MAPGR: Multi-Agent Prompt-Guided Residual Diffusion for ancient mural restoration
Source abstract
Abstract
Digital inpainting provides a non-destructive approach to ancient mural conservation, but reconstructed content must remain grounded in verifiable historical evidence under the minimum intervention principle. Existing diffusion-based restoration models often rely on unconstrained generative priors, which can introduce implausible content in severely degraded regions. We propose Multi-Agent Prompt Guided Restoration (MAPGR), an evidence-driven framework that reformulates mural restoration as a multi-stage reasoning process. MAPGR constrains generation with diagnostic evidence and incorporates a dual-constraint diffusion architecture with a Semantic Modulation Block (SMB) and Mask-Guided Self-Attention (MG-SA) for stylistic and spatial control. Experiments on the DUNHUANG and DhMurals datasets demonstrate that MAPGR achieves strong restoration performance while improving interpretability and historical plausibility, offering an evidence-driven computational framework for ancient mural conservation.
Source: openalex_inverted_index · Quality: baseline
Research context
OpenAlex topics: Building materials and conservation · Image Processing and 3D Reconstruction · Cultural Heritage Materials Analysis
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (42)
Show source-provided references
1. Chen, Y., Fan, Z. & Zhang, S. A novel Brownian bridge diffusion-based generative inpainting algorithm for ancient murals. Eng. Appl. Artif. Intell. 165, 113367 (2026).10.1016/j.engappai.2025.113367 ↗
2. Yu, Z. et al. SemiSketch: An ancient mural sketch extraction network based on reference prior and gradient frequency compensation. Pattern Recognit. 171, 112146 (2026).10.1016/j.patcog.2025.112146 ↗
3. Li, Y., Yu, Y. & Wu, X. A dual-encoder hierarchical feature fusion network for ancient mural disease detection. npj Herit. Sci. 13, 284 (2025).10.1038/s40494-025-01844-2 ↗
4. Qiu, S., Zhang, P., Li, S. & Hu, B. Tang dynasty mural restoration algorithms based on generative adversarial networks with multi-source information fusion. Inf. Fusion 127, 103883 (2026).10.1016/j.inffus.2025.103883 ↗
5. Kachkine, A. Physical restoration of a painting with a digitally constructed mask. Nature 642, 343–350 (2025).10.1038/s41586-025-09045-4 ↗
6. Xu, Z. et al. A comprehensive dataset for digital restoration of Dunhuang murals. Sci. Data 11, 955 (2024).10.1038/s41597-024-03785-0 ↗
7. Piatkowska, K. & Darecka, K. Restoring the past digitally: A 3d visualization study of the Great Mill in Gdansk. Des. Stud. 103, 101376 (2026).10.1016/j.destud.2025.101376 ↗
8. Rossi, M. & Bournas, D. Structural health monitoring and management of cultural heritage structures: a state-of-the-art review. Appl. Sci. 13, 6450 (2023).10.3390/app13116450 ↗
9. Li, C. A partial differential equation-based image restoration method in environmental art design. Adv. Math. Phys. 2021, 4040497 (2021).
10. Bhele, S., Shriramwar, S. & Agarkar, P. An efficient texture-structure conserving patch matching algorithm for inpainting mural images. Multimed. Tools Appl. 82, 46741–46762 (2023).10.1007/s11042-023-15370-5 ↗
11. Zhao, F., Ren, H., Sun, K. & Zhu, X. GAN-based heterogeneous network for ancient mural restoration. Herit. Sci. 12, 418 (2024).10.1186/s40494-024-01517-6 ↗
12. Cao, J., Zhang, Z., Zhao, A., Cui, H. & Zhang, Q. Ancient mural restoration based on a modified generative adversarial network. Herit. Sci. 8, 1–14 (2020).10.1186/s40494-020-0355-x ↗
13. Cao, J., Peng, C., Yang, Z. & Wang, F. Mural classification network based on the fusion of shallow feature enhancement and dense attention aggregation. npj Herit. Sci. 13, 136 (2025).10.1038/s40494-025-01692-0 ↗
14. Suvorov, R. et al. Resolution-robust large mask inpainting with Fourier convolutions. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, 2149–2159 (2022).10.1109/WACV51458.2022.00323 ↗
15. Zhang, K. et al. Mc-blur: A comprehensive benchmark for image deblurring. IEEE Trans. Circuits Syst. Video Technol. 34, 3755–3767 (2023).10.1109/TCSVT.2023.3319330 ↗
16. Zhang, K. et al. Adversarial spatio-temporal learning for video deblurring. IEEE Trans. Image Process. 28, 291–301 (2018).10.1109/TIP.2018.2867733 ↗
17. Cui, Y. & Knoll, A. Dual-domain strip attention for image restoration. Neural Netw. 171, 429–439 (2024).10.1016/j.neunet.2023.12.003 ↗
18. Cui, Y., Ren, W. & Knoll, A. Omni-kernel network for image restoration. In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, 1426–1434 (2024).10.1609/aaai.v38i2.27907 ↗
19. Jin, Z., Qiu, Y., Zhang, K., Li, H. & Luo, W. Mb-taylorformer v2: Improved multi-branch linear transformer expanded by Taylor formula for image restoration. IEEE Trans. Pattern Anal. Mach. Intell. 47, 5990–6005 (2025).10.1109/TPAMI.2025.3559891 ↗
20. Cui, Y. & Knoll, A. Exploring the potential of channel interactions for image restoration. Knowl.-Based Syst. 282, 111156 (2023).10.1016/j.knosys.2023.111156 ↗
21. Cui, Y., Wang, Q., Li, C., Ren, W. & Knoll, A. Eenet: an effective and efficient network for single image dehazing. Pattern Recognition. 158, 111074 (2025).10.1016/j.patcog.2024.111074 ↗
22. Nair, N. G., Mei, K. & Patel, V. M. AT-DDPM: Restoring faces degraded by atmospheric turbulence using denoising diffusion probabilistic models. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 3434–3443 (2023).
