FusionGCN (Expert Systems With Applications, 2025)：

欢迎参考和引用我们的工作(Welcome to refer to and cite our work)

文章发表在Expert Systems with Applications Volume 262, 1 March 2025上

Code for paper “FusionGCN: Multi-focus image fusion using superpixel features generation GCN and pixel-level feature reconstruction CNN”.

Delivery timeline

• Received 27 July 2024;
• Received in revised form 27 October 2024;
• Accepted 29 October 2024;
• Available online 7 November 2024

Highlights

• To our knowledge, this is the first time that GCN has been used to solve the problem of multi-focus image fusion.
• Compared with existing methods, FusionGCN provides a new solution for multi-focus image fusion tasks running on devices with limited computing resources.
• FusionGCN has achieved good results on multiple datasets.

Reference information ※

@article{Ouyang2025FusionGCN,
  title={FusionGCN: Multi-focus image fusion using superpixel features generation GCN and pixel-level feature reconstruction CNN},  
  author={Yuncan Ouyang and Hao Zhai and Hanyue Hu and Xiaohang Li and Zhi Zeng},  
  journal={Expert Systems with Applications},  
  pages={125665},  
  year={2025},  
  publisher={Elsevier}  
}

Or

Ouyang Y, Zhai H, Hu H, et al. FusionGCN: Multi-focus image fusion using superpixel features generation GCN and pixel-level feature reconstruction CNN[J]. Expert Systems with Applications, 2025: 125665.

Dependencies

• python >= 3.6
• pytorch >= 1.5.0
• CUDA >= 12.0
• train.py -- 训练我们的网络(Train our network)
• inference.py -- 利用训练好的网络参数进行图像融合(Fusion images through network)

How to use ※

• 代码提供Lytro, MFFW, MFI-WHU和GrayScale4个数据集进行测试
• The code provides four datasets for testing: Lytro, MFFW, MFI-WHU and GrayScale
• 仅需替换inference.py中的数据集名称即可一键运行
• Simply replace the dataset name in the inference.py to run with one click

Results

The output results will be stored in ./Result/(datasetname).

Our results in Lytro, MFFW, MFI-WHU, GrayScale datasets can be downloaded.

Acknowledgments ※

以上代码在 MFFT(EAAI, 2024) 的基础上构建而来
感谢他的杰出工作

The above code is generated by MFFT(EAAI, 2024) Built on the foundation
Thank you for his outstanding work

Contact information

E-mail addresses: [email protected] (Y. Ouyang)