Research on Restoration and Completion Algorithm of Cultural Relics Based on Generative Adversarial Network and Diffusion Model
DOI:
https://doi.org/10.67541/jdc2613Keywords:
Restoration of cultural relics; Generative adversarial network; Diffusion model; Decision level integration; Dynamic routingAbstract
A decision-level fusion framework was proposed to solve the problem that the texture generated by the generative adversarial network (GAN) is sharp, but the large-scale loss is easy to drift, and the complementarity of the diffusion model is coherent but the fine-grained texture is easy to over-smooth in the restoration of cultural relics. An asymmetric dual-stream generator with feature representation isolation is constructed, and a defect-aware routing controller is designed. The local edge density, fragment compactness, and neighborhood semantic gradient are taken as inputs, and the path is generated by pixelwise allocation through online preference mining. The routing weight modulates the cross-attention of the diffusion stream to the adversarial stream texture features by multiplicative gating, and reverse-infuses the low-level structure prior of the diffusion stream. An adaptive fusion operator guided by routing gradient is proposed to generate the bandwidth point by point according to the weighted gradient tensor and stretch it anisotropic along the main direction of the structure tensor. The three style anchoring constraints of stroke direction, pigment color moment and texture energy spectrum slope are introduced. On the hierarchical incomplete test set, the PSNR of the method reaches 28.94 dB and 25.31 dB in the transition interval from L2 to L3, which is 1.48 dB and 1.29 dB higher than the optimal shallow fusion baseline. The boundary LPIPS is reduced to 0.087, and the style distance is reduced to 0.108. Ablation experiments verify the independent contribution of routing mechanism and feature-level interaction and show that decision-level fusion can effectively realize the complementarity of the two streams under the capability constraints of the existing generation model.
References
[1] Dong, G., Huang, W., Smith, W. A., & Ren, P. (2020). A shadow constrained conditional generative adversarial net for SRTM data restoration. Remote Sensing of Environment, 237, 111602. https://doi.org/10.1016/j.rse.2019.111602
[2] Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., Busababodhin, P., & Du, Y. (2025). Hybrid convolutional neural network-graph attention network-gradient boosting decision tree model for seismic impedance inversion prediction. Journal of Seismic Exploration, 34(5), 81. https://doi.org/10.36922/JSE025310051
[3] Saxena, D., & Cao, J. (2021). Generative adversarial networks (GANs) challenges, solutions, and future directions. ACM Computing Surveys (CSUR), 54(3), 1-42. https://doi.org/10.1145/3446374
[4] Deng, Z., Zhang, H., Liang, X., Yang, L., Xu, S., Zhu, J., & Xing, E. P. (2017). Structured generative adversarial networks. Advances in neural information processing systems, 30. https://doi.org/10.48550/arXiv.1711.00889
[5] Wang, C., Xu, C., Yao, X., & Tao, D. (2019). Evolutionary generative adversarial networks. IEEE Transactions on Evolutionary Computation, 23(6), 921-934. https://doi.org/10.1109/tevc.2019.2895748
[6] Li, R., Chen, L., Zhang, Z., Jampani, V., Patel, V. M., & Zhang, L. (2025, April). Syncnoise: Geometrically consistent noise prediction for instruction-based 3d editing. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 5, pp. 4905-4913). https://doi.org/10.1609/aaai.v39i5.32519
[7] Liu, S., Yu, C., Cao, C., Qian, W., & Wang, F. (2024, September). VCD-Texture: Variance alignment based 3D-2D co-denoising for text-guided texturing. In European Conference on Computer Vision (pp. 373-389). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-72640-8_21
[8] Zhao, T., Chen, G., Pang, C., Seenoi, P., Papukdee, N., & Busababodhin, P. (2025). Time-lapse earthquake difference prediction based on physics-informed long short-term memory coupled with interpretability boosting. Journal of Seismic Exploration, 34(3), 25. https://doi.org/10.36922/JSE025310049
[9] Zheng, H., Yao, Y., Zhou, Y., Luo, J., & Lin, Z. (2026). Pixperfect: Seamless latent diffusion local editing with discriminative pixel-space refinement. Advances in Neural Information Processing Systems, 38, 128266-128292. https://doi.org/10.48550/arXiv.2512.03247
[10] Xu, L., Luo, Q., Chen, L., Ni, T., Lin, Q., Zhao, S., ... & Feng, H. (2026). From Adversarial Competition to Iterative Refinement: A Critical Review of Generative AI for Remote Sensing Scene Classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. https://doi.org/10.1109/jstars.2026.3670966
[11] Li, X., Ren, Y., Jin, X., Lan, C., Wang, X., Zeng, W., ... & Chen, Z. (2025). Diffusion models for image restoration and enhancement: A comprehensive survey. International Journal of Computer Vision, 133(11), 8078-8108. https://doi.org/10.1007/s11263-025-02570-9
