Name
Abstract | ||
|---|---|---|
We study an image denoising problem: Given a set of image denoisers, each having a different denoising capability, can we design a framework that allows us to integrate the individual denoisers to produce an overall better result? If we can do so, then potentially we can integrate multiple weak denoisers to denoise complex scenes. The goal of this paper is to present a meta-procedure called the Consensus Neural Network (ConsensusNet). Given a set of initial denoisers, ConsensusNet takes the initial estimates and generates a linear combination of the results. The combined estimate is then fed to a booster neural network to reduce the amount of method noise. ConsensusNet is a modular framework that allows any image denoiser to be used in the initial stage. Experimental results show that ConsensusNet can consistently improve denoising performance for both deterministic denoisers and neural network denoisers. |
Year | Venue | Field |
|---|---|---|
2017 | arXiv: Computer Vision and Pattern Recognition | Noise reduction,Linear combination,Pattern recognition,Computer science,Image denoising,Artificial intelligence,Modular design,Artificial neural network,Machine learning |
DocType | Volume | Citations |
Journal | abs/1711.06712 | 1 |
PageRank | References | Authors |
0.36 | 32 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Choi, Joon Hee | 1 | 52 | 2.51 |
| Omar A. Elgendy | 2 | 64 | 4.68 |
| Stanley H. Chan | 3 | 403 | 30.95 |