Table 1 Quantitative comparison (MI, SF, VIF, \(Q^{AB/F}\), and SSIM) of nine methods on the M3FD, FMB, and RoadScene datasets.

From: A dual-stream feature decomposition network with weight transformation for multi-modality image fusion

Datasets

Metrics

IFCNN

U2Fusion

SwinFusion

SeAFusion

SuperFusion

TarDAL

SDDGAN

CoCoNet

Diff-IF

Ours

M3FD

MI

2.989

2.837

4.469

3.641

4.442

3.169

3.193

2.789

4.907

5.164

SF

15.302

8.39

14.404

14.598

11.767

12.77

7.871

25.47

14.764

14.626

VIF

0.66

0.56

0.767

0.706

0.67

0.596

0.534

0.733

0.769

0.844

\(Q^{AB/F}\)

0.614

0.412

0.626

0.605

0.525

0.41

0.291

0.392

0.593

0.665

SSIM

0.947

0.888

1.016

0.944

0.991

0.88

0.788

0.682

0.935

1.001

FMB

MI

3.2

3.052

4.586

3.752

4.318

3.431

3.252

2.992

4.873

5.132

SF

16.233

8.694

15.388

15.557

12.824

12.024

8.256

27.409

15.706

15.757

VIF

0.688

0.595

0.819

0.748

0.712

0.602

0.54

0.753

0.83

0.886

\(Q^{AB/F}\)

0.656

0.446

0.679

0.654

0.585

0.41

0.301

0.414

0.656

0.714

SSIM

0.963

0.907

1.015

0.958

1.001

0.909

0.798

0.69

0.974

1

RoadScene

MI

3.011

3.226

3.714

3.178

3.974

3.391

3.192

2.75

4.26

4.698

SF

10.135

6.777

8.655

13.364

8.566

10.656

5.814

16.992

10.308

8.089

VIF

0.582

0.574

0.632

0.62

0.664

0.578

0.497

0.623

0.705

0.821

\(Q^{AB/F}\)

0.502

0.399

0.441

0.455

0.445

0.427

0.297

0.371

0.494

0.485

SSIM

0.933

0.902

0.941

0.866

0.958

0.871

0.789

0.717

0.861

0.967

  1. For all five metrics, a higher value indicates better fusion quality. The best three values in each metric are denoted in bolditalic, italic, and bold, respectively.