Name
Abstract | ||
|---|---|---|
This paper presents Bayesian edge inference (BE!), a single-frame super-resolution method explicitly grounded in Bayesian inference that addresses issues common to existing methods. Though the best give excellent results at modest magnification factors, they suffer from gradient stepping and boundary coherence problems by factors of 4x. Central to BEI is a causal framework that allows image capture and recapture to be modeled differently, a principled way of undoing downsampling blur, and a technique for incorporating Markov random field potentials arbitrarily into Bayesian networks. Besides addressing gradient and boundary issues, BEI is shown to be competitive with existing methods on published correctness measures. The model and framework are shown to generalize to other reconstruction tasks by demonstrating BEI's effectiveness at CCD demosaicing and inpainting with only trivial changes. |
Year | DOI | Venue |
|---|---|---|
2009 | 10.1109/CVPRW.2009.5206691 | CVPR: 2009 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, VOLS 1-4 |
Keywords | Field | DocType |
image resolution,bayesian network,bayesian methods,reconstruction,image processing,magnification,image reconstruction,data mining,bayesian inference,markov processes,super resolution,kernel,bayesian,image | Computer vision,Bayesian inference,Markov process,Pattern recognition,Markov random field,Computer science,Inference,Inpainting,Demosaicing,Bayesian network,Artificial intelligence,Bayesian probability | Conference |
ISSN | Citations | PageRank |
1063-6919 | 3 | 0.43 |
References | Authors | |
31 | 4 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Neil Toronto | 1 | 19 | 3.22 |
| Bryan S. Morse | 2 | 672 | 90.28 |
| Kevin D. Seppi | 3 | 335 | 41.46 |
| Dan Ventura | 4 | 121 | 15.55 |