Noise-Resilient Multi-Focus Image Fusion Using A Siamese Network With A Learnable Prefusion Shield
DOI:
https://doi.org/10.29284/2yaem287Keywords:
Multi-focus image fusion, Siamese network, Noise resilience, Learnable pre-fusion shield, Lightweight CNN, Zero-shot resolution transfer, Edge preservation and Perceptual quality.Abstract
Image fusion networks for multi-focus images usually assume clean inputs. In real-world conditions, images often contain noise. This causes a significant drop in fusion performance. Despite its importance, noise robustness in multi-focus image fusion is still underexplored. In this work, we propose a noise-resilient fusion framework with two main components. First, we design a Siamese-based fusion baseline. The model achieves 31.43 dB PSNR on the Lytro dataset and 36.4 dB on the RealMFF dataset. These results are comparable to recent state-of-the-art methods. Second, we introduce a lightweight pre-fusion module called a fusion shield. The module can be added to existing fusion networks without modifying their architecture. It acts as an input-adaptive noise suppression layer before fusion. The shield contains only 1,027 parameters and adds negligible computational overhead. When combined with the baseline, it improves robustness under noisy conditions. We observe up to a +3.65 dB gain under high-density impulse noise. At the same time, performance on clean images remains almost unchanged, with only −0.01 dB variation. The same module handles multiple noise types, including impulse, mixed and Gaussian noise. No retraining of the base fusion model is required. These results show that a simple drop-in module can improve the robustness of existing fusion networks and make them more suitable for real-world imaging conditions.
References
[1]. X. Zhang, "Deep Learning-Based Multi-Focus Image Fusion: A Survey and a Comparative Study," IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021.
[2]. B. Cao, B. Cai, C. Zhang, and Q. Hu, "Dig2DIG: Dig into Diffusion Information Gains for Image Fusion," arXiv preprint arXiv:2503.18627, 2025.
[3]. A. Gómez‑Ríos, S. Luque, J. González, and E. López‑Rubio, "The Impact of Noise and Brightness on Object Detection Methods," Sensors, vol. 24, no. 3, p. 821, 2024.
[4]. H. Kaur and D. Koundal, "Reliable and robust low rank representation based noisy images multi‑focus image fusion," Multimedia Tools and Applications, vol. 82, pp. 8235–8259, 2023.
[5]. J. Rudahunga, Y. Tang, and J. Park, "Multi‑Channel Blind Restoration of Mixed Noise Images under Atmospheric Turbulence," Atmosphere, vol. 13, no. 11, p. 1842, 2022.
[6]. Ma, Jiayi et al. “FusionGAN: A generative adversarial network for infrared and visible image fusion.” Inf. Fusion 48 (2019): 11-26.
[7]. Zhai, Hao and Zhuang Yi. “Multi-focus image fusion method using energy of Laplacian and a deep neural network.” Applied optics 59 6 (2020): 1684-1694.
[8]. Li, Bo et al. “Multi-focus image fusion based on dynamic threshold neural P systems and surfacelet transform.” Knowl. Based Syst. 196 (2020): 105794.
[9]. H. Xu, J. Ma, J. Jiang, X. Guo and H. Ling, "U2Fusion: A Unified Unsupervised Image Fusion Network," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 1, pp. 502-518, 1 Jan. 2022.
[10]. Li, Lin et al. “Multi-Focus Image Fusion with Convolutional Neural Network Based on Dempster-Shafer Theory.” Optik (2022).
[11]. Wei, Bingzhe, et al. "The multi-focus-image-fusion method based on convolutional neural network and sparse representation." Entropy 23.7 (2021): 827.
[12]. Liu, Deyu and Changhao Ma. “Multi-focus image fusion method based on U-net and multi-scale pyramid attention mechanism.” Proceedings of the 4th International Conference on Artificial Intelligence and Computer Engineering (2023).
[13]. Wang, Xuejiao et al. “Multi-focus image fusion framework based on transformer and feedback mechanism.” Ain Shams Engineering Journal (2022.
[14]. Wu, Pan et al. “Multi-focus image fusion: Transformer and shallow feature attention matters.” Displays 76 (2022): 102353.
[15]. Yang, Chunfeng et al. “Multi-Scale Visual Attention Cascaded Network for Multi-Focus Image Fusion.” 2023 4th International Conference on Information Science and Education (ICISE-IE) (2023): 89-92.
