Rethinking Low-Light Image Enhancement: A Log-Domain Intensity--Chromaticity Decoupling Perspective

TL;DR

This paper introduces a novel low-light image enhancement method based on log-domain intensity--chromaticity decoupling, improving image quality and downstream face detection performance.

Contribution

It proposes a new decoupling-based approach with explicit constraints to suppress noise and abnormal channel amplification in low-light images.

Findings

Achieves 29.71 dB PSNR and 0.89 SSIM on LOLv2-Real.

Outperforms existing methods in quantitative and visual quality.

Enhances downstream face detection under low-light conditions.

Abstract

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.