Astro-SReC: attention-enhanced neural networks for lossless compression of super-resolution solar observations
文献类型：期刊论文
| 作者 | Wu, Shichao1; Liu, Yingbo1; Zeng, Li2; Ma, Xuan1; Yang L(杨磊)3 |
| 刊名 | EXPERT SYSTEMS WITH APPLICATIONS
 |
| 出版日期 | 2026-09-01 |
| 卷号 | 325 |
| 关键词 | Lossless compression Astro-SReC Super-resolution Attention Sparse and low rank |
| ISSN号 | 0957-4174 |
| DOI | 10.1016/j.eswa.2026.132608 |
| 产权排序 | 第3完成单位 |
| 文献子类 | Article |
| 英文摘要 | Modern observational instruments produce massive super-resolution (SR) data that require domain-specific compression solutions beyond general compression methods. This paper presents Astro-SReC, an SR lossless compression framework designed for solar images. It integrates efficient channel and shuffle attention mechanisms to capture fine-grained solar features, employs a surrogate gradient for ReLU activation functions to maintain network expressiveness, and leverages sparse and low-rank decomposition to model redundant structures and complex textures. This architecture enables adaptive compression that preserves critical astronomical features on multiple scales while maximizing redundancy reduction. We evaluated Astro-SReC on solar observation datasets with 1911 & times; 1911 pixel resolution, where it achieved a 5.46% reduction in bits per subpixel (bpsp) compared to baseline SReC model compression while maintaining an average compression time of 1.59 seconds. Among deep compression models, it also achieves a 57.54% reduction in bpsp relative to LC-FDNet. The framework also generalizes to natural images, achieving 1.21% and 2.50% improvements on the DIV2K and Flickr2K datasets, respectively. These results offer a new approach to astronomical data compression under the growing data demands of modern solar observatories. |
| 学科主题 | 天文学 ; 太阳与太阳系 |
| URL标识 | 查看原文 |
| 出版地 | THE BOULEVARD, LANGFORD LANE, KIDLINGTON, OXFORD OX5 1GB, ENGLAND |
| WOS关键词 | IMAGE COMPRESSION |
| 资助项目 | National Natural Science Foundation of China[62262068]; National Natural Science Foundation of China[62462064]; Yunnan University of Finance and Economics Postgraduate Innovation Foundation[2025YUFEYC018]; Yunnan University of Finance and Economics Postgraduate Innovation Foundation[2025YUFEYC116]; Yunnan Fundamental Research Projects[202301AT070417] |
| WOS研究方向 | Computer Science ; Engineering ; Operations Research & Management Science |
| 语种 | 英语 |
| WOS记录号 | WOS:001761684700001 |
| 出版者 | PERGAMON-ELSEVIER SCIENCE LTD |
| 资助机构 | National Natural Science Foundation of China[62262068, 62462064] ; Yunnan University of Finance and Economics Postgraduate Innovation Foundation[2025YUFEYC018, 2025YUFEYC116] ; Yunnan Fundamental Research Projects[202301AT070417] |
| 版本 | 出版稿 |
| 源URL | [http://ir.ynao.ac.cn/handle/114a53/29199]  |
| 专题 | 云南天文台_抚仙湖太阳观测站 |
| 通讯作者 | Liu, Yingbo |
| 作者单位 | 1.Yunnan University of Finance and Economics, 237 Longquan Road, Kunming, 650221, Yunnan, China; 2.School of Mathematics and Statistics, Honghe University, Mengzi, 661100, Yunnan, China; 3.Yunnan Observatories, Chinese Academy of Sciences, P.0.Box110, Kunming, 650011, Yunnan, China |
| 推荐引用方式 GB/T 7714 | Wu, Shichao,Liu, Yingbo,Zeng, Li,et al. Astro-SReC: attention-enhanced neural networks for lossless compression of super-resolution solar observations[J]. EXPERT SYSTEMS WITH APPLICATIONS,2026,325. |
| APA | Wu, Shichao,Liu, Yingbo,Zeng, Li,Ma, Xuan,&杨磊.(2026).Astro-SReC: attention-enhanced neural networks for lossless compression of super-resolution solar observations.EXPERT SYSTEMS WITH APPLICATIONS,325. |
| MLA | Wu, Shichao,et al."Astro-SReC: attention-enhanced neural networks for lossless compression of super-resolution solar observations".EXPERT SYSTEMS WITH APPLICATIONS 325(2026). |
入库方式： OAI收割
来源：云南天文台
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。