AGLLDiff: Guiding Diffusion Models Towards Unsupervised Training-free Real-world Low-light Image Enhancement

TL;DR

AGLLDiff is a training-free diffusion-based framework that effectively enhances real-world low-light images by guiding the process with attribute modeling, overcoming the need for paired data and complex degradation modeling.

Contribution

It introduces a novel attribute guidance approach for diffusion models, enabling unsupervised, training-free low-light image enhancement in real-world scenarios.

Findings

Outperforms existing unsupervised methods on multiple benchmarks.

Handles complex wild degradations effectively.

Does not require paired training data or explicit degradation modeling.

Abstract

Existing low-light image enhancement (LIE) methods have achieved noteworthy success in solving synthetic distortions, yet they often fall short in practical applications. The limitations arise from two inherent challenges in real-world LIE: 1) the collection of distorted/clean image pairs is often impractical and sometimes even unavailable, and 2) accurately modeling complex degradations presents a non-trivial problem. To overcome them, we propose the Attribute Guidance Diffusion framework (AGLLDiff), a training-free method for effective real-world LIE. Instead of specifically defining the degradation process, AGLLDiff shifts the paradigm and models the desired attributes, such as image exposure, structure and color of normal-light images. These attributes are readily available and impose no assumptions about the degradation process, which guides the diffusion sampling process to a reliable high-quality solution space. Extensive experiments demonstrate that our approach outperforms the current leading unsupervised LIE methods across benchmarks in terms of distortion-based and perceptual-based metrics, and it performs well even in sophisticated wild degradation.