PGT-NeuS: Progressive-Growing Tri-Plane Representation for Neural Surface Reconstruction

doi:10.1109/TVCG.2025.3590394

	PGT-NeuS: Progressive-Growing Tri-Plane Representation for Neural Surface Reconstruction
	Xiang, Xue-Kun 1,2; Yuan, Yu-Jie 1,2; Hu, Wen-Bo 3; Liu, Yu-Tao 1,2; Ma, Yue-Wen 3; Gao, Lin 1,2
	2025-10-01
发表期刊	IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
ISSN	1077-2626
卷号	31 期号:10 页码:9213-9224
摘要	3D reconstruction from multi-view images is a long-standing problem in computer graphic. Neural 3D reconstruction, especially NeuS and its variants, has improved reconstruction quality compared to traditional methods. However, it is still a challenge for these methods to reconstruct fine-grained geometric details since the spherical harmonic positional encoding lacks the ability to express high-frequency signals. In this paper, we propose a multi-resolution tri-plane feature encoding that leverages the detail reconstruction capabilities of high-resolution tri-plane while using the smoothness of low-resolution tri-plane to suppress high-frequency artifacts. Additionally, a progressive training strategy is introduced, gradually merging scene details from coarse to fine granularity, enhancing reconstruction quality while maintaining training stability and reducing difficulty. Furthermore, to address reconstruction challenges arising from sparse viewpoints and inconsistent lighting in image datasets, we introduce normal priors as supervision and propose consistency verification for multi-view normal priors, which assesses the accuracy of normal priors and effectively supervise the reconstructed surfaces. Moreover, we propose a perturbing and fine-tuning strategy on regions of unreliable normal priors to further improve the quality of geometric surface reconstruction.
关键词	Image reconstruction Surface reconstruction Neural radiance field Encoding Training Rendering (computer graphics) Accuracy Three-dimensional displays Image color analysis Refining surface reconstruction progressive learning normal priors verification
DOI	10.1109/TVCG.2025.3590394
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China[62322210] ; Beijing Municipal Science and Technology Commission[Z231100005923031] ; Innovation Funding of ICT, CAS[E461020]
WOS研究方向	Computer Science
WOS类目	Computer Science, Software Engineering
WOS记录号	WOS:001566979000032
出版者	IEEE COMPUTER SOC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/41717
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Gao, Lin
作者单位	1.Chinese Acad Sci, Beijing Key Lab Mobile Comp & Pervas Device, Inst Comp Technol, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 101408, Peoples R China 3.ByteDance Pico, Beijing 100098, Peoples R China
推荐引用方式 GB/T 7714	Xiang, Xue-Kun,Yuan, Yu-Jie,Hu, Wen-Bo,et al. PGT-NeuS: Progressive-Growing Tri-Plane Representation for Neural Surface Reconstruction[J]. IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS,2025,31(10):9213-9224.
APA	Xiang, Xue-Kun,Yuan, Yu-Jie,Hu, Wen-Bo,Liu, Yu-Tao,Ma, Yue-Wen,&Gao, Lin.(2025).PGT-NeuS: Progressive-Growing Tri-Plane Representation for Neural Surface Reconstruction.IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS,31(10),9213-9224.
MLA	Xiang, Xue-Kun,et al."PGT-NeuS: Progressive-Growing Tri-Plane Representation for Neural Surface Reconstruction".IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS 31.10(2025):9213-9224.