## Layered Representation of Motion Video using Robust Maximum-Likelihood Estimation of Mixture Models and MDL Encoding (1995)

### Cached

### Download Links

- [eagle.almaden.ibm.com]
- [www.sarnoff.com]
- DBLP

### Other Repositories/Bibliography

Citations: | 196 - 4 self |

### BibTeX

@INPROCEEDINGS{Ayer95layeredrepresentation,

author = {Serge Ayer and Harpreet S. Sawhney},

title = {Layered Representation of Motion Video using Robust Maximum-Likelihood Estimation of Mixture Models and MDL Encoding},

booktitle = {},

year = {1995},

pages = {777--784}

}

### Years of Citing Articles

### OpenURL

### Abstract

Representing and modeling the motion and spatial support of multiple objects and surfaces from motion video sequences is an important intermediate step towards dynamic image understanding. One such representation, called layered representation, has recently been proposed. Although a number of algorithms have been developed for computing these representations, there has not been a consolidated effort into developing a precise mathematical formulation of the problem. This paper presents such a formulation based on maximum likelihood estimation of mixture models and the minimum description length (MDL) encoding principle. The three major issues in layered motion representation are: (i) how many motion models adequately describe image motion, (ii) what are the motion model parameters, and (iii) what is the spatial support layer for each motion model. In order to allow multiple models in the description of image motion, the likelihood function for change in intensity of a pixel is modeled a...

### Citations

3768 |
Stochastic relaxation, Gibbs distributions and the Bayesian restoration of images
- Geman, Geman
- 1984
(Show Context)
Citation Context ...rojected down from the next coarser level. The local optimization step implements an MRF based smoothness criterion in the estimation of support layers. A first order 2-clique neighborhood is defined =-=[7]-=-. In a scan-line order, we perform a local minimization at each point following equation (14). In other words, this step tries to find the locally optimal labeling by taking into account both the resi... |

591 |
Robust Statistics: An Approach Based on Influence Functions
- Hample, ronchetti, et al.
(Show Context)
Citation Context ... in equation (9) is normal, but the M-estimation is done through a robust estimator rather than the LS estimator corresponding to the normal density. (For a detailed account of robust estimators, see =-=[8]-=- and [14].) We have experimented with two ae functions, the Lorentzian and the Geman & McClure function [3]: ae LO (r; oe) = log(1 + 1 2 r 2 oe 2 ); ae GM (r; oe) = r 2 oe 2 1 + r 2 oe 2 : The ae LO a... |

582 | H.: Hierarchical model-based motion estimation
- BERGEN, ANANDAN, et al.
- 1992
(Show Context)
Citation Context ...he log-likelihood term for the measurements conditioned on the models being derived using the mixture model formulation. Our algorithm implements the ML estimation using a direct, hierarchical method =-=[2]-=-. Amongst the algorithms that compute layered representations using simultaneous (and not sequential) multiple motion estimation, ours is the only one that we know of in which support layers at the co... |

432 | Efficient and Effective Querying by Image Content
- Faloutsos, Barber, et al.
- 1994
(Show Context)
Citation Context ...ir change over time can then be used for contentbased querying and indexing. Video indexing systems in the spirit of static image indexing like in the Photobook [17] and Query-By-Image-Content (QBIC) =-=[6]-=- systems can effectively use layered description as an intermediate representation of video content. A number of algorithms have been developed for computing these representations [1, 5, 9, 10, 24]. H... |

222 | Constructing Simple Stable Descriptions for Image Partitioning
- Leclerc
- 1989
(Show Context)
Citation Context ...inimizes the encoding length of the model parameters and the MLE residuals. This is a generalization of the encoding length computed using the log-likelihood of the Maximum APosteriori (MAP) estimate =-=[13]-=-, with the log-likelihood term for the measurements conditioned on the models being derived using the mixture model formulation. Our algorithm implements the ML estimation using a direct, hierarchical... |

146 | Mixture models for optical flow computation
- Jepson, Black
- 1993
(Show Context)
Citation Context ...ever they did not address the problem of automatic determination of the appropriate number of models. Their results were shown using 2D affine flow computed in hand-selected regions. Jepson and Black =-=[11]-=- applied the EM algorithm using the mixture model formulation to optical flow computation without estimating the number of models. Also, they used a 2D translational model only. 3 Overview of the form... |

