Unveiling Advanced Frequency Disentanglement Paradigm for Low-Light Image Enhancement

TL;DR

This paper introduces an advanced frequency disentanglement paradigm for low-light image enhancement that improves existing methods' performance with minimal additional computational cost, leveraging frequency decomposition and versatile integration.

Contribution

It reveals that a simple disentanglement paradigm suffices to enhance state-of-the-art LLIE models, using Laplace-based frequency consistency for better optimization and broad model compatibility.

Findings

Achieves up to 7.68dB PSNR improvement on benchmarks.

Enhances multiple models including CNNs, Transformers, and diffusion models.

Adds only 88K parameters to existing architectures.

Abstract

Previous low-light image enhancement (LLIE) approaches, while employing frequency decomposition techniques to address the intertwined challenges of low frequency (e.g., illumination recovery) and high frequency (e.g., noise reduction), primarily focused on the development of dedicated and complex networks to achieve improved performance. In contrast, we reveal that an advanced disentanglement paradigm is sufficient to consistently enhance state-of-the-art methods with minimal computational overhead. Leveraging the image Laplace decomposition scheme, we propose a novel low-frequency consistency method, facilitating improved frequency disentanglement optimization. Our method, seamlessly integrating with various models such as CNNs, Transformers, and flow-based and diffusion models, demonstrates remarkable adaptability. Noteworthy improvements are showcased across five popular benchmarks, with up to 7.68dB gains on PSNR achieved for six state-of-the-art models. Impressively, our approach maintains efficiency with only 88K extra parameters, setting a new standard in the challenging realm of low-light image enhancement.