数字农科院2.0

MSP-MVS: Multi-Granularity Segmentation Prior Guided Multi-View Stereo

文献类型： 外文期刊

作者： Zhenlong Yuan;Cong Liu;Fei Shen;Zhaoxin Li;Jinguo Luo;Tianlu Mao;Zhaoqi Wang

作者机构：

关键词： 3D Reconstruction; Multi-view stereo; Low-texture; PatchMatch

期刊名称： Proceedings of the AAAI Conference on Artificial Intelligence

ISSN： 2159-5399

年卷期： 2025 年 39 卷 9 期

页码：

收录情况： EI（2025版）

摘要： Recently, patch deformation-based methods have demonstrated significant strength in multi-view stereo by adaptively expanding the reception field of patches to help reconstruct textureless areas. However, such methods mainly concentrate on searching for pixels without matching ambiguity (i.e., reliable pixels) when constructing deformed patches, while neglecting the deformation instability caused by unexpected edge-skipping, resulting in potential matching distortions. Addressing this, we propose MSP-MVS, a method introducing multi-granularity segmentation prior for edge-confined patch deformation. Specifically, to avoid unexpected edge-skipping, we first aggregate and further refine multi-granularity depth edges gained from Semantic-SAM as prior to guide patch deformation within depth-continuous (i.e., homogeneous) areas. Moreover, to address attention imbalance caused by edge-confined patch deformation, we implement adaptive equidistribution and disassemble-clustering of correlative reliable pixels (i.e., anchors), thereby promoting attention-consistent patch deformation. Finally, to prevent deformed patches from falling into local-minimum matching costs caused by the fixed sampling pattern, we introduce disparity-sampling synergistic 3D optimization to help identify global-minimum matching costs. Evaluations on ETH3D and Tanks & Temples benchmarks prove our method obtains state-of-the-art performance with remarkable generalization.

分类号：

  • 相关文献

[1]Image-Based Dynamic Quantification And High-Accuracy 3.D Evaluation Of Canopy S tructure Of Plant Populations. Hui, F, Zhu, JY, Hu, PC, Meng, L, Zhu, BL, Guo, Y, Li, BG, Ma, YT. 2018

[2]Evolution of holometaboly revealed by developmental transformation of internal thoracic structures in a green lacewing Chrysopa pallens (Neuroptera: Chrysopidae). Zhao, Chenjing,Wang, Mengqing,Gao, Caixia,Li, Min,Zhang, Kuiyan,Yang, Ding,Liu, Xingyue. 2021

[3]NeRF-based Polarimetric Multi-view Stereo. Cao, Jiakai,Yuan, Zhenlong,Mao, Tianlu,Wang, Zhaoqi,Li, Zhaoxin. 2024

[4]TSAR-MVS: Textureless-aware segmentation and correlative refinement guided multi-view stereo. Zhenlong Yuan,Jiakai Cao,Zhaoqi Wang,Zhaoxin Li. 2024

[5]Soybean yield estimation and lodging classification based on UAV multi-source data and self-supervised contrastive learning. Longyu Zhou,Yong Zhang,Haochong Chen,Guangyao Sun,Lei Wang,Mingxue Li,Xuhong Sun,Puyu Feng,Long Yan,Lijuan Qiu,Yinghui Li,Yuntao Ma. 2025

[6]Genetic resolution of multi-level plant height in common wheat using the 3D canopy model from ultra-low altitude unmanned aerial vehicle imagery. Shuaipeng Fei,Yidan Jia,Lei Li,Shunfu Xiao,Jie Song,Shurong Yang,Duoxia Wang,Guangyao Sun,Bohan Zhang,Keyi Wang,Junjie Ma,Jindong Liu,Yonggui Xiao,Yuntao Ma. 2025

[7]Soybean yield estimation and lodging discrimination based on lightweight UAV and point cloud deep learning. Longyu Zhou,Dezhi Han,Guangyao Sun,Yaling Liu,Xiaofei Yan,Hongchang Jia,Long Yan,Puyu Feng,Yinghui Li,Lijuan Qiu,Yuntao Ma. 2025

[8]Potential Correlation Between Bombus lantschouensis Thoracic Morphology and Flight Behavior. Li, Wenjie,Liu, Sipei,Zong, Le,Huang, Zhengzhong,Jiang, Lei,Liu, Xiaokun,Yang, Pingping,Zhang, Yitian,Du, Zhong,Fan, Weili,Qin, Zhuanghui,Wang, Xieshuang,Zhang, Xinying,Wang, Xiaolong,Yin, Haodong,An, Jiandong,Zhu, Chaodong,Orr, Michael C.,Wang, Jiangning,Ge, Siqin. 2025

作者其他论文 更多>>