MUFusion

This is the code of the paper titled as "MUFusion: A general unsupervised image fusion network based on memory unit".

The paper can be found here.

Environment

• Python 3.7.3
• torch 1.9.1
• scipy 1.2.0 (alternatively, you can use the cv2.imread to replace the imread from scipy.misc)

To test on the pre-trained model

Put your image pairs in the "test_images" directory and run the following prompt:

python test_image.py

You may need to modify related variables in "test_image.py" and the model name in "args_fusion.py"

Tips: If you have difficulty resolving the "imread"&"imsave" from the scipy package, in all files, please replace "imread" with "cv2.imread(path,0)" & replace "imsave" with "cv2.imwrite(path, fuseImage)";

For calculating the image quality assessments, please refer to this repository.

2023-5-25: RGB inference code for different tasks are avaialble now.

To train

Put the training patches in the "XX_Patches" directory and run the following prompt:

python train.py

The training informaiton (number of samples, batch size etc.) can be changed in the "args_fusion.py"

The raw training patches will be available here.

To create your own dataset and compare algorithms in the same environment, please refer to this code for generating the patches.

Links to the original datasets: TNO (password：18sb), RoadScene, Milti-exposure, Multi-focus (password：4js3), Medical.

Contact Informaiton

If you have any questions, please contact me at chunyang_cheng@163.com.

Acknowledgement

Code of this implementation is based on the DenseFuse.

Citation

If this work is helpful to you, please cite it as (BibTeX):

@article{cheng2023mufusion,
  title={MUFusion: A general unsupervised image fusion network based on memory unit},
  author={Cheng, Chunyang and Xu, Tianyang and Wu, Xiao-Jun},
  journal={Information Fusion},
  volume={92},
  pages={80--92},
  year={2023},
  publisher={Elsevier}
}