File Download
There are no files associated with this item.
Links for fulltext
(May Require Subscription)
- Publisher Website: 10.1109/ICIP46576.2022.9898028
- Scopus: eid_2-s2.0-85146646386
- WOS: WOS:001058109502029
Supplementary
- Citations:
- Appears in Collections:
Conference Paper: Manet: Improving Video Denoising with a Multi-Alignment Network
Title | Manet: Improving Video Denoising with a Multi-Alignment Network |
---|---|
Authors | |
Keywords | attention image alignment image synthesis video denoising video enhancement |
Issue Date | 16-Oct-2022 |
Abstract | In video denoising, the adjacent frames often provide very useful information, but accurate alignment is needed before such information can be harnassed. In this work, we present a multi-alignment network, which generates multiple flow proposals followed by attention-based averaging. It serves to mimic the non-local mechanism, suppressing noise by averaging multiple observations. Our approach can be applied to various state-of-the-art models that are based on flow estimation. Experiments on a large-scale video dataset demonstrate that our method improves the denoising baseline model by 0.2 dB, and further reduces the parameters by 47% with model distillation. Code is available at https://github.com/IndigoPurple/MANet. |
Persistent Identifier | http://hdl.handle.net/10722/333718 |
ISI Accession Number ID |
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Zhao, Yaping | - |
dc.contributor.author | Zheng, Haitian | - |
dc.contributor.author | Wang, Zhongrui | - |
dc.contributor.author | Luo, Jiebo | - |
dc.contributor.author | Lam, Edmund | - |
dc.date.accessioned | 2023-10-06T08:38:31Z | - |
dc.date.available | 2023-10-06T08:38:31Z | - |
dc.date.issued | 2022-10-16 | - |
dc.identifier.uri | http://hdl.handle.net/10722/333718 | - |
dc.description.abstract | <p>In video denoising, the adjacent frames often provide very useful information, but accurate alignment is needed before such information can be harnassed. In this work, we present a multi-alignment network, which generates multiple flow proposals followed by attention-based averaging. It serves to mimic the non-local mechanism, suppressing noise by averaging multiple observations. Our approach can be applied to various state-of-the-art models that are based on flow estimation. Experiments on a large-scale video dataset demonstrate that our method improves the denoising baseline model by 0.2 dB, and further reduces the parameters by 47% with model distillation. Code is available at https://github.com/IndigoPurple/MANet.<br></p> | - |
dc.language | eng | - |
dc.language | eng | - |
dc.relation.ispartof | 2022 IEEE International Conference on Image Processing (ICIP) (16/10/2022-19/10/2022, Bordeaux, France) | - |
dc.subject | attention | - |
dc.subject | image alignment | - |
dc.subject | image synthesis | - |
dc.subject | video denoising | - |
dc.subject | video enhancement | - |
dc.title | Manet: Improving Video Denoising with a Multi-Alignment Network | - |
dc.type | Conference_Paper | - |
dc.identifier.doi | 10.1109/ICIP46576.2022.9898028 | - |
dc.identifier.scopus | eid_2-s2.0-85146646386 | - |
dc.identifier.spage | 2036 | - |
dc.identifier.epage | 2040 | - |
dc.identifier.isi | WOS:001058109502029 | - |