In order to fully integrate the complementary information of infrared and visible images and improve the feature preservation and task adaptability of fusion results, an enhancement-feature and Transformer-integrated dual-semantic image fusion method (enhancement-feature transformer-integrated dual semantic image fusion, ETDFusion) was proposed to investigate multimodal image fusion. Feature enhancement modules were designed to strengthen the edge details of infrared images and the brightness information of visible images. A hybrid architecture combining convolutional neural network(CNN) and Vision Transformer was constructed to achieve coordinated optimization of local feature extraction and global dependency modeling. In addition, a dual-semantic loss function was introduced by incorporating semantic segmentation and object detection constraints, and a three-stage training strategy was employed to guide the network in learning fusion features with strong semantic structures. The results show that the proposed method produces fused images with higher visual quality in terms of detail preservation and target saliency, and achieves superior performance to existing algorithms on multiple subjective and objective metrics. It is concluded that the ETDFusion network effectively balances multimodal information complementation and high-level visual task adaptation, providing an engineering-feasible technical pathway for infrared and visible image fusion. It can be further applied to perception tasks in complex environments such as security monitoring, intelligent transportation, and autonomous driving.
LI Shi-wei, TIAN Lin-bo, LIU Kun, XIE Liu-zhou, WANG Jian-qiang.
Dual-semantic-guided Fusion Framework for Infrared and Visible Images toward Advanced Vision Tasks[J].
Science Technology and Engineering, 2026, 26(17): 7365-7388 DOI:10.12404/j.issn.1671-1815.2507158
第二阶段,语义任务网络训练。固定融合网络参数,利用第一阶段输出的融合图像分别训练语义分割网络与目标检测网络。分割网络采用在线难例挖掘(online hard example mining, OHEM)策略增强对复杂语义区域的判别能力;检测网络则基于交并比指标进行优化。此阶段目的在于获得具备高质量语义理解能力的任务网络,为后续的联合优化提供稳定的监督信号。
ZhengBowen, WangZhuo, CaoXinyu. Infrared and visible image registration based on compound 2S network[J]. Science Technology and Engineering, 2024, 24(16): 6783-6791.
ZhuZiwen, SongXiao’ou, CuiWei, et al. Review of visible and infrared image fusion for intelligent object detection[J]. Computer Engineering and Applications, 2025, 61(17): 17-32.
LiLinhui, ZhangXinliang, FuYifan, et al. Research on visible light and infrared post-fusion detection based on TC-YOLOv7 algorithm[J]. Automotive Engineering, 2023, 45(12): 2280-2290.
WangFangbin, JinXun, ZhuDarong, et al. Infrared polarized face recognition based on RGB color space[J]. Laser and Optoelectronics Progress, 2022, 59(12): 535-543.
ShenYu, QiangZhenkai, WeiZiyi, et al. Review of infrared and visible image fusion based on deep learning: developments and prospects[J]. Control and Decision, 2025, 40(6): 1793-1806.
ZhangHonggang, YangHaitao, ZhengFengjie, et al. Review of feature-level infrared and visible image fusion[J]. Computer Engineering and Applications, 2024, 60(18): 17-31.
WuYifei, YangRui, LüQishen, et al. Infrared and visible image fusion: statistical analysis, deep learning approaches and future prospects[J]. Laser and Optoelectronics Progress, 2024, 61(14): 42-60.
[19]
LiuS, LanX, ChenW, et al. AGMFusion: a real-time end-to-end infrared and visible image fusion network based on adaptive guidance module[J]. IEEE Sensors Journal, 2024, 24(17): 28338-28350.
[20]
HeD, LiW, WangG, et al. MMIF-INet: multimodal medical image fusion by invertible network[J]. Information Fusion, 2025, 114: 102666.
LiShuhui, CaiWei, WangXin, et al. Review of infrared and visible image fusion methods in deep learning frameworks[J]. Computer Engineering and Applications, 2025, 61(9): 25-40.
ZhangZhouyu, CaoYunfeng, MengDing, et al. Infrared and visible image fusion via multi-layer convolutional sparse representation[J]. Journal of Harbin Institute of Technology, 2021, 53(12): 51-59.
HeYou, LiuYu, LiYaowen, et al. Development and prospects of multisource information fusion[J]. Acta Aeronautica et Astronautica Sinica, 2025, 46(6): 29-54.
