Hybrid-domain attention learning for lightweight underwater image enhancement and real-time vision deployment
Shan Jiang, Chenfeng Sun, Xuan Liu · 2026
Read Hybrid-domain attention learning for lightweight underwater image enhancement and real-time vision deployment by Shan Jiang, Chenfeng Sun, Xuan Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13574762.txt.
Description
Shan Jiang, with two colleagues, presents a 2026 lightweight network that cleans up underwater photos for robots and divers. Water warps colour, flattens contrast, and blurs detail, yet heavy models cannot run on small underwater gear. The network aims for sharp results at a cost low enough for live use.
The design joins two views of each image: its frequency pattern, which holds broad colour casts, and its spatial layout, which holds edges and detail. Attention blocks in both views, plus varied kernel shapes, balance quality against compute. On public test sets, the method beat mainstream rivals on two standard quality scores while using less compute. Code is posted in public for others to verify. The authors target marine exploration and robot vision on tight hardware.
Catalog record
| CREATOR | Shan Jiang, Chenfeng Sun, Xuan Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72331 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |