𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡丨Physics-Guided RL Algorithm for Image Dehazing

AI Researcher & Co-author

𝐌𝐨𝐭𝐢𝐯𝐚𝐭𝐢𝐨𝐧 & 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞: Unpaired dehazing needs reliable restoration without costly diffusion sampling. For this, our work Dehaze-RL learns physical parameters, enhancement factors and spatial gating, guided by perceptual rewards.

𝐑𝐞𝐬𝐮𝐥𝐭𝐬 & 𝐕𝐞𝐧𝐮𝐞𝐬: Improved restoration quality with a physics-grounded hybrid action space and multi-granularity rewards. Accepted at ECCV 2026.

𝐋𝐢𝐧𝐤：https://media.eventhosts.cc/Conferences/ECCV2026/pdfs/3070.pdf