thfylsty / Objective-evaluation-for-image-fusion

image fusion : EI CrossEntropy SF EN Qabf SCD FMI_w FMI_dct SSIM MS-SSIM FMI-pixel Nabf MI VIF SD EN DF QSF QMI QS QY QC QNCIE Qabf AG MIabf QG CC VIFF QP QW QE QCV QCB

如果可以帮到你，感谢简单引用一下我们的文章。

If you got some help, thank you for simply quoting our article.

Methods

1 PerceptualFusion

Fu, Yu, Xiao-Jun Wu, and Tariq Durrani. "Image fusion based on generative adversarial network consistent with perception." Information Fusion (2021).

@article{fu2021image,
  title={Image fusion based on generative adversarial network consistent with perception},
  author={Fu, Yu and Wu, Xiao-Jun and Durrani, Tariq},
  journal={Information Fusion},
  year={2021},
  publisher={Elsevier}
}

2 Dualbranch_Fusion

@inproceedings{fu2021dual,
  title={A Dual-branch Network for Infrared and Visible Image Fusion},
  author={Fu, Yu and Wu, Xiao-Jun},
  booktitle={2020 25th International Conference on Pattern Recognition},
  year={2020}
}

3 DeepDecFusion

@article{fu2021deep,
  title={A Deep Decomposition Network for Image Processing: A Case Study for Visible and Infrared Image Fusion},
  author={Fu, Yu and Wu, Xiao-Jun and Kittler, Josef},
  journal={arXiv preprint arXiv:2102.10526},
  year={2021}
}

4 An Effective Method for Fusing Infrared and Visible Images.

(later)

5 PPT Fusion

(later)

run

"main.m"

代码没写注释，有点乱

1.多进程

默认开启了多进程，核心为5个，

如果不需要，注释第二行和最后一行，以及将parfor改成for

line1: p = parpool('local',5) ;
......
line13: parfor ->>>  for
......
line62: delete(gcp('nocreate'));

2.多个文件夹评价

设置这里可以进行多个文件夹的评价

line5: fusion_name = ["0","5","10","15","20"];

比如，多种方法可以写成

line5: fusion_name = ["CVT","DCWT","Deepfuse"..........];

比如，多次迭代的结果可以写成

line5: fusion_name = ["0","5","10","15","20"];

然后由第8,9行来控制评价的文件夹

for fm=1:5
    name = fusion_name(fm);

输入图像

%         source_ir  = ['../TNO/ir/',num2str(i),'.bmp'];
%         source_vis = ['../TNO/vi/',num2str(i),'.bmp'];
source_ir  = ['../road/ir/',num2str(i),'.jpg'];
source_vis = ['../road/vi/',num2str(i),'.jpg'];
        
fused = strcat(fused_path,name,'/road/sm/',num2str(i),'.bmp');

输出结果

在所在文件夹保存成.mat文件

    save_path = strcat(fused_path,'/',name);
    save(save_path,'a','b') ;
    
save_path = strcat(fused_path,'/all');
save(save_path,'c') ;

然后用excel自己处理吧。

这是表头

EI	CrossEntropy	SF	EN	Qabf	SCD	FMI_w	FMI_dct	SSIM	MS_SSIM	FMI_pixel	Nabf	MI	VIF	SD	EN	DF	QSF	QMI	QS	QY	QC	QNCIE	Q^{AB/F}	AG	MIabf	QG	CC	VIFF	QP	QW	QE	QCV	QCB

=======================

=======================

=======================

run

"main.m"

The code is not commented, it's a bit messy

1. Multi-process

Multi-process is enabled by default, with 5 cores,

If not needed, comment the second and last lines, and change parfor to for

line1: p = parpool('local',5);
......
line13: parfor ->>> for
......
line62: delete(gcp('nocreate'));

2. Multiple folder evaluation

Set here to evaluate multiple folders

line5: fusion_name = ["0","5","10","15","20"];

For example, multiple methods can be written as

line5: fusion_name = ["CVT","DCWT","Deepfuse"..........];

For example, the result of multiple iterations can be written as

line5: fusion_name = ["0","5","10","15","20"];

Then lines 8 and 9 control the evaluation folder

for fm=1:5
    name = fusion_name(fm);

Input image

% source_ir = ['../TNO/ir/',num2str(i),'.bmp'];
% source_vis = ['../TNO/vi/',num2str(i),'.bmp'];
source_ir = ['../road/ir/',num2str(i),'.jpg'];
source_vis = ['../road/vi/',num2str(i),'.jpg'];
        
fused = strcat(fused_path,name,'/road/sm/',num2str(i),'.bmp');

Output result

Save as a .mat file in the folder where it is located

    save_path = strcat(fused_path,'/',name);
    save(save_path,'a','b');
    
save_path = strcat(fused_path,'/all');
save(save_path,'c');

Then use excel to handle it yourself.

This is the header

EI	CrossEntropy	SF	EN	Qabf	SCD	FMI_w	FMI_dct	SSIM	MS_SSIM	FMI_pixel	Nabf	MI	VIF	SD	EN	DF	QSF	QMI	QS	QY	QC	QNCIE	Q^{ AB/F}	AG	MIabf	QG	CC	VIFF	QP	QW	QE	QCV	QCB