Frequency-Decomposed INR for NIR-Assisted Low-Light RGB Image Denoising

2026-04-17

AI Summary

This work proposes a frequency-domain decoupled implicit neural representation to address severe noise and high-frequency structural degradation in low-light visible images. By employing multiscale wavelet transforms to explicitly separate low- and high-frequency components, a dual-branch framework is constructed: low-light RGB images guide the reconstruction of low-frequency luminance and chrominance, while high signal-to-noise ratio near-infrared (NIR) images constrain high-frequency texture synthesis. A cross-modal differential frequency supervision mechanism and an uncertainty-aware adaptive weighting loss are introduced to enable complementary fusion in the frequency domain. This approach effectively mitigates color shifts and artifacts commonly caused by rigid spatial-domain fusion, consistently outperforming state-of-the-art methods across arbitrary resolutions while enhancing luminance consistency, structural detail recovery, and low-light perceptual reliability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Addressing the issues of severe noise and high frequency structural degradation in visible images under low-light conditions, this paper proposes a Near Infrared (NIR) aided low light image restoration method based on Frequency Decoupled Implicit Neural Representation (FDINR). Based on the statistical prior of RGB-NIR cross-modal frequency correlations, specifically that low-frequency RGB signals are more reliable, whereas high frequency NIR signals exhibit higher correlation, we explicitly decompose images into distinct frequency components via multi-scale wavelet transforms and construct a dual-branch implicit neural representation framework. Within this framework, we design a cross modal differentiated frequency supervision mechanism, leveraging low light RGB to guide the reconstruction of low frequency luminance and color, and utilizing high-SNR NIR signals to constrain the generation of high frequency texture details, thereby achieving complementary advantages in the frequency domain. Furthermore, an uncertainty-based adaptive weighting loss function is introduced to automatically balance the contributions of different frequency tasks, solving the problems of color distortion and artifacts caused by rigid fusion in the spatial domain common in traditional methods. Experimental results demonstrate that FD-INR not only effectively restores image luminance consistency and structural details but also, benefitting from its implicit continuous representation, outperforms existing methods in arbitrary-resolution reconstruction tasks, significantly enhancing the reliability of low light perception.

Problem

• low-light image denoising
• noise
• structural degradation
• NIR-assisted restoration
• frequency decomposition

Innovation

• Frequency-Decomposed INR
• NIR-assisted denoising
• cross-modal frequency correlation
• implicit neural representation
• adaptive weighting loss