Chinese Optics, Volume. 18, Issue 2, 317(2025)
Infrared and visible image fusion guided by cross-domain interactive attention and contrastive learning
Figures & Tables(14)
Fig. 2. Detail-enhanced network architecture with dual-branch skip connections
Fig. 3. Image fusion network architecture with dual-branch joint encoder
Fig. 4. Network framework for contrastive learning of attributes and content
Fig. 6. Fusion results for seven groups of scenes processed by different algorithms in the TNO
Table 1. Mean values of objective evaluation indices for 42 groups of images in the TNO
View tableView in ArticleTable 1. Mean values of objective evaluation indices for 42 groups of images in the TNO
方法 评价指标 AG EN SD VIF SF MI PSNR SSIM Time RFN-Nest 2.669 6.963 36.897 0.559 5.874 2.113 62.193 0.649 0.249 U2Fusion ${\underline{\underline{5.023}}} $ 6.997 37.697 0.619 11.864 2.005 62.808 0.605 0.354 PIAFusion 3.828 6.814 37.141 0.740 9.620 3.352 61.776 0.468 0.682 SuperFusion 2.421 6.558 30.663 0.422 6.275 2.330 60.979 ${\underline{\underline{0.753}}} $ 0.715 SwinFusion 3.560 6.819 34.825 0.658 8.985 2.297 62.577 0.686 ${\underline{\underline{0.553}}} $ SeAFusion 4.980 ${\underline{\underline{7.133}}} $ 44.244 ${\underline{\underline{0.704}}} $ 12.253 ${\underline{\underline{2.833}}} $ 61.392 0.628 0.604 TarDAL 2.998 6.840 ${\underline{\underline{45.212}}} $ 0.539 7.959 2.802 62.304 0.597 3.159 DIVFusion 5.560 7.593 47.526 0.625 13.463 2.217 59.979 0.408 2.149 DDFM 5.111 6.854 37.081 0.629 ${\underline{\underline{12.952}}} $ 2.048 63.466 0.618 3.517 LRRNet 3.600 6.838 39.499 0.551 9.331 2.515 62.656 0.546 0.927 SFCFusion 4.324 6.700 31.297 0.675 11.401 1.997 63.133 0.687 2.578 MTDFusion 4.612 6.695 33.669 0.578 11.643 2.256 62.150 0.758 1.597 本文方法 5.601 7.443 50.879 0.792 14.742 3.374 ${\underline{\underline{62.875}}} $ 0.831 0.593
Table 2. Mean value of objective evaluation indices for 40 groups of images in MSRS
View tableView in ArticleTable 2. Mean value of objective evaluation indices for 40 groups of images in MSRS
方法 评价指标 AG EN SD VIF SF MI PSNR SSIM Time RFN-Nest 1.557 5.209 25.976 0.555 4.725 2.498 67.123 0.565 0.428 U2Fusion 2.409 5.332 25.303 0.555 7.709 2.244 66.599 0.595 0.536 PIAFusion 3.598 6.536 $\underline{\underline{46.263}} $ 1.008 10.945 3.825 64.464 0.545 0.892 SuperFusion 3.598 6.468 43.469 0.913 9.464 $\underline{\underline{3.999}} $ 64.851 0.545 0.874 SwinFusion 3.598 6.491 44.209 0.913 9.712 4.173 64.821 0.545 0.724 SeAFusion 3.598 6.547 42.902 $\underline{\underline{0.952}} $ 10.047 3.776 64.570 0.581 $\underline{\underline{0.647}} $ TarDAL 3.598 3.312 26.792 0.162 13.973 1.245 63.544 0.278 4.589 DIVFusion 4.313 7.406 54.228 0.784 $\underline{\underline{11.575}} $ 2.545 56.314 0.243 3.248 DDFM 1.848 5.642 21.144 0.561 5.922 2.414 67.088 0.705 3.774 LRRNet 3.508 $\underline{\underline{6.780}} $ 25.976 0.852 10.058 3.202 58.759 $\underline{\underline{0.685}} $ 1.938 SFCFusion $\underline{\underline{3.759}} $ 5.933 30.836 0.636 11.407 2.002 66.788 0.526 2.549 MTDFusion 2.114 5.586 30.836 0.399 6.727 2.090 65.110 0.632 2.874 本文方法 4.731 7.331 56.891 1.138 13.216 4.215 $\underline{\underline{66.821}} $ 0.732 0.698
Table 3. Mean value of objective evaluation indices for 221 groups of images in RoadSence
View tableView in ArticleTable 3. Mean value of objective evaluation indices for 221 groups of images in RoadSence
方法 评价指标 AG EN SD VIF SF MI PSNR SSIM Time RFN-Nest 3.362 7.336 46.025 0.500 7.852 2.738 61.366 0.617 0.357 U2Fusion 6.099 7.183 40.092 0.564 15.282 2.578 61.366 0.696 0.684 PIAFusion 4.308 6.981 42.702 0.681 12.132 3.557 61.680 0.659 0.534 SuperFusion 4.469 6.990 41.358 0.608 12.185 $\underline{\underline{3.562}} $ 62.107 0.566 0.824 SwinFusion 4.516 7.000 44.067 0.614 16.720 3.334 61.297 0.529 $\underline{\underline{0.545}} $ SeAFusion $\underline{\underline{6.491}} $ 7.330 49.645 0.600 $\underline{\underline{16.625}} $ 3.022 61.714 0.584 0.657 TarDAL 6.691 7.550 59.398 0.418 16.123 2.191 59.566 0.552 3.924 DIVFusion 5.010 ${\underline{\underline{7.539}}} $ 54.188 0.572 13.295 2.900 61.779 0.441 2.842 DDFM 3.952 6.868 33.551 0.532 10.174 2.845 64.484 0.660 3.667 LRRNet 5.692 7.526 ${\underline{\underline{54.772}}} $ ${\underline{\underline{0.631}}} $ 15.223 3.510 62.025 0.730 1.259 SFCFusion 6.304 7.222 41.496 0.591 15.994 2.842 63.781 0.670 1.842 MTDFusion 4.407 7.059 37.356 0.577 11.417 2.896 ${\underline{\underline{64.440}}} $ ${\underline{\underline{0.728}}} $ 2.067 本文方法 6.924 7.596 60.891 0.703 16.810 4.027 62.303 0.804 0.573
Table 4. Mean values of objective evaluation indices in 10 groups of ablation experiment scenes
View tableView in ArticleTable 4. Mean values of objective evaluation indices in 10 groups of ablation experiment scenes
模型 AG EN SD VIF SF MI PSNR SSIM 实验1 7.328 7.248 52.314 0.652 16.257 3.628 54.217 0.766 实验2 5.628 6.995 48.302 0.563 18.005 3.495 57.457 0.501 实验3 6.715 7.137 45.541 0.589 18.186 3.214 56.259 0.627 实验4 6.357 7.033 50.249 0.637 17.894 3.455 57.224 0.643 实验5 7.780 7.324 52.413 0.633 17.924 3.527 58.149 0.702 本文方法 8.326 7.421 61.672 0.709 18.259 3.989 59.248 0.791
Tools
Get Citation
Copy Citation Text
Jing DI, Chan LIANG, Ji-zhao LIU, Jing LIAN. Infrared and visible image fusion guided by cross-domain interactive attention and contrastive learning[J]. Chinese Optics, 2025, 18(2): 317
Paper Information
Category:
Received: Aug. 14, 2024
Accepted: --
Published Online: May. 19, 2025
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20