Dif-Fusion

This is Pytorch implementation for “Dif-Fusion: Toward High Color Fidelity in Infrared and Visible Image Fusion with Diffusion Models”

Overview

The overall framework for Dif-Fusion

Fusion examples

Running the code

Run train_ddpm.py to train a diffusion model for infrared and visible images.

Run train_fusion_head.py to train the fusion model.

Run t_fusion.py to obtain the fused color images.

Please set paths to infrared and visible image folds in train_xxx.py its corresponding json files. If there are trained models, set paths to these files in xxxx.json.

• The trained weights can be downloaded from: https://pan.baidu.com/s/1zOz9LRQsSnM1TDF7yYvXIw Password: q17w

• The datasets can be downloaded from: https://github.com/Linfeng-Tang/Image-Fusion

• For the codebase, please refer to: https://github.com/wgcban/ddpm-cd

• For the evaluation, please refer to: https://github.com/Linfeng-Tang/Image-Fusion

BibTeX

@ARTICLE{Diffusioncolor ,
  author={Yue, Jun and Fang, Leyuan and Xia, Shaobo and Deng, Yue and Ma, Jiayi},
  journal={IEEE Transactions on Image Processing}, 
  title={Dif-Fusion: Toward High Color Fidelity in Infrared and Visible Image Fusion With Diffusion Models}, 
  year={2023},
  volume={32},
  pages={5705-5720},
  doi={10.1109/TIP.2023.3322046}}