Rethinking Low-Light Image Enhancement: A Log-Domain Intensity--Chromaticity Decoupling Perspective
For computer vision practitioners, this method offers a new perspective on low-light enhancement that reduces chromatic noise and improves downstream task performance.
The paper proposes a log-domain intensity-chromaticity decoupling method for low-light image enhancement, achieving 29.71 dB PSNR and 0.89 SSIM on LOLv2-Real, with improved downstream face detection on DarkFace.
Explicit reconstruction constraints derived from the decoupled representation are further imposed to suppress abnormal channel amplification and chromatic noise. Experiments on LOLv2-Real, MIT-Adobe FiveK, and LSRW show that the proposed method achieves competitive or superior quantitative and visual performance, reaching 29.71 dB PSNR and 0.89 SSIM on LOLv2-Real. DarkFace experiments further indicate improved downstream face detection under low-light conditions. Code and pretrained models are available at: https://github.com/mubaisam/ICD.