RASR: Retinex-Guided Adaptive One-Step Diffusion for Low-Light Image Super-Resolution

Ziyu Yue, Junran Zhang, Zhixun Su · 2026

Read RASR: Retinex-Guided Adaptive One-Step Diffusion for Low-Light Image Super-Resolution by Ziyu Yue, Junran Zhang, Zhixun Su (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/pmc-13611183.txt.

Description

Ziyu Yue and colleagues brighten and sharpen dark, blurry photos in a single diffusion step, in this study from twenty twenty-six. Low-light super-resolution is doubly cursed: little light and little detail, tangled with noise. Diffusion models restore such images beautifully but slowly, step by step. One-step shortcuts exist, yet they guess one fixed noise setting for every picture, or feed the darkness straight in and bake its bias into the result.

Their framework, RASR, first splits the photo into what things are made of and how they are lit, using the lit-independent part as the content. A teacher model supplies per-image targets, and a light student predicts each image's noise level from that split alone. One pass through a large image model then finishes the job. On the standard benchmark it beat rival methods on fidelity and perceptual quality alike.

Catalog record

CREATORZiyu Yue, Junran Zhang, Zhixun Su
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13611183

Browse the full library · About the Tower of Babel