Name
Abstract | ||
|---|---|---|
Layered models allow scene segmentation and motion estimation to be formulated together and to inform one another. Traditional layered motion methods, however, employ fairly weak models of scene structure, relying on locally connected Ising/Potts models which have limited ability to capture long-range correlations in natural scenes. To address this, we formulate a fully-connected layered model that enables global reasoning about the complicated segmentations of real objects. Optimization with fully-connected graphical models is challenging, and our inference algorithm leverages recent work on efficient mean field updates for fully-connected conditional random fields. These methods can be implemented efficiently using high-dimensional Gaussian filtering. We combine these ideas with a layered flow model, and find that the long-range connections greatly improve segmentation into figure-ground layers when compared with locally connected MRF models. Experiments on several benchmark datasets show that the method can recover fine structures and large occlusion regions, with good flow accuracy and much lower computational cost than previous locally-connected layered models. |
Year | DOI | Venue |
|---|---|---|
2013 | 10.1109/CVPR.2013.317 | CVPR |
Keywords | Field | DocType |
optimisation,high-dimensional gaussian filtering,fully-connected graphical models,previous locally-connected layered model,mrf model,locally connected mrf models,occlusion regions,inference mechanisms,scene segmentation,layered model,foreground flow,image segmentation,long-range connections,traditional layered motion method,fully-connected conditional random fields,global reasoning,layered flow model,fully-connected graphical model,long-range correlations,motion estimation,inference algorithm,layered motion methods,complicated segmentation,locally-connected layered models,figure-ground layers,gaussian processes,natural scenes,benchmark testing,ising-potts models,background flow,real object segmentations,fully-connected conditional random field,scene structure,fully-connected layered model,markov processes,potts model,correlation methods,computational modeling,optical imaging,approximation algorithms,estimation | Conditional random field,Computer vision,Pattern recognition,Computer science,Segmentation,Filter (signal processing),Image segmentation,Gaussian,Gaussian process,Artificial intelligence,Motion estimation,Graphical model | Conference |
Volume | Issue | ISSN |
2013 | 1 | 1063-6919 |
Citations | PageRank | References |
34 | 1.03 | 29 |
Authors | ||
5 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Deqing Sun | 1 | 1061 | 44.84 |
| Jonas Wulff | 2 | 438 | 17.59 |
| Erik B. Sudderth | 3 | 1420 | 119.04 |
| Hanspeter Pfister | 4 | 5933 | 340.59 |
| Michael J. Black | 5 | 11233 | 1536.41 |