Seeing Beyond Haze: Generative Nighttime Image Dehazing

Beibei Lin Stephen Lin Robby Tan · 2025

Read Seeing Beyond Haze: Generative Nighttime Image Dehazing by Beibei Lin Stephen Lin Robby Tan (2025) 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-2503-08073.txt.

Description

Beibei Lin, Stephen Lin, and Robby Tan present a 2025 method for clearing haze from night photos. Night haze is hard because dense haze and bright glow can hide the scene behind them. Older methods often lack strong guesses about what the hidden background should look like, so they leave much of the scene unclear.

Their method, BeyondHaze, adapts a diffusion image model to the night task. The model learns night clearing skill while keeping its power to build clean images, and it trains on pairs made to boost detail in blocked areas. Users can set how much the model may invent, trading realism against faith to the source. Tests on real night photos showed much better sight of the scene.

Catalog record

CREATORBeibei Lin Stephen Lin Robby Tan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-08073

Browse the full library · About the Tower of Babel