |

453 |
Three Dimensional Computer Vision
- Faugeras
- 1993
(Show Context)
Citation Context ...a single image. A useful spin-off is the ability to detect target objects which are not aligned with the Manhattan grid. Most indoor and outdoor city scenes are based on a cartesian coordinate system =-=[3, 6]-=- which we can refer to as a Manhattan grid. This grid defines an ~ i; ~ j; ~ k coordinate system. This gives a natural reference frame for the viewer. If the viewer can determine his/her position rela... |

54 | Fundamental Bounds on Edge Detection: An Information Theoretic Evaluation
- Konishi, Yuille, et al.
- 1999
(Show Context)
Citation Context ...ese distributions were very consistent for a range of images. and overexposed -- makes edge detection particularly difficult). Instead we use the power of statistics. Following work by Konishi et al. =-=[4]-=-, we determine probabilitiessP on (E ~u ) and P off (E ~u ) for the probabilities of the response E ~u of an edge filter at position ~u in the image conditioned on whether we are on or off an edge. Th... |

51 |
Contribution to the determination of vanishing points using Hough transform
- Lutton, Maitre, et al.
- 1994
(Show Context)
Citation Context ...ork and Three- Dimensional Geometry There has been an enormous amount of work in projective geometry [3, 6]. Techniques from projective geometry have been applied to finding the vanishing points [1], =-=[5]-=-. For a recent application to vision systems for the blind see [9] for the detection of pedestrian crossings using projection geometry. This work, however, has typically proceeded through the stages o... |

34 |
Aided and Automatic Target Recognition Based Upon Sensory Inputs from Image Forming Systems
- Ratches, Walters, et al.
- 1997
(Show Context)
Citation Context ...jects in background clutter. To perform such a task effectively requires modelling the properties of the background clutter in addition to those of the target object. It has recently been appreciated =-=[7]-=- that simple models of background clutter based on Gaussian probability Figure 5: Another indoor scene and its exterior. Same conventions as above. The vanishing points are estimated to within 5 ffi (... |

30 |
Robust Computation and Parameterization of Multiple View Relations
- Torr, Zisserman
- 1998
(Show Context)
Citation Context ...hrough the stages of edge detection, Hough transforms, and finally the calculation of the geometry. Alternatively, a sequence of images over time can be used to estimate the geometry, see for example =-=[8]-=-. In this paper, we demonstrate that accurate results can be obtained from a single image directly without the need for techniques such as edge detection and Hough transforms. For completeness, we giv... |

26 |
New method for vanishing point detection
- Brillault-Oâ€™Mahony
- 1991
(Show Context)
Citation Context ...ous Work and Three- Dimensional Geometry There has been an enormous amount of work in projective geometry [3, 6]. Techniques from projective geometry have been applied to finding the vanishing points =-=[1]-=-, [5]. For a recent application to vision systems for the blind see [9] for the detection of pedestrian crossings using projection geometry. This work, however, has typically proceeded through the sta... |

18 | Grouping based on projective geometry constraints and uncertainty
- Utcke
- 1998
(Show Context)
Citation Context ...gnificantly easier to detect target objects which are not aligned with the grid. 1 Introduction Recently there has been growing interest in building computer vision navigational systems for the blind =-=[9]-=-, [10]. These systems can be used, for example, for navigation and for the detection and reading of informational signs. The goal of this paper is to determine the orientation of the viewer in the sce... |

9 | Clutter modeling and performance and analysis in automatic target recognition
- Zhu, Lanterman, et al.
- 1998
(Show Context)
Citation Context ...Outdoor 12). Figure 6: Another indoor scene. Same conventions as above. (Indoor 8). distributions are often inadequate and that better performance can be obtained using alternative probability models =-=[11]-=-. The Manhattan world assumption gives an alternative way of probabilistically modelling background clutter. The background clutter will correspond to the regular structure of buildings and roads and ... |

2 |
Signfinder
- Yuille, Snow, et al.
- 1998
(Show Context)
Citation Context ...cantly easier to detect target objects which are not aligned with the grid. 1 Introduction Recently there has been growing interest in building computer vision navigational systems for the blind [9], =-=[10]-=-. These systems can be used, for example, for navigation and for the detection and reading of informational signs. The goal of this paper is to determine the orientation of the viewer in the scene (in... |