To address the issues of fixed and monotonous information selection strategies in current image fusion algorithms, which lead to the loss of critical source image information and interference from invalid noise degrading fusion quality, this paper proposed an interpretable infrared and visible image fusion method based on Class Activation Mapping (CAM). By leveraging the CAM mechanism, class activation weights were derived from different source images, reflecting the network’s attention to feature importance. These weights were utilized to assign channel-specific feature priorities, enabling weighted fusion of deep features to preserve richer salient targets, texture details, and critical information from source images while suppressing noise. Experimental results demonstrate that the proposed method outperformed most state-of-the-art algorithms on the TNO and RoadScene datasets. On the TNO dataset, it achieves superior information entropy (EN) and visual fidelity (VIF) scores of 7.327 2 and 0.692 7, respectively, significantly surpassing existing approaches. This indicates that the proposed method effectively retains key features of source images while exhibiting exceptional visual perception performance.
HA Q, WATANABEK, KARASAWAT, et al. MFNet: Towards real-time semantic segmentation for autonomous vehicles with multi-spectral scenes[C]//2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017: 5108-5115.
[2]
TANGL, XIANGX, ZHANGH, et al. DIVFusion: Darkness-free infrared and visible image fusion[J]. Information Fusion, 2023, 91: 477-493.
[3]
CHENJ, LIX, LUOL, et al. Infrared and visible image fusion based on target-enhanced multiscale transform decomposition[J]. Information Sciences, 2020, 508: 64-78.
[4]
WUM, MAY, FANF, et al. Infrared and visible image fusion via joint convolutional sparse representation[J]. Journal of the Optical Society of America A, 2020, 37(7): 1105-1115.
[5]
LIS, ZOUY, WANGG, et al. Infrared and visible image fusion method based on a principal component analysis network and image pyramid[J]. Remote Sensing, 2023, 15(3): 685.
[6]
CHENJ, WUK, CHENGZ, et al. A saliency-based multiscale approach for infrared and visible image fusion[J]. Signal Processing, 2021, 182: 107936.
[7]
LIX, TANH, ZHOUF, et al. Infrared and visible image fusion based on domain transform filtering and sparse representation[J]. Infrared Physics & Technology, 2023, 131: 104701.
[8]
SONGZ, QINP, ZENGJ, et al. EdgeFusion: Infrared and visible image fusion algorithm in low light[C]//Pattern Recognition and Computer Vision . Singapore: Springer Nature Singapore, 2023: 259-270.
[9]
LIH, WUX J. DenseFuse: A fusion approach to infrared and visible images[J]. IEEE Transactions on Image Processing, 2019, 28(5): 2614-2623.
[10]
LIH, WUX J, KITTLERJ. RFN-Nest: An end- to-end residual fusion network for infrared and visible images[J]. Information Fusion, 2021, 73: 72-86.
[11]
XUH, WANGX, MAJ. DRF: Disentangled representation for visible and infrared image fusion[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-13.
[12]
XUH, MAJ, JIANGJ, et al. U2Fusion: A unified unsupervised image fusion network[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(1): 502-518.
[13]
MAJ, TANGL, XUM, et al. STDFusionNet: An infrared and visible image fusion network based on salient target detection[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-13.
[14]
TANGL, YUANJ, ZHANGH, et al. PIAFusion: A progressive infrared and visible image fusion network based on illumination aware[J]. Information Fusion, 2022, 83: 79-92.
[15]
ZHAOW, XIES, ZHAOF, et al. MetaFusion: Infrared and visible image fusion via meta-feature embedding from object detection[C]//2023 the IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), 2023: 13955-13965.
[16]
MAJ, YUW, LIANGP, et al. FusionGAN: A generative adversarial network for infrared and visible image fusion[J]. Information Fusion, 2019, 48: 11-26.
[17]
MAJ, XUH, JIANGJ, et al. DDcGAN: A dual-discriminator conditional generative adversarial network for multi-resolution image fusion[J]. IEEE Transactions on Image Processing, 2020, 29: 4980-4995.
[18]
MAJ, ZHANGH, SHAOZ, et al. GANMcC: A generative adversarial network with multiclassification constraints for infrared and visible image fusion[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70: 1-14.
[19]
RAOD, XUT, WUX J. TGFuse: An infrared and visible image fusion approach based on transformer and generative adversarial network[J]. IEEE Transactions on Image Processing, 2023: 3273451.
ZHOUB, KHOSLAA, LAPEDRIZAA, et al. Learning deep features for discriminative localization[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2016: 2921-2929.
[23]
TOETA. The TNO multiband image data collection[J]. Data in Brief, 2017, 15: 249-251.
JINHaibin, ZhixianLÜ, HOUMuzhou, et al. Research on endoscopic image classification algorithm based on feature fusion and SVM[J]. Journal of North University of China(Natural Science Edition), 2023, 44(1): 86-96.(in Chinese)
[26]
WANGC, XIAOJ, HANY, et al. Towards learning spatially discriminative feature representations[C]//2021 IEEE/CVF International Conference on Computer Vision(ICCV), 2021: 1306-1315.
[27]
LIH, WUX J, KITTLERJ. MDLatLRR: A novel decomposition method for infrared and visible image fusion[J]. IEEE Transactions on Image Processing, 2020, 29: 4733-4746.
[28]
WANGD, LIUJ, LIUR, et al. An interactively reinforced paradigm for joint infrared-visible image fusion and saliency object detection[J]. Information Fusion, 2023, 98: 101828.
[29]
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.
[30]
TANGW, HEF, LIUY, et al. DATFuse: Infrared and visible image fusion via dual attention transformer[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2023, 33(7): 3159-3172.