M2Retinexformer

This repository contains the official weights for M2Retinexformer (Multi-Modal Retinexformer), introduced in the paper M2Retinexformer: Multi-Modal Retinexformer for Low-Light Image Enhancement.

• Paper: M2Retinexformer: Multi-Modal Retinexformer for Low-Light Image Enhancement
• Code: GitHub Repository

Introduction

Low-light image enhancement is challenging due to complex degradations, including amplified noise, artifacts, and color distortion. M2Retinexformer is a novel framework that extends Retinexformer by incorporating depth cues, luminance priors, and semantic features within a progressive refinement pipeline.

Depth provides geometric context invariant to lighting variations, while luminance and semantic features offer explicit guidance on brightness distribution and scene understanding. These modalities are fused through cross-attention with adaptive gating to dynamically balance illumination-guided self-attention and cross-attention based on the reliability of auxiliary cues.

Citation

If you find this work useful, please cite:

@misc{aboelwafa2026m2retinexformermultimodalretinexformerlowlight,
      title={M2Retinexformer: Multi-Modal Retinexformer for Low-Light Image Enhancement}, 
      author={Youssef Aboelwafa and Hicham G. Elmongui and Marwan Torki},
      year={2026},
      eprint={2605.12556},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2605.12556}, 
}

Acknowledgements

This project is built on the baseline architecture of Retinexformer.