[16]. Li, H.; Yang, T.; Wang, R.; Li, C.; Zhou, S.; Guo, X. Residual Attention-Based Image Fusion Method with Multi-Level Feature Encoding. Sensors 2025, 25, 717.
[17]. Ge, Lei and Lei Dou. “G-Loss: a loss function with gradient information for super-resolution.” Optik (2023).
[18]. Nikulin, Dmitry et al. “Perceptual Gradient Networks.” ArXiv abs/2105.01957 (2021).
[19]. An, Tai et al. “Patch loss: A generic multi-scale perceptual loss for single image super-resolution.” Pattern Recognit. 139 (2023): 109510.
[20]. Shi, Haoyue et al. “Loss functions for pose guided person image generation.” Pattern Recognit. 122 (2022): 108351.
[21]. Manongga, William Eric et al. “Optimal Contrast Enhancement with Linear Image Fusion and Patch-Based Contrast Quality Index.” 2023 International Conference on Consumer Electronics - Taiwan (ICCE-Taiwan) (2023): 787-788.
[22]. Kumar, Vinit et al. “A Two-Level Hybrid Image Fusion Technique for Color Image Contrast Enhancement.” 2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT) (2021): 1-6.
[23]. Hessel, Charles and Jean-Michel Morel. “An Extended Exposure Fusion and its Application to Single Image Contrast Enhancement.” 2020 IEEE Winter Conference on Applications of Computer Vision (WACV) (2020): 137-146.
[24]. Saravanan, Vimala, et al. "Advancements in Image Enhancement: Comparing Spatial and Frequency Domain Methods Using COPRAS Analysis."
[25]. Nagaiah, Dr. K. “Image Quality and Performance Analysis Using Frequency Domain Technique.” International Journal of Innovative Science and Modern Engineering (2025): n. pag.
[26]. Satyanarayana, V. et al. “Image Fusion Using Contrast-Enhanced Images Captured in Low Light and Infrared.” 2023 International Conference on Advanced & Global Engineering Challenges (AGEC) (2023): 11-15.
[27]. Zhang, Bozhi et al. “Infrared and Visible Image Fusion Based on Contrast Enhancement Guided Filter and Infrared Feature Decomposition.” Infrared Physics & Technology (2022).
[28]. Rudahunga, Christian, Henry Kiragu, and Mary Ahuna. "Fusion Based MR Images Denoising Technique Using Frequency Domain and Non-Local Means Filters." WSEAS Transactions on Signal Processing 18 (2022): 153-163.
[29]. Bouzos, Odysseas, Ioannis Andreadis, and Nikolaos Mitianoudis. "A convolutional neural network-based conditional random field model for structured multi-focus image fusion robust to noise." IEEE Transactions on Image Processing 32 (2023): 2915-2930.
[30]. Pan, Xin, et al. "Improving multi-focus image fusion through Noisy image and feature difference network." Image and Vision Computing 142 (2024): 104891.
[31]. Zhang, Xingchen. "Multi-focus image fusion: A benchmark." arXiv preprint arXiv:2005.01116 (2020).
[32]. Wang, Y., et al. "LightMFF: Lightweight Multi-Focus Image Fusion via Edge-Preserving." IEEE Transactions on Image Processing, 2023.
[33]. Zhang, Y., et al. "Deep Residual Parallel Learning for Multi-Focus Image Fusion." IEEE Transactions on Image Processing, 2020.
[34]. Zhou, H., et al. "Gradient Aware Convolutional Network for Multi-Focus Image Fusion." IEEE Transactions on Image Processing, 2021.
[35]. Li, J., et al. "SwinFusion: Cross-domain Long-range Learning for General Image Fusion." IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022.
[36]. Ma, J., et al. "MUFusion: A Multi-units Fusion Network for Infrared and Visible Image Fusion." IEEE Transactions on Image Processing, 2021.
[37]. Zhang, Q., et al. "A Cross-view Transform for Multi-focus Image Fusion." Pattern Recognition, 2011.
[38]. Li, H., et al. "Multi-level Deep Laplacian Regularization for Multi-Focus Image Fusion with Random Forest Model." IEEE Transactions on Image Processing, 2020.
[39]. Xu, H., et al. "U2Fusion: A Unified Unsupervised Image Fusion Network." IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020.
[40]. Saharia, C., et al. "Image Super-Resolution via Iterative Refinement." IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
