FusionNet: Multi-model Linear Fusion Framework for Low-light Image Enhancement

2025-04-27

AI Summary

To address parameter explosion, optimization instability, and cross-color-space/architecture feature misalignment in multi-model low-light image enhancement, this paper proposes a parallel multi-model linear fusion framework. The framework concurrently extracts global and local features across multiple color spaces—including sRGB, HSV, and HVI—and introduces a novel Hilbert-space-theoretic constrained linear fusion mechanism, mathematically guaranteeing fusion stability and preventing network collapse while substantially reducing training overhead. It is the first method to enable efficient synergistic fusion of CNN and Transformer architectures with heterogeneous color-space representations. Our approach won first place in the CVPR 2025 NTIRE Low-Light Enhancement Challenge and consistently outperforms state-of-the-art methods on both synthetic and real-world datasets, achieving significant gains in PSNR and SSIM as well as superior visual quality.

Technology Category

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

Application Category

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

Abstract

The advent of Deep Neural Networks (DNNs) has driven remarkable progress in low-light image enhancement (LLIE), with diverse architectures (e.g., CNNs and Transformers) and color spaces (e.g., sRGB, HSV, HVI) yielding impressive results. Recent efforts have sought to leverage the complementary strengths of these paradigms, offering promising solutions to enhance performance across varying degradation scenarios. However, existing fusion strategies are hindered by challenges such as parameter explosion, optimization instability, and feature misalignment, limiting further improvements. To overcome these issues, we introduce FusionNet, a novel multi-model linear fusion framework that operates in parallel to effectively capture global and local features across diverse color spaces. By incorporating a linear fusion strategy underpinned by Hilbert space theoretical guarantees, FusionNet mitigates network collapse and reduces excessive training costs. Our method achieved 1st place in the CVPR2025 NTIRE Low Light Enhancement Challenge. Extensive experiments conducted on synthetic and real-world benchmark datasets demonstrate that the proposed method significantly outperforms state-of-the-art methods in terms of both quantitative and qualitative results, delivering robust enhancement under diverse low-light conditions.

Problem

• Overcoming parameter explosion and feature misalignment in multi-model fusion
• Enhancing low-light images across diverse color spaces effectively
• Reducing training costs while improving enhancement performance robustly

Innovation

• Multi-model linear fusion for global and local features
• Hilbert space theory ensures stability and efficiency
• Parallel processing across diverse color spaces