Optics and Precision Engineering, Volume. 33, Issue 1, 148(2025)
Conditional diffusion and multi-channel high-low frequency parallel fusion of infrared and visible light images
Figures & Tables(13)
Fig. 1. General network framework based on conditional diffusion and multi-channel high-low frequency parallel fusion
Fig. 9. Fusion results of day and night scenes in ablation experiment
Table 1. Objective evaluation indicators for nighttime scenes in MSRS dataset '00004N'
View tableView in ArticleTable 1. Objective evaluation indicators for nighttime scenes in MSRS dataset '00004N'
Method EN MI AG SF SSIM PSNR CC VIFF RFN-Nest 5.701 2 1.915 9 3.588 6 7.301 9 0.722 2 22.625 7 0.902 6 0.567 7 SDNet 5.657 6 1.663 2 6.462 2 12.254 3 0.705 1 22.102 2 0.895 3 0.621 8 U2Fusion 5.255 8 1.135 9 6.208 1 14.113 4 0.727 9 21.261 6 0.782 9 0.354 8 CDDFuse 5.000 4 2.123 7 5.082 2 12.464 1 0.796 9 22.076 9 0.600 6 0.631 7 TarDAL 5.202 4 2.451 2 4.769 0 7.825 5 0.629 2 20.986 6 0.797 9 0.234 0 DIVFusion 7.128 1 2.057 4 12.004 1 16.413 4 0.374 5 7.288 7 0.608 5 0.517 7 SwinFusion 5.855 1 1.962 2 6.721 2 10.190 6 0.634 8 22.573 4 0.894 3 0.600 5 DDFM 5.343 5 2.052 3 4.133 2 6.457 9 0.806 7 22.934 9 0.821 5 0.330 8 BTSFusion 5.989 6 2.088 7 7.850 2 14.650 8 0.826 6 21.951 9 0.904 8 0.656 5 Ours 6.112 9 2.471 5 10.129 4 14.706 8 0.837 5 22.967 1 0.908 1 0.884 9
Table 2. Objective evaluation indicators for daytime scenes in MSRS dataset '00241D '
View tableView in ArticleTable 2. Objective evaluation indicators for daytime scenes in MSRS dataset '00241D '
Method EN MI AG SF SSIM PSNR CC VIFF RFN-Nest 7.077 5 2.052 3 14.292 5 17.701 4 0.591 7 14.636 8 0.915 3 0.527 7 SDNet 6.582 4. 2.127 5 23.135 9 34.406 3 0.303 6 13.124 2 0.916 2 0.750 3 U2Fusion 6.928 9 1.754 4 25.950 2 41.272 2 0.608 7 12.271 1 0.829 6 0.410 1 CDDFuse 7.464 2 2.525 3 21.759 2 28.965 8 0.424 0 15.845 8 0.498 6 0.808 5 TarDAL 6.571 1 0.881 7 9.837 5 12.525 3 0.546 6 14.078 8 0.128 0.081 6 DIVFusion 7.661 6 1.857 9 21.950 9 27.648 5 0.441 2 12.978 1 0.869 3 0.817 0 SwinFusion 7.052 1 1.658 5 20.103 6 25.243 2 0.364 2 14.444 2 0.853 9 0.489 1 DDFM 6.220 7 2.439 6 9.774 4 13.658 7 0.645 5 15.031 1 0.879 1 0.260 0 BTSFusion 7.332 4 1.915 7 28.704 1 37.816 8 0.853 9 14.854 1 0.900 3 0.737 5 Ours 7.708 1 3.291 0 27.736 1 34.514 3 0.924 1 15.871 7 0.968 3 0.985 6
Table 3. Objective evaluation indicators for 'FLIR_0579' in RoadScene dataset
View tableView in ArticleTable 3. Objective evaluation indicators for 'FLIR_0579' in RoadScene dataset
Method EN MI AG SF SSIM PSNR CC VIFF RFN-Nest 7.534 2 1.616 5 10.135 13.084 8 0.712 3 14.425 0 0.710 1 0.666 7 SDNet 7.335 3 1.373 6 14.347 2 17.548 6 0.716 2 16.725 0.667 8 0.568 6 U2Fusion 7.437 4 0.529 5 18.508 3 23.046 1 0.723 0 12.752 6 0.672 2 0.595 2 CDDFuse 7.516 3 1.927 4 14.561 7 20.789 1 0.677 5 13.139 2 0.265 5 0.605 2 TarDAL 7.570 1 1.813 3 10.691 4 13.364 6 0.666 7 16.513 9 0.623 8 0.558 5 DIVFuison 7.458 7 1.688 5 17.733 5 23.027 4 0.678 4 15.026 7 0.664 0 0.662 4 SwinFusion 7.393 7 1.490 5 13.204 2 17.351 1 0.710 1 17.351 1 0.388 3 0.556 6 DDFM 7.357 5 0.723 7 12.292 4 15.468 9 0.435 7 14.483 9 0.687 5 0.067 8 BTSFusion 6.607 4 1.703 4 18.423 9 28.190 2 0.700 8 15.361 6 0.614 5 0.671 0 Ours 7.574 0 2.394 9 17.750 2 21.508 2 0.754 8 15.041 4 0.730 0 0.674 7
Table 4. Mean objective evaluation indicators of 30 images with four different network structures in ablation experiment
View tableView in ArticleTable 4. Mean objective evaluation indicators of 30 images with four different network structures in ablation experiment
Network En MI AG SF SSIM PSNR CC VIFF M_LRM+SdeNet 6.505 8 1.349 3 19.588 6 25.111 3 0.642 3 13.742 7 0.800 4 0.642 5 M_LRM+FuNet 6.301 3 1.668 14.567 1 19.573 3 0.692 5 14.658 3 0.784 7 0.449 7 FuNet+SdeNet 6.343 1 1.004 6 20.033 5 23.544 7 0.598 7 12.705 8 0.692 0 0.370 7 ALL 6.776 2 2.159 5 20.473 2 27.175 7 0.866 5 17.881 7 0.813 4 0.954 2
Tools
Get Citation
Copy Citation Text
Jing DI, Heran WANG, Chan LIANG, Jizhao LIU, Jing LIAN. Conditional diffusion and multi-channel high-low frequency parallel fusion of infrared and visible light images[J]. Optics and Precision Engineering, 2025, 33(1): 148
Paper Information
Category:
Received: Aug. 26, 2024
Accepted: --
Published Online: Apr. 1, 2025
The Author Email: Heran WANG (838129431@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20