MinLi, CaoSijian, ZhaoHuaici, et al. Infrared and visible image fusion with improved residual dense generative adversarial network[J]. Control and Decision, 2023, 38(3): 721-728.
[40]
HanK, WangY, ChenH, et al. A survey on vision transformer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 45(1): 87-110.
YangTianyu, HuoHongtao, LiuXiaowen, et al. Information interaction linear Transformer network for infrared and visible image fusion[J]. Journal of Image and Graphics, 2025, 30(7): 2499-2513.
[43]
RaoD, XuT, WuX. TGFuse: an infrared and visible image fusion approach based on transformer and generative adversarial network[J]. IEEE Transactions on Image Processing, 2023, 10: 3273451.
[44]
QuL, LiuS, WangM, et al. TransMEF: a Transformer-based multi-exposure image fusion framework using self-supervised multi-task learning[C]// Proceedings of the AAAI Conference on Artificial Intelligence. New York: AAAI, 2022: 2126-2134.
FengFucun, ChangLihong. Image fusion method of enhancement based on non-subsampled contourlet transform[J]. Science Technology and Engineering, 2020, 20(24): 9970-9975.
[50]
YueJ, FangL, XiaS, et al. Dif-Fusion: toward high color fidelity in infrared and visible image fusion with diffusion models[J]. IEEE Transactions on Image Processing, 2023, 32: 5705-5720.
JiangMai, Shaguijun, LiNing. Infrared and visible light images fusion based on Tetrolet transform and infrared saliency feature extraction[J]. Science Technology and Engineering, 2022, 22(30): 13398-13405.
[53]
TangH, LiuG, QianY, et al. EgeFusion: towards edge gradient enhancement in infrared and visible image fusion with multi-scale Transform[J]. IEEE Transactions on Computational Imaging, 2024, 10: 385-398.
ChengYihang, QiaoZhengyu, HuangYong, et al. Luminance-adaptive infrared and visible image fusion based on Retinex theory (invited)[J]. Laser & Optoelectronics Progress, 2024, 61(20): 127-140.
[56]
KhazalR B, HabeebN J. Enhancement infrared-visible image fusion using the integration of stationary wavelet transform and fuzzy histogram equalization[J]. Journal of Techniques, 2022, 4(4): 119-127.
HuJianping, HaoMengyun, DuYing, et al. Fusion of infrared and visible images via structure and texture-aware Retinex[J]. Optics and Precision Engineering, 2022, 30(24): 3225-3238.
TianBowen, DingJianwei, HuZirui. Low-light image enhancement and denoising algorithm integrating zero-reference depth curves[J]. Science Technology and Engineering, 2025, 25(2): 704-712.
[61]
ZhangH, TangL, XiangX, et al. Dispel darkness for better fusion: a controllable visual enhancer based on cross-modal conditional adversarial learning[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2024: 26487-26496.
[62]
TangL, YuanJ, MaJ. Image fusion in the loop of high-level vision tasks: a semantic-aware real-time infrared and visible image fusion network[J]. Information Fusion, 2022, 82: 28-42.
[63]
LiuJ, FanX, HuangZ, et al. Target-aware dual adversarial learning and a multi-scenario multi-modality benchmark to fuse infrared and visible for object detection[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 5802-5811.
[64]
ZhangX, ZhaiH, LiuJ, et al. Real-time infrared and visible image fusion network using adaptive pixel weighting strategy[J]. Information Fusion, 2023, 99: 101863.
[65]
TangL, LiC, MaJ. Mask-DiFuser: a masked diffusion model for unified unsupervised image fusion[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025, 48(1): 591-608.
[66]
LiuX, HuoH, LiJ, et al. A semantic-driven coupled network for infrared and visible image fusion[J]. Information Fusion, 2024, 108: 102352.
[67]
LiuJ, LiuZ, WuG, et al. Multi-interactive feature learning and a full-time multi-modality benchmark for image fusion and segmentation[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2023: 8115-8124.
[68]
ZhangH, MaJ. SDNet: a versatile squeeze-and-decomposition network for real-time image fusion[J]. International Journal of Computer Vision, 2021, 129(10): 2761-2785.
[69]
LiH, WuX. CrossFuse: a novel cross attention mechanism based infrared and visible image fusion approach[J]. Information Fusion, 2024, 103: 102147.