ICLR: Inter-Chrominance and Luminance Interaction for Natural Color Restoration in Low-Light Image Enhancement

TL;DR

This work targets low-light image enhancement by decoupling luminance and chrominance using the HVI color space and addressing two key challenges: distributional differences between luminance and chrominance and weak correlations within chrominance channels in homogeneous regions. It introduces ICLR, a three-level U-Net framework equipped with a Dual-stream Interaction Enhancement Module (DIEM) and a Covariance Correction Loss (CCL). DIEM comprises Multidimensional Attention-guided Fusion Modules (MAFM) and Cross Dynamic Enhancement Modules (CDEM) to fuse and enhance cross-branch information, while CCL imposes luminance-guided and covariance-based constraints to stabilize optimization and reduce gradient conflicts between chrominance channels. Experiments on LOLv1/LOLv2 and unpaired datasets demonstrate state-of-the-art restoration quality, with notable color fidelity improvements and favorable computational efficiency, indicating strong generalization to diverse lighting conditions and color distributions. The work advances LLIE by integrating distribution-aware cross-branch learning with covariance-based optimization in HVI space, offering practical benefits for robust color restoration in real-world low-light imaging.

Abstract

Low-Light Image Enhancement (LLIE) task aims at improving contrast while restoring details and textures for images captured in low-light conditions. HVI color space has made significant progress in this task by enabling precise decoupling of chrominance and luminance. However, for the interaction of chrominance and luminance branches, substantial distributional differences between the two branches prevalent in natural images limit complementary feature extraction, and luminance errors are propagated to chrominance channels through the nonlinear parameter. Furthermore, for interaction between different chrominance branches, images with large homogeneous-color regions usually exhibit weak correlation between chrominance branches due to concentrated distributions. Traditional pixel-wise losses exploit strong inter-branch correlations for co-optimization, causing gradient conflicts in weakly correlated regions. Therefore, we propose an Inter-Chrominance and Luminance Interaction (ICLR) framework including a Dual-stream Interaction Enhancement Module (DIEM) and a Covariance Correction Loss (CCL). The DIEM improves the extraction of complementary information from two dimensions, fusion and enhancement, respectively. The CCL utilizes luminance residual statistics to penalize chrominance errors and balances gradient conflicts by constraining chrominance branches covariance. Experimental results on multiple datasets show that the proposed ICLR framework outperforms state-of-the-art methods.

Figures (8)

• Figure 1: Our two key observations. (a) Substantial distributional differences between luminance and chrominance limit the complementary feature extraction. (b) In images with large homogeneous-color regions, chrominance branches typically exhibit weak correlation due to their relatively concentrated distributions, which leads to gradient conflicts.
• Figure 2: The overall architecture of the proposed ICLR framework. ICLR is a three-level U-Net architecture integrated with Dual-stream Interaction Enhancement Modules (DIEM) and a Covariance Correction Loss (CCL). The DIEM module mainly consists of Multidimensional Attention-guided Fusion Modules (MAFM) and Cross Dynamic Enhancement Modules (CDEM).
• Figure 3: Comparison of luminance and chrominance branches feature distributions before and after MAFM.
• Figure 4: The qualitative comparison of the methods we compared in terms of color restoration. Zoom in for the best view.
• Figure 5: (a) Comparison of NIQE↓ on five unpaired datasets, where lower values indicate better performance. (b) Comparison of Flops and Inference Time.