Cross-Fusion Transformer-Based Infrared and Visible Image Fusion Method

Fig. 4. Fused results by different algorithms on the TNO dataset. (a) Infrared image; (b) visible image; (c) DenseFuse; (d) FusionGAN; (e) GANMcC; (f) U2Fusion; (g) SwinFusion; (h) YDTR; (i) ResCCFusion; (j) DATFuse; (k) TGFuse; (l) proposed algorithm

Download full size

Fig. 5. Fused results by different algorithms on the INO dataset. (a) Infrared image; (b) visible image; (c) DenseFuse; (d) FusionGAN; (e) GANMcC; (f) U2Fusion; (g) SwinFusion; (h) YDTR; (i) ResCCFusion; (j) DATFuse; (k) TGFuse; (l) proposed algorithm

Download full size

Fig. 6. Fused results by different algorithms on the RoadScene dataset. (a) Infrared image; (b) visible image; (c) DenseFuse; (d) FusionGAN; (e) GANMcC; (f) U2Fusion; (g) SwinFusion; (h) YDTR; (i) ResCCFusion; (j) DATFuse; (k) TGFuse; (l) proposed algorithm

Download full size

Fig. 7. Fused results by different algorithms on the MSRS dataset. (a) Infrared image; (b) visible image; (c) DenseFuse; (d) FusionGAN; (e) GANMcC; (f) U2Fusion; (g) SwinFusion; (h) YDTR; (i) ResCCFusion; (j) DATFuse; (k) TGFuse; (l) proposed algorithm

Download full size

Table 1. Average indicators values of different algorithms on the TNO dataset

View table

Table 1. Average indicators values of different algorithms on the TNO dataset

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
DenseFuse	0.4048	0.8060	0.5643	0.7165	0.3790	2.6913	0.6321	0.5930
FusionGAN	0.3579	0.8054	0.4809	0.5186	0.2312	2.3367	0.4248	0.3553
GANMcC	0.4081	0.8061	0.5231	0.6307	0.3066	2.7359	0.5477	0.4920
U2Fusion	0.3449	0.8050	0.5567	0.7535	0.4308	2.2903	0.6070	0.5829
SwinFusion	0.5040	0.8095	0.7427	0.8503	0.5019	3.4595	0.7797	0.5658
YDTR	0.4453	0.8072	0.5730	0.7721	0.4188	2.9585	0.6611	0.5512
ResCCFusion	0.5324	0.8090	0.6896	0.8305	0.4767	3.5471	0.7313	0.5620
DATFuse	0.5639	0.8098	0.7464	0.9114	0.5208	3.7199	0.7871	0.5131
TGFuse	0.4908	0.8102	0.9513	0.8956	0.5533	3.3978	0.8362	0.5500
Proposed	0.7175	0.8155	0.9636	0.9339	0.5546	4.8048	0.9194	0.5215

Table 2. Average indicators values of different algorithms on the INO dataset

View table

Table 2. Average indicators values of different algorithms on the INO dataset

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
DenseFuse	0.4577	0.8079	0.3790	0.7090	0.3957	3.2418	0.5455	0.4914
FusionGAN	0.3706	0.8062	0.3446	0.5559	0.2777	2.6072	0.3317	0.3356
GANMcC	0.4130	0.8071	0.3614	0.6007	0.3260	2.9724	0.4698	0.3913
U2Fusion	0.3779	0.8063	0.4009	0.7581	0.4801	2.6853	0.4915	0.5045
SwinFusion	0.5273	0.8116	0.6535	0.8665	0.5841	3.9012	0.6709	0.4854
YDTR	0.4920	0.8094	0.4626	0.7759	0.5006	3.5091	0.5949	0.4831
ResCCFusion	0.5978	0.8126	0.4817	0.8177	0.5438	4.3048	0.6434	0.4966
DATFuse	0.5767	0.8119	0.5021	0.8666	0.5534	4.1377	0.5869	0.4699
TGFuse	0.5068	0.8112	0.9467	0.8945	0.6344	3.7760	0.7479	0.4940
Proposed	0.7020	0.8170	0.7393	0.9230	0.6065	5.1248	0.7337	0.5160

Table 3. Average indicators values of different algorithms on the RoadScene dataset

View table

Table 3. Average indicators values of different algorithms on the RoadScene dataset

