Beyond Illumination: A Conditional Mutual Information-Guided Network for Low-Light Image Enhancement

2026-08-03

AI Summary

This work addresses the limitation of existing low-light image enhancement methods, which often neglect the interplay between luminance and chrominance components, thereby constraining their representational capacity. To overcome this, the authors propose CMIG-Net, the first framework to incorporate conditional mutual information into this task. Built upon a decoupled representation in the HVI color space, CMIG-Net introduces a Conditional Mutual Information Calibration (CMIC) module that quantifies the contribution of chrominance features conditioned on luminance, generating a conditional information map for adaptive chrominance calibration based on local illumination statistics. Additionally, a Dual-branch Dynamic Interaction Refinement (D2IR) module is devised to dynamically modulate information exchange between the luminance and chrominance branches. Experiments demonstrate that the proposed method outperforms CIDNet across multiple paired low-light datasets, achieving PSNR gains of up to 0.619 dB, including a 0.382 dB improvement on the Sony-Total-Dark dataset.

Abstract

Low-light image enhancement (LLIE) seeks to restore structural fidelity, natural color rendition, and proper exposure from images captured under inadequate lighting conditions. Recent state-of-the-art approaches, such as CIDNet, adopt a dual-branch architecture comprising a chrominance (HV) branch and an intensity (I) branch to separately model decoupled chromatic and luminance information within the HVI color space. However, these methods overlook the mutual interaction between intensity and chrominance components, which inherently limits their representational capacity and leads to suboptimal enhancement performance. To address this limitation, we propose the Conditional Mutual Information-Guided Network (CMIG-Net), which leverages conditional mutual information as a principled metric to quantitatively assess the contribution of chrominance features conditioned on the available intensity information. In particular, we design a Conditional Mutual Information Calibration (CMIC) module that generates a conditional information map, enabling region-adaptive recalibration of chrominance representations according to local illumination statistics. Furthermore, we introduce a Dynamic Dual-branch Information Restoration (D2IR) module, which adaptively governs bidirectional information flow between the intensity and chrominance branches, guided by both the conditional prior and the instantaneous restoration state. Extensive experiments on paired LLIE benchmarks demonstrate that CMIG-Net consistently outperforms CIDNet, achieving up to a 0.619 dB gain in PSNR, with a 0.382 dB improvement specifically on the challenging Sony-Total-Dark dataset.

Problem

• Low-light image enhancement
• chrominance-intensity interaction
• conditional mutual information
• HVI color space
• image restoration

Innovation

• Conditional Mutual Information
• Low-Light Image Enhancement
• Chrominance-Intensity Interaction
• Adaptive Feature Recalibration
• Dual-Branch Architecture