Journal of Infrared and Millimeter Waves, Volume. 44, Issue 2, 275(2025)
BDMFuse: Multi-scale network fusion for infrared and visible images based on base and detail features
Figures & Tables(15)
Fig. 1. The overall network of BDMFuse: (a) overall structure of the training phase; (b) overall structure of the testing phase
Fig. 8. Fusion image of ablation experiments: (a) IR; (b) VIS; (c) SE-Attention; (d) ECA-Attention; (e) No-Attention; (f) No-Comp Encoder; (g) No-Fusion Module; (h) No-Strategy; (i) No-Strategy & No-Multi-scale; (j) proposed
Fig. 9. Fusion image of TNO: (a) IR; (b) VIS; (c) DenseFuse; (d) RFN-Nest; (e) FusionGAN; (f) U2Fusion; (g) CSF; (h) SEDR; (i) SwinFusion; (j) SeAFusion; (k) CDDFuse; (l) proposed
Fig. 10. Fusion image of RoadScene: (a) IR; (b) VIS; (c) DenseFuse; (d) RFN-Nest; (e) FusionGAN; (f) U2Fusion; (g) CSF; (h) SEDR; (i) SwinFusion; (j) SeAFusion; (k) CDDFuse; (l) proposed
Fig. 11. Fusion image of LLVIP: (a) IR; (b) VIS; (c) DenseFuse; (d) RFN-Nest; (e) FusionGAN; (f) U2Fusion; (g) CSF; (h) SEDR; (i) SwinFusion; (j) SeAFusion; (k) CDDFuse; (l) (Our proposed)
Table 1. Average quality evaluation metrics for ablation experiments
View tableView in ArticleTable 1. Average quality evaluation metrics for ablation experiments
EN MI SD AG VIF SCD MS_SSIM SE-Attention 6.8501 13.7001 34.7315 3.7415 0.9612 1.8192 0.9416 ECA-Attention 6.9504 13.9008 40.6716 3.5503 0.9548 1.6514 0.9148 No-Attention 7.1774 14.3547 40.2576 4.2096 1.3605 1.8801 0.9306 No-Comp Encoder 6.8728 13.7456 38.0053 2.9682 0.9219 1.8060 0.9403 No-Fusion Module 7.2129 14.4259 40.8263 3.7076 1.2727 1.8870 0.9277 No-Strategy 7.0291 14.0581 37.9805 3.3666 1.0851 1.8935 0.9511 No-Strategy&
No-Multi-scale
6.5910 13.1819 35.0411 3.2785 0.8063 1.6111 0.9216 Proposed 7.2525 14.5050 43.6183 4.1627 1.4128 1.8715 0.9207
Table 2. Average quality evaluation metrics for 40 pairs of TNO fused images
View tableView in ArticleTable 2. Average quality evaluation metrics for 40 pairs of TNO fused images
EN MI SD AG VIF SCD MS_SSIM DenseFuse 6.7933 13.5867 33.0055 3.8958 1.1141 1.8951 0.9379 RFN-Nest 6.9889 13.9777 35.2026 2.8619 1.1218 1.8710 0.9138 FusionGAN 6.5073 13.0147 26.4478 2.4276 0.7408 1.1339 0.7511 U2Fusion 6.4745 12.9489 24.9858 3.8414 0.7371 1.6542 0.9433 CSF 6.9295 13.8591 34.2342 3.8565 1.2221 1.8503 0.9190 SEDR 6.8795 13.7590 39.0527 4.1491 1.5546 1.8554 0.8946 SwinFusion 6.6156 13.2311 31.0339 3.4971 0.7228 1.7147 0.8992 SeAFusion 7.1474 14.2949 39.9100 5.6162 1.7369 1.7323 0.8553 CDDFuse 7.0582 14.1164 39.1751 5.2135 1.3976 1.7966 0.8795 Proposed 7.3544 14.7088 44.5808 5.9907 1.8274 1.8784 0.8935
Table 3. Average quality evaluation metrics for 100 pairs of RoadScene fused images
View tableView in ArticleTable 3. Average quality evaluation metrics for 100 pairs of RoadScene fused images
EN MI SD AG VIF SCD MS_SSIM DenseFuse 7.2684 14.5368 43.1679 4.5979 0.6504 1.6552 0.9220 RFN-Nest 7.3492 14.6984 46.1085 3.1668 0.6091 1.6837 0.8671 FusionGAN 7.0540 14.1079 39.0645 3.2246 0.4805 1.0496 0.7547 U2Fusion 7.0815 14.1629 37.8472 5.2514 0.6338 1.3866 0.9148 CSF 7.4257 14.8514 47.9828 5.0241 0.7952 1.7282 0.9261 SEDR 7.4499 14.8999 49.3581 4.9322 0.8286 1.6978 0.9005 SwinFusion 6.9712 13.9424 45.1634 4.3312 0.7275 1.5772 0.8470 SeAFusion 7.5117 15.0234 56.1119 6.8737 1.0949 1.6732 0.8786 CDDFuse 7.5003 15.0007 56.4331 6.5083 1.1120 1.7105 0.8740 Proposed 7.5623 15.1246 53.1000 6.5506 1.0359 1.7766 0.9353
Table 4. Average quality evaluation metrics for 100 pairs of LLVIP fused images
View tableView in ArticleTable 4. Average quality evaluation metrics for 100 pairs of LLVIP fused images
EN MI SD AG VIF SCD MS_SSIM DenseFuse 6.9740 13.9480 36.7708 2.8641 0.4617 1.3835 0.9224 RFN-Nest 7.0028 14.0055 37.8919 2.2445 0.4216 1.4154 0.8981 FusionGAN 6.4153 12.8306 26.1540 2.0102 0.2746 0.7521 0.7865 U2Fusion 6.6531 13.3062 34.9659 3.3391 0.5283 1.2755 0.9098 CSF 6.8283 13.6566 35.3773 2.7960 0.4564 1.3621 0.9109 SEDR 6.8877 13.7795 36.5430 2.6319 0.4495 1.2546 0.8834 SwinFusion 7.3844 14.7688 50.8104 4.2064 0.9087 1.5887 0.9451 SeAFusion 7.4193 14.8386 50.4468 4.1898 0.9062 1.6259 0.9435 CDDFuse 7.3134 14.6267 48.3450 3.8052 0.8171 1.5889 0.9337 Proposed 7.5398 15.0795 52.9415 3.8464 0.9889 1.7071 0.9250
Tools
Get Citation
Copy Citation Text
Hai-Ping SI, Wen-Rui ZHAO, Ting-Ting LI, Fei-Tao LI, Bacao FERNADO, Chang-Xia SUN, Yan-Ling LI. BDMFuse: Multi-scale network fusion for infrared and visible images based on base and detail features[J]. Journal of Infrared and Millimeter Waves, 2025, 44(2): 275
Paper Information
Category: Interdisciplinary Research on Infrared Science
Received: Jul. 25, 2024
Accepted: --
Published Online: Mar. 14, 2025
The Author Email: Yan-Ling LI (lyl_lingling@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20