Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising
Practical low-light denoising/enhancement solutions often require fast computation,high memory efficiency,and can achieve visually high-quality restoration results.Most existing methods aim to restore quality but compromise on speed and memory requirements,which limits their usefulness to a large ex...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial office of Computer Science
2022-08-01
|
| Series: | Jisuanji kexue |
| Subjects: | |
| Online Access: | https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-49-8-120.pdf |
| _version_ | 1797845120030081024 |
|---|---|
| author | WEI Kai-xuan, FU Ying |
| author_facet | WEI Kai-xuan, FU Ying |
| author_sort | WEI Kai-xuan, FU Ying |
| collection | DOAJ |
| description | Practical low-light denoising/enhancement solutions often require fast computation,high memory efficiency,and can achieve visually high-quality restoration results.Most existing methods aim to restore quality but compromise on speed and memory requirements,which limits their usefulness to a large extent.This paper proposes a new deep denoising architecture,a re-parameterized multi-scale fusion network for extreme low-light raw denoising,which greatly improves the inference speed without losing high-quality denoising performance.Specifically,image features are extracted in multi-scale space,and a lightweight spatial-channel parallel attention module is used to focus on core features within space and channel dynamically and adaptively.The representation ability of the model is further enriched by re-parameterized convolutional unit without increasing computational cost at inference.The proposed model can restore UHD 4K resolution images within about 1s on a CPU(e.g.,Intel i7-7700K) and run at 24 fps on a GPU(e.g.,NVIDIA GTX 1080Ti),which is almost four times faster than existing advanced methods(e.g.,UNet) while still maintaining competitive restoration quality. |
| first_indexed | 2024-04-09T17:34:46Z |
| format | Article |
| id | doaj.art-41ae4b967d2342c88db000bdbe1a0863 |
| institution | Directory Open Access Journal |
| issn | 1002-137X |
| language | zho |
| last_indexed | 2024-04-09T17:34:46Z |
| publishDate | 2022-08-01 |
| publisher | Editorial office of Computer Science |
| record_format | Article |
| series | Jisuanji kexue |
| spelling | doaj.art-41ae4b967d2342c88db000bdbe1a08632023-04-18T02:32:21ZzhoEditorial office of Computer ScienceJisuanji kexue1002-137X2022-08-0149812012610.11896/jsjkx.220200179Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw DenoisingWEI Kai-xuan, FU Ying0School of Computer Science and Technology,Beijing Institute of Technology,Beijing 100081,ChinaPractical low-light denoising/enhancement solutions often require fast computation,high memory efficiency,and can achieve visually high-quality restoration results.Most existing methods aim to restore quality but compromise on speed and memory requirements,which limits their usefulness to a large extent.This paper proposes a new deep denoising architecture,a re-parameterized multi-scale fusion network for extreme low-light raw denoising,which greatly improves the inference speed without losing high-quality denoising performance.Specifically,image features are extracted in multi-scale space,and a lightweight spatial-channel parallel attention module is used to focus on core features within space and channel dynamically and adaptively.The representation ability of the model is further enriched by re-parameterized convolutional unit without increasing computational cost at inference.The proposed model can restore UHD 4K resolution images within about 1s on a CPU(e.g.,Intel i7-7700K) and run at 24 fps on a GPU(e.g.,NVIDIA GTX 1080Ti),which is almost four times faster than existing advanced methods(e.g.,UNet) while still maintaining competitive restoration quality.https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-49-8-120.pdfre-parameterization convolutional unit|multi-scale fusion|spatial-channel parallel attention module|extreme low-light denoising |
| spellingShingle | WEI Kai-xuan, FU Ying Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising Jisuanji kexue re-parameterization convolutional unit|multi-scale fusion|spatial-channel parallel attention module|extreme low-light denoising |
| title | Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising |
| title_full | Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising |
| title_fullStr | Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising |
| title_full_unstemmed | Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising |
| title_short | Re-parameterized Multi-scale Fusion Network for Efficient Extreme Low-light Raw Denoising |
| title_sort | re parameterized multi scale fusion network for efficient extreme low light raw denoising |
| topic | re-parameterization convolutional unit|multi-scale fusion|spatial-channel parallel attention module|extreme low-light denoising |
| url | https://www.jsjkx.com/fileup/1002-137X/PDF/1002-137X-2022-49-8-120.pdf |
| work_keys_str_mv | AT weikaixuanfuying reparameterizedmultiscalefusionnetworkforefficientextremelowlightrawdenoising |