RetinexDualV2: Physically-Grounded Dual Retinex for Generalized UHD Image Restoration
Mohab KishawyJun Chen {kishawym, chenjun}@mcmaster.ca McMaster University · 2026
Read RetinexDualV2: Physically-Grounded Dual Retinex for Generalized UHD Image Restoration by Mohab KishawyJun Chen {kishawym, chenjun}@mcmaster.ca McMaster University (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/arxiv-2603-27979.txt.
Description
Kishawy and Chen propose RetinexDualV2, a single model for restoring ultra-high-definition images. A grounding module first extracts cues tied to each flaw, such as rain masks or dark channels. Those cues steer a Retinex split of the image into reflection and light parts, which the model then corrects. One design handles rain, noise, and low light without task-specific changes. The system placed fourth and fifth in two NTIRE 2026 challenges.
Catalog record
| CREATOR | Mohab KishawyJun Chen {kishawym, chenjun}@mcmaster.ca McMaster University |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30467 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.27979