[12] Ding, X., Chen, X., Sui, Y., Wang, Y., & Zhang, J. (2025). Underwater image enhancement using a diffusion model with adversarial learning. Journal of Imaging, 11(7), 212. https://doi.org/10.3390/jimaging11070212
[13] Lu, Z., Li, B., Fu, C., Wu, J., Xu, L., Jia, S., & Zhang, H. (2024, October). Remaining useful life prediction method based on dual-path interaction network with multiscale feature fusion and dynamic weight adaptation. In Actuators (Vol. 13, No. 10, p. 413). MDPI. https://doi.org/10.3390/act13100413
[14] Zhao, J., Liu, T., & Sun, L. (2025). Multi-scale fusion maxvit for medical image classification with hyperparameter optimization using super beluga whale optimization. Electronics, 14(5), 912. https://doi.org/10.3390/electronics14050912
[15] Ye, F., & Bors, A. G. (2022). Dynamic self-supervised teacher-student network learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(5), 5731-5748. https://doi.org/10.1109/tpami.2022.3220928
[16] Wang, Y., & Wang, X. (2024). Artificial intelligence in physical education: comprehensive review and future teacher training strategies. Frontiers in public health, 12, 1484848. https://doi.org/10.3389/fpubh.2024.1484848
[17] Liu, M., & He, Q. (2026). Relation-Aware and Constraint-Guided Discrete Diffusion for Design Color Scheme Generation. Journal of Discovery Core, 1(2), 80-109. https://doi.org/10.67541/jdc2610
[18] Liu, L. (2026). Research on Intelligent Generation Algorithm of Interface Icon Based on Diffusion Model. International Scientific Technical and Economic Research, 4(1), 149-167. https://doi.org/10.71451/ISTAER2607
[19] Du, J., & Liu, M. (2026). Intelligent Defect Detection of Design Materials Based on Edge AI. Journal of Digital Frontier, 1(2), 27-57. https://doi.org/10.67541/jdf2609
[20] Ma, X., & Wu, H. (2026). Construction and Real Time Status Monitoring of a Changxin Palace Lantern Digital Twin Based on Multi Mode Sensing Data Fusion. Journal of Digital Frontier, 1(1), 185-218. https://doi.org/10.67541/jdf2607
[21] Liu, Y., Feng, S., & Liu, M. (2026). Research on Transformer-Based Action Sequence Modeling of Intangible Cultural Heritage Shadow Play Using Attention Mechanisms. International Scientific Technical and Economic Research, 4(2), 51-77. https://doi.org/10.71451/ISTAER2615
[22] Mouadden, N., Laousy, O., Marini, R., Ong, V., Revel, M. P., Chassagnon, G., ... & Vakalopoulou, M. (2025, September). Conditional Latent Diffusion Models for Irregularly Spaced Longitudinal Radiological Data. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 106-115). Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-05182-0_11
[23] Hussain, M., O’Nils, M., Lundgren, J., & Mousavirad, S. J. (2024). A comprehensive review on deep learning-based data fusion. IEEE Access, 12, 180093-180124. https://doi.org/10.1109/access.2024.3508271
[24] Almutairi, A., & Owais, M. (2025). Reliable vehicle routing problem using traffic sensors augmented information. Sensors, 25(7), 2262. https://doi.org/10.3390/s25072262
[25] Wang, J., Zhang, L., Wang, H., Zhang, C., Su, H., Duan, J., & Hu, J. (2026). Multi-objective collaborative optimization of conical pick group based on factor space theory. Scientific Reports. https://doi.org/10.1038/s41598-026-64321-1
[26] Zhang, W., Zhao, Y., Breckon, T. P., & Chen, L. (2017). Noise robust image edge detection based upon the automatic anisotropic Gaussian kernels. Pattern Recognition, 63, 193-205. https://doi.org/10.1016/j.patcog.2016.10.008
[27] Xiao, C., Staring, M., Wang, Y., Shamonin, D. P., & Stoel, B. C. (2012). Multiscale bi-Gaussian filter for adjacent curvilinear structures detection with application to vasculature images. IEEE Transactions on Image Processing, 22(1), 174-188. https://doi.org/10.1109/tip.2012.2216277
[28] Qian, J., Yang, J., & Gao, G. (2013). Discriminative histograms of local dominant orientation (D-HLDO) for biometric image feature extraction. Pattern Recognition, 46(10), 2724-2739. https://doi.org/10.1016/j.patcog.2013.03.005
[29] Shu, C., Ding, X., & Fang, C. (2011). Histogram of the oriented gradient for face recognition. Tsinghua Science and Technology, 16(2), 216-224. https://doi.org/10.1016/s1007-0214(11)70032-3
[30] Robitaille, T. P. (2017). A modular set of synthetic spectral energy distributions for young stellar objects. Astronomy & Astrophysics, 600, A11. https://doi.org/10.1051/0004-6361/201425486
[31] Yu, R., Chen, W., & Zhou, D. (2020). Infrared and visible image fusion based on gradient transfer optimization model. IEEE Access, 8, 50091-50106. https://doi.org/10.1109/access.2020.2979760
[32] Zhang, H., Xu, H., Xiao, Y., Guo, X., & Ma, J. (2020, April). Rethinking the image fusion: A fast unified image fusion network based on proportional maintenance of gradient and intensity. In Proceedings of the AAAI conference on artificial intelligence (Vol. 34, No. 07, pp. 12797-12804). https://doi.org/10.1609/aaai.v34i07.6975
Published
Data Availability Statement
The data that support the findings of this study are available upon request from the corresponding authors, M.M.
Issue
Section
License
Copyright (c) 2026 Min Ma (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.