111 |
Detecting and tracking multiple moving objects using temporal integration
- Irani, Rousso, et al.
- 1992
(Show Context)
Citation Context ...e-Content (QBIC) [6] systems can effectively use layered description as an intermediate representation of video content. A number of algorithms have been developed for computing these representations =-=[1, 5, 9, 10, 24]-=-. However, there has not been a 1 consolidated effort into developing a precise mathematical formulation of the problem. This paper presents such a formulation based on maximum likelihood estimation o... |

95 |
Robust estimation of a multilayered motion representation
- Darrel, Pentland
- 1991
(Show Context)
Citation Context ...e-Content (QBIC) [6] systems can effectively use layered description as an intermediate representation of video content. A number of algorithms have been developed for computing these representations =-=[1, 5, 9, 10, 24]-=-. However, there has not been a 1 consolidated effort into developing a precise mathematical formulation of the problem. This paper presents such a formulation based on maximum likelihood estimation o... |

76 | Direct recovery of shape from multiple views: A parallax based approach
- Kumar, Anandan, et al.
- 1994
(Show Context)
Citation Context ...also an important issue. Research towards describing image motion with multiple parametric models and the residuals [4], or in terms of a 3D model that combines a 2D parametric and 3D parallax models =-=[12, 20, 22]-=-, is in progress. However, creating compact layered descriptions using these representations still needs to be addressed. In the context of the specific algorithm proposed in this paper, we are furthe... |

37 |
Segmentation of moving objects by robust motion parameter estimation over multiple frames
- Ayer, Schroeter, et al.
- 1994
(Show Context)
Citation Context ...e-Content (QBIC) [6] systems can effectively use layered description as an intermediate representation of video content. A number of algorithms have been developed for computing these representations =-=[1, 5, 9, 10, 24]-=-. However, there has not been a 1 consolidated effort into developing a precise mathematical formulation of the problem. This paper presents such a formulation based on maximum likelihood estimation o... |

33 | Accurate computation of optical flow by using layered motion representations
- Hsu, Anandan, et al.
- 1994
(Show Context)
Citation Context |

33 |
Robust Regression
- Li
- 1985
(Show Context)
Citation Context ...ied to the encoding of the model parameters and the maximum-likelihood estimates of the residuals. The important problem of estimating the threshold (or the scale parameter of the influence function) =-=[14]-=- for the truncated quadratic is not addressed in their approach. Moreover, the truncated quadratic is too severe and can reject non-outliers too especially in the initial stages of the iterative proce... |

24 |
The robust estimation of multiple motions: Affine and piecewise-smooth flow fields Xerox Systems Practices Lab
- Black, Anandan
- 1993
(Show Context)
Citation Context ...nd Adelson, and Irani and al. They do not, however, give a firm quantitative formulation of the framework. In this work, we have formally addressed all the issues discussed by them. Black and Anandan =-=[3]-=- incorporated robust M-estimators in their solution to multiple motion estimation, again through successive separation of dominant motions. They dwell at length on the advantage of using redescending ... |

17 | Mixture models for optical ow computation - Jepson, Black - 1993 |

16 | Robust Statistics: The Approach Based on - Hampel, Rousseeuw, et al. - 1986 |

12 | Estimating multiple independent motions in segmented images using parametric models with local deformations
- Black, Jepson
- 1994
(Show Context)
Citation Context ...faces. The viability of using a combination of 2D and 3D models, as appropriate, is also an important issue. Research towards describing image motion with multiple parametric models and the residuals =-=[4]-=-, or in terms of a 3D model that combines a 2D parametric and 3D parallax models [12, 20, 22], is in progress. However, creating compact layered descriptions using these representations still needs to... |

6 | Accurate computation of optical ow by using layered motion representation - Hsu, Anandan, et al. - 1994 |

3 | The robust estimation of multiple motions: A ne and piecewisesmooth ow elds - Black, Anandan - 1996 |

1 | cient and e ective querying by image content - unknown authors - 1994 |