23. Garber, T. & Tirer, T. Image restoration by denoising diffusion models with iteratively preconditioned guidance. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 25245–25254 (2024).10.1109/CVPR52733.2024.02385 ↗
24. Li, X. et al. Diffusion models for image restoration and enhancement: A comprehensive survey. Int. J. Comput. Vis. 133, 8078–8108 (2025).10.1007/s11263-025-02570-9 ↗
25. Wei, P. TransDDPM: Transformer-based denoising diffusion probabilistic model for image restoration. In Chinese Conference on Pattern Recognition and Computer Vision (PRCV), 250–263 (Springer, 2023).10.1007/978-981-99-8552-4_20 ↗
26. Jiang, C., Ren, T. & Cheng, Z. All-in-one mural restoration with prompt-guided residual diffusion. npj Herit. Sci. 13, 667 (2025).10.1038/s40494-025-02232-6 ↗
27. Tang, X., Sui, Y., Sun, K. & Xiang, L. DiffInpaint: line drawing guided murals restoration with diffusion model. Measurement 119223 (2025).10.1016/j.measurement.2025.119223 ↗
28. Lei, T., Li, Y., Cheng, Z. & Zhou, W. Low-rank structure guided diffusion for Shaanxi temple mural restoration. npj Herit. Sci. 13, 295 (2025).10.1038/s40494-025-01827-3 ↗
29. Li, Y., Zhang, C., Li, Y., Sui, D. & Guo, M. An improved mural image restoration method based on diffusion model. npj Herit. Sci. 13, 347 (2025).10.1038/s40494-025-01914-5 ↗
30. Xu, Z. et al. MuralAgent: Enhancing ancient mural outpainting with RAG-based texts and multimodal integration. ACM Trans. Multimed. Comput. Commun. Appl. 21, 1–17 (2025).
31. Zhang, L., Rao, A. & Agrawala, M. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 3836–3847 (2023).10.1109/ICCV51070.2023.00355 ↗
32. Xu, Z. et al. Automated mural restoration via semi supervised segmentation and prompt guided diffusion inpainting. npj Herit. Sci. 13, 504 (2025).10.1038/s40494-025-02047-5 ↗
33. Wen, Y. et al. Structure-preserving frequency-regularized text-guided optimal transport for unpaired rain streaks and raindrops removal. IEEE Trans. Multimed. (2026).10.1016/j.patcog.2025.112862 ↗
34. Wen, Y. et al. Cross-level interaction and intra-level fusion network for remote sensing image dehazing. IEEE Trans. Geosci. Remote Sens. 63, 1–14 (2025).
35. Li, L. et al. Line drawing-guided progressive inpainting for mural damage. ACM J. Comput. Cult. Herit. 18, 1–20 (2025).
36. Zheng, S. Intangible heritage restoration of damaged tomb murals through augmented reality technology: A case study of Zhao Yigong tomb murals in tang dynasty of China. J. Cult. Herit. 69, 135–147 (2024).10.1016/j.culher.2024.08.005 ↗
37. Yan, Y., Chai, B. & Li, J. ArtDiff: Integrating IoT and AI to enhance precision in ancient mural restoration. Alex. Eng. J. 111, 511–520 (2025).10.1016/j.aej.2024.09.120 ↗
38. Li, X., Lu, D. & Pan, Y. Color restoration and image retrieval for Dunhuang fresco preservation. IEEE Multimed. 7, 38–42 (2000).10.1109/93.848425 ↗
39. Yu, T. et al. Dunhuang grottoes painting dataset and benchmark. arXiv preprint arXiv:1907.04589, (2019).
40. Xu, Z. et al. MuralDiff: Diffusion for ancient murals restoration on large-scale pre-training. IEEE Trans. Emerg. Top. Comput. Intell. 8, 2169–2181 (2024).10.1109/TETCI.2024.3359038 ↗
41. Akegarasu. lora-scripts: Scripts for training LORA (low-rank adaptation) for stable diffusion. https://github.com/Akegarasu/lora-scripts (2024).
42. Luo, Z., Gustafsson, F. K., Zhao, Z., Sjölund, J. & Schön, T. B. Image restoration with mean-reverting stochastic differential equations. 23045–23066 (ICML, 2023).