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
DenseFuse	0.3787	0.8067	0.3939	0.6441	0.3475	2.6424	0.4781	0.4426
FusionGAN	0.3885	0.8072	0.3754	0.5102	0.2654	2.8168	0.3692	0.2973
GANMcC	0.3456	0.8063	0.3827	0.5878	0.3223	2.5188	0.4855	0.3854
U2Fusion	0.3726	0.8069	0.4857	0.7631	0.4667	2.6238	0.5175	0.4882
SwinFusion	0.5152	0.8092	0.5276	0.7727	0.4562	3.5876	0.6188	0.4514
YDTR	0.4100	0.8070	0.4329	0.7082	0.4182	2.8519	0.5556	0.4718
ResCCFusion	0.5037	0.8090	0.4570	0.9963	0.4480	3.5155	0.5653	0.4106
DATFuse	0.5792	0.8104	0.5263	0.8429	0.4848	3.9343	0.5875	0.4420
TGFuse	0.4027	0.8070	0.6635	0.7360	0.5118	2.8418	0.6023	0.4428
Proposed	0.7100	0.8149	0.7044	0.8654	0.5093	4.9370	0.7615	0.4771

Table 4. Average indicators values of different algorithms on the MSRS dataset

View table

Table 4. Average indicators values of different algorithms on the MSRS dataset

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
DenseFuse	0.4308	0.8065	0.4230	0.6587	0.3646	2.7288	0.6583	0.4467
FusionGAN	0.3397	0.8040	0.3450	0.4060	0.1230	1.9372	0.4030	0.2531
GANMcC	0.4120	0.8059	0.3827	0.5796	0.2537	2.6054	0.5859	0.3877
U2Fusion	0.3584	0.8046	0.4161	0.5639	0.3715	2.1248	0.4906	0.3795
SwinFusion	0.6774	0.8178	0.9221	0.9121	0.6575	4.5662	0.9168	0.4661
YDTR	0.4597	0.8079	0.4688	0.6313	0.3955	2.9138	0.5920	0.3797
ResCCFusion	0.6435	0.8145	0.8934	0.8944	0.6549	4.2571	0.8507	0.4649
DATFuse	0.6033	0.8137	0.6848	0.8692	0.6422	4.0097	0.8462	0.4537
TGFuse	0.5311	0.8128	1.3483	0.8872	0.6757	3.6547	0.9360	0.4790
Proposed	0.6867	0.8187	0.9110	0.9061	0.6542	4.6780	0.9185	0.4612

Table 5. Ablation experimental results of CFTFB

View table

Table 5. Ablation experimental results of CFTFB

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
w/o AWM	0.7098	0.8153	0.9290	0.9301	0.5511	4.7718	0.9107	0.5187
Swap	0.6957	0.8146	0.8919	0.9424	0.5376	4.6397	0.8952	0.5177
w/o Cross	0.6871	0.8143	0.8838	0.9425	0.5395	4.5727	0.8951	0.5198
w/o CFTFB	0.6330	0.8123	0.8265	0.9097	0.5223	4.2467	0.8500	0.5122
Proposed	0.7175	0.8155	0.9636	0.9339	0.5546	4.8048	0.9194	0.5215

Table 6. Ablation experimental results of MSFFB

View table

Table 6. Ablation experimental results of MSFFB

Algorithm	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
w/o CAM	0.6982	0.8149	0.9474	0.9313	0.5514	4.6756	0.9055	0.5162
w/o MSFFB	0.6917	0.8143	0.8879	0.9268	0.5460	4.6066	0.8712	0.5210
Proposed	0.7175	0.8155	0.9636	0.9339	0.5546	4.8048	0.9194	0.5215

Table 7. Ablation experimental results on number of ECB modules

View table

Table 7. Ablation experimental results on number of ECB modules

Number of ECB	$Q_{M I}$	$Q_{N I C E}$	$Q_{M}$	$Q_{Y}$	$Q_{A B / F}$	MI	VIF	SSIM
1	0.5516	0.8097	0.6809	0.8527	0.4706	3.6887	0.7561	0.5084
2	0.6314	0.8125	0.7399	0.9065	0.5067	4.2123	0.8341	0.5204
3	0.7175	0.8155	0.9636	0.9339	0.5546	4.8048	0.9194	0.5215
4	0.6855	0.8144	0.8546	0.9353	0.5419	4.5798	0.8983	0.5210

Tools

Get Citation

Copy Citation Text

Haitao Yin, Changsheng Zhou. Cross-Fusion Transformer-Based Infrared and Visible Image Fusion Method[J]. Laser & Optoelectronics Progress, 2025, 62(6): 0637002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Digital Image Processing

Received: Apr. 7, 2024

Accepted: Aug. 1, 2024

Published Online: Mar. 4, 2025

The Author Email:

DOI:10.3788/LOP241049

CSTR:32186.14.LOP241049

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology