Y.A. Maniakov1, P. O. Arkhipov2, P. L. Stavtsev3

1–3 Orel Branch of Federal Research Center «Computer Science and Control» of the RAS (Orel, Russia)
1 maniakov_yuri@mail.ru; 2 arpaul@mail.ru; 3 pavelstavcev@gmail.com

Three-dimensional reconstruction refers to the process of generating and processing volumetric models of real-world objects and scenes based on available data about the target subject. Currently, this area represents one of the most rapidly developing and in-demand fields within computer vision. Techniques and tools for 3D reconstruction have already found widespread application across architecture, robotics, medicine, manufacturing, computer graphics, and numerous other domains of human activity. Constructing a volumetric model provides more comprehensive information about the original object, thereby facilitating its analysis, description, and subsequent manipulation.

However, due to existing limitations in scanning equipment, as well as the imperfections of current data acquisition and processing algorithms, reconstructed models often contain various defects. The majority of these are geometric in nature, including noise, out­liers, and incomplete data. Such defects can significantly degrade both the visual quality of the reconstructed object and its suitability for further analysis and practical applications.

Recent research increasingly focuses on addressing these issues using neural network-based approaches. For instance, models such as PointCleanNet, PointProNets, and PointFilter are specifically designed to denoise point clouds, while Point Completion Network, Point Projection Network, and Rl-GAN-Net aim to complete missing geometry. A key limitation of these models is their narrow specia­lization, as each is typically trained to address only one type of defect. A few more comprehensive solutions, such as DeCo, employ a dual-branch architecture that separately handles denoising and point completion tasks.

The objective of this work is to propose a conceptual framework for a unified and integrated post-processing method for 3D reconstructed models. This approach is designed to simultaneously address both noise reduction and data completion, thereby enhancing the overall quality and usability of the reconstructed models.

Maniakov Y.A., Arkhipov P.O., Stavtsev P.L. The information model of 3D-reconstruction postprocessing method. Highly Available Systems. 2025. V. 21. № 2. P. 66−73. DOI: https://doi.org/ 10.18127/j20729472-202502-06 (in Russian)

1. Medvedev M.V., Kirpichnikov A.P. Trexmernaya rekonstrukciya ob``ektov v sisteme texnicheskogo zreniya mobil`nogo robota. Vestnik Kazanskogo texnologicheskogo universiteta. 2014. T. 17. № 15. S. 326–330.
2. Chertkova Ya.V., Katasonov D.N., Batalova K.V. Obzor sredstv tryoxmernoj rekonstrukcii. Intere`kspo Geo-Sibir`. 2021. T 6. DOI: 10.33764/2618-981X-2021-6-293-298
3. Man`yakov Yu.A., Arxipov P.O., Stavcev P.L. Issledovanie i analiz sushhestvuyushhix metodov postobrabotki oblakov tochek. Sistemy` vy`sokoj dostupnosti. 2024. T. 20. № 3. S. 51–58. DOI: 10.18127/j20729472-202403-05
4. Marie-Julie Rakotosaona, Vittorio La Barbera, Paul Guerrero, Niloy J. Mitra, Maks Ovsjanikov. PointCleanNet: Learning to Denoise and Remove Outliers from Dense Point Clouds. arXiv:1901.01060v3. 2019. DOI: 10.48550/arXiv.1901.01060
5. Riccardo Roveri, A. Cengiz Öztirel, Ioana Pandele, Markus Gross. PointProNets: Consolidation of Point Clouds with Convolutional Neural Networks. EUROGRAPHICS. 2018. V. 37. № 2.
6. Dongbo Zhang, Xuequan Lu, Hong Qin, Ying He. Pointfilter: Point Cloud Filtering via Encoder-Decoder Modeling. arXiv:2002.05968v2. 2020. DOI: 10.48550/arXiv.2002.05968
7. Wentao Yuan, Tejas Khot, David Held, Christoph Mertz, Martial Hebert. PCN: Point Completion Network. arXiv:1808.00671v3. 2019. DOI: 10.48550/arXiv.1808.00671
8. Weichao Wu, Zhong Xie, Yongyang Xu, Ziyin Zeng, Jie Wan. Point Projection Network: A Multi-View-Based PointCompletion Net-work with Encoder-Decoder Architecture. Remote Sens. 2021. V. 13. № 23. P. 4917. DOI: 10.3390/rs13234917
9. Muhammad Sarmad, Hyunjoo Jenny Lee, Young Min Kim. RL-GAN-Net: A Reinforcement Learning Agent Controlled GAN Net-work for Real-Time Point Cloud Shape Completion. arXiv:1904.12304v1. 2019. DOI: 10.48550/arXiv.1904.12304
10. Antonio Alliegro, Diego Valsesia, Giulia Fracastoro, Enrico Magli, Tatiana Tommas. Denoise and Contrast for Category Agnostic Shape Completion. arXiv:2103.16671v1. 2021. DOI: 10.48550/arXiv.2103.16671
11. Chang A.X., Funkhouser T., Guibas L., Hanrahan P., Huang Q. and Li Z., Savarese S., Savva M., Song S., Su H., Xiao J., Yi L., Yu F. Sha­peNet: An Information-Rich 3D Model Repository. arXiv:1512.03012v1. 2015. DOI: 10.48550/arXiv.1512.03012