RL-AWB: Deep Reinforcement Learning for Auto White Balance Correction in Low-Light Night-time Scenes
Yuan-Kang Lee, Kuan-Lin Chen, Chia-Che Chang, Yu-Lun Liu
2026-01-09
Summary
This paper tackles the tricky problem of getting colors to look right in photos taken at night, a process called color constancy. It introduces a new system, RL-AWB, that uses a combination of traditional image processing techniques and a more modern approach called deep reinforcement learning to automatically adjust the white balance in nighttime photos.
What's the problem?
Taking good photos at night is hard because there's very little light, which creates a lot of noise and makes it difficult for cameras to accurately capture colors. Existing methods often struggle with the unique challenges of nighttime scenes, where the lighting is complex and can vary greatly. Essentially, colors get distorted and don't look realistic, and it's hard for a computer to figure out what the 'true' colors should be.
What's the solution?
The researchers developed RL-AWB, which works in two main steps. First, it uses a statistical method to identify pixels that are likely to be gray and estimate the overall lighting conditions in the scene. Then, it uses deep reinforcement learning – a type of artificial intelligence – to fine-tune the color balance. This AI learns by mimicking how professional photo editors adjust white balance, dynamically changing settings for each image to get the best result. To help test their method, they also created a new dataset of nighttime photos taken with different cameras.
Why it matters?
This research is important because it improves the quality of photos taken in low-light conditions. Better color accuracy makes nighttime photos more visually appealing and useful. The new dataset they created will also help other researchers develop and test their own nighttime image processing algorithms, ultimately leading to better cameras and photos for everyone.
Abstract
Nighttime color constancy remains a challenging problem in computational photography due to low-light noise and complex illumination conditions. We present RL-AWB, a novel framework combining statistical methods with deep reinforcement learning for nighttime white balance. Our method begins with a statistical algorithm tailored for nighttime scenes, integrating salient gray pixel detection with novel illumination estimation. Building on this foundation, we develop the first deep reinforcement learning approach for color constancy that leverages the statistical algorithm as its core, mimicking professional AWB tuning experts by dynamically optimizing parameters for each image. To facilitate cross-sensor evaluation, we introduce the first multi-sensor nighttime dataset. Experiment results demonstrate that our method achieves superior generalization capability across low-light and well-illuminated images. Project page: https://ntuneillee.github.io/research/rl-awb/