RL-AWB: Deep Reinforcement Learning for Auto White Balance Correction in Low-Light Night-time Scenes

Yuan-Kang Lee 1,3 Kuan-Lin Chen 2,3 Chia-Che Chang 1 Yu-Lun Liu 3 1 MediaTek Inc. 2 National Taiwan University 3 National Yang Ming Chiao Tung University · 2026

Read RL-AWB: Deep Reinforcement Learning for Auto White Balance Correction in Low-Light Night-time Scenes by Yuan-Kang Lee 1,3 Kuan-Lin Chen 2,3 Chia-Che Chang 1 Yu-Lun Liu 3 1 MediaTek Inc. 2 National Taiwan University 3 National Yang Ming Chiao Tung 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-2601-05249.txt.

Description

Yuan-Kang Lee and 3 co-authors correct the colours of night photographs with reinforcement learning in a 2026 paper. Auto white balance fails at night because low light brings heavy noise and complex mixed illumination, so photographs keep unwanted colour casts. Their framework, RL-AWB, starts from a statistical algorithm tailored to night scenes, combining salient grey pixel detection with a new illuminant estimation method.

On that foundation they build what they describe as the first deep reinforcement learning approach to colour constancy, using the statistical method as its core and mimicking professional white balance workflows. Better night colours matter for phone photography, security cameras, and vehicle vision, where decisions and memories alike depend on scenes looking as they truly were.

Catalog record

CREATORYuan-Kang Lee 1,3 Kuan-Lin Chen 2,3 Chia-Che Chang 1 Yu-Lun Liu 3 1 MediaTek Inc. 2 National Taiwan University 3 National Yang Ming Chiao Tung University
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-05249

Browse the full library · About the Tower of Babel