MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention
Most learning-based low-light image enhancement methods typically suffer from two problems. First, they require a large amount of paired data for training, which are difficult to acquire in most cases. Second, in the process of enhancement, image noise is difficult to be removed and may even be ampl...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Tsinghua University Press
2022-06-01
|
| Series: | Big Data Mining and Analytics |
| Subjects: | |
| Online Access: | https://www.sciopen.com/article/10.26599/BDMA.2021.9020020 |
| _version_ | 1811344618585849856 |
|---|---|
| author | Renjun Wang Bin Jiang Chao Yang Qiao Li Bolin Zhang |
| author_facet | Renjun Wang Bin Jiang Chao Yang Qiao Li Bolin Zhang |
| author_sort | Renjun Wang |
| collection | DOAJ |
| description | Most learning-based low-light image enhancement methods typically suffer from two problems. First, they require a large amount of paired data for training, which are difficult to acquire in most cases. Second, in the process of enhancement, image noise is difficult to be removed and may even be amplified. In other words, performing denoising and illumination enhancement at the same time is difficult. As an alternative to supervised learning strategies that use a large amount of paired data, as presented in previous work, this paper presents an mixed-attention guided generative adversarial network called MAGAN for low-light image enhancement in a fully unsupervised fashion. We introduce a mixed-attention module layer, which can model the relationship between each pixel and feature of the image. In this way, our network can enhance a low-light image and remove its noise simultaneously. In addition, we conduct extensive experiments on paired and no-reference datasets to show the superiority of our method in enhancing low-light images. |
| first_indexed | 2024-04-13T19:49:53Z |
| format | Article |
| id | doaj.art-38af9fceb081492ba8cc29581c4fe969 |
| institution | Directory Open Access Journal |
| issn | 2096-0654 |
| language | English |
| last_indexed | 2024-04-13T19:49:53Z |
| publishDate | 2022-06-01 |
| publisher | Tsinghua University Press |
| record_format | Article |
| series | Big Data Mining and Analytics |
| spelling | doaj.art-38af9fceb081492ba8cc29581c4fe9692022-12-22T02:32:34ZengTsinghua University PressBig Data Mining and Analytics2096-06542022-06-015211011910.26599/BDMA.2021.9020020MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-AttentionRenjun Wang0Bin Jiang1Chao Yang2Qiao Li3Bolin Zhang4College of Computer Science and Electronic Engineering, Hunan University, Changsha 410006, ChinaCollege of Computer Science and Electronic Engineering, Hunan University, Changsha 410006, ChinaCollege of Computer Science and Electronic Engineering, Hunan University, Changsha 410006, ChinaCollege of Computer Science and Electronic Engineering, Hunan University, Changsha 410006, ChinaCollege of Computer Science and Electronic Engineering, Hunan University, Changsha 410006, ChinaMost learning-based low-light image enhancement methods typically suffer from two problems. First, they require a large amount of paired data for training, which are difficult to acquire in most cases. Second, in the process of enhancement, image noise is difficult to be removed and may even be amplified. In other words, performing denoising and illumination enhancement at the same time is difficult. As an alternative to supervised learning strategies that use a large amount of paired data, as presented in previous work, this paper presents an mixed-attention guided generative adversarial network called MAGAN for low-light image enhancement in a fully unsupervised fashion. We introduce a mixed-attention module layer, which can model the relationship between each pixel and feature of the image. In this way, our network can enhance a low-light image and remove its noise simultaneously. In addition, we conduct extensive experiments on paired and no-reference datasets to show the superiority of our method in enhancing low-light images.https://www.sciopen.com/article/10.26599/BDMA.2021.9020020low-light image enhancementunsupervised learninggenerative adversarial network (gan)mixed-attention |
| spellingShingle | Renjun Wang Bin Jiang Chao Yang Qiao Li Bolin Zhang MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention Big Data Mining and Analytics low-light image enhancement unsupervised learning generative adversarial network (gan) mixed-attention |
| title | MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention |
| title_full | MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention |
| title_fullStr | MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention |
| title_full_unstemmed | MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention |
| title_short | MAGAN: Unsupervised Low-Light Image Enhancement Guided by Mixed-Attention |
| title_sort | magan unsupervised low light image enhancement guided by mixed attention |
| topic | low-light image enhancement unsupervised learning generative adversarial network (gan) mixed-attention |
| url | https://www.sciopen.com/article/10.26599/BDMA.2021.9020020 |
| work_keys_str_mv | AT renjunwang maganunsupervisedlowlightimageenhancementguidedbymixedattention AT binjiang maganunsupervisedlowlightimageenhancementguidedbymixedattention AT chaoyang maganunsupervisedlowlightimageenhancementguidedbymixedattention AT qiaoli maganunsupervisedlowlightimageenhancementguidedbymixedattention AT bolinzhang maganunsupervisedlowlightimageenhancementguidedbymixedattention |