新颖的视图合成技术迅速发展，从神经辐射场发展到3D高斯点云渲染（3DGS），后者提供了实时渲染和快速训练，且不牺牲视觉保真度。然而，3DGS依赖于准确的相机姿态和高质量点云初始化，这在稀疏视角场景中难以获得。尽管传统结构从运动（SfM）管道在这些设置中往往失效，现有基于学习的点估计替代方法通常需要可靠的参考视角，并且对姿态或深度误差敏感。在本工作中，我们提出了一种利用π³的鲁棒方法，这是一种无参考点云估计网络。我们整合了来自π³的密集初始化，并设计了一种正则化方案以减轻几何不准确性。具体而言，我们采用不确定性引导的深度监督、法线一致性损失和深度变形。实验结果表明，我们的方法在 Tanks and Temples、LLFF、DTU 和 MipNeRF360 数据集上实现了最先进的性能。
英文摘要：
Novel view synthesis has evolved rapidly, advancing from Neural Radiance Fields to 3D Gaussian Splatting (3DGS), which offers real-time rendering and rapid training without compromising visual fidelity. However, 3DGS relies heavily on accurate camera poses and high-quality point cloud initialization, which are difficult to obtain in sparse-view scenarios. While traditional Structure from Motion (SfM) pipelines often fail in these settings, existing learning-based point estimation alternatives typically require reliable reference views and remain sensitive to pose or depth errors. In this work, we propose a robust method utilizing π^3, a reference-free point cloud estimation network. We integrate dense initialization from π^3 with a regularization scheme designed to mitigate geometric inaccuracies. Specifically, we employ uncertainty-guided depth supervision, normal consistency loss, and depth warping. Experimental results demonstrate that our approach achieves state-of-the-art performance on the Tanks and Temples, LLFF, DTU, and MipNeRF360 datasets.
MeshSplats: Mesh-Based Rendering with Gaussian Splatting Initialization
Faculty of Mathematics and Computer Science, Jagiellonian University in Kraków（克拉科夫雅盖隆大学数学与计算机科学学院）；Department of Engineering, University of Cambridge（剑桥大学工程系）；IDEAS Research Institute（IDEAS研究所）
3D Gaussian Splatting with Fisheye Images: Field of View Analysis and Depth-Based Initialization
Department of Computing Sciences, Tampere University（塔尔基大学计算机科学系）；Department of Computer Science, Aalto University（阿尔托大学计算机科学系）；Department of Computer Science and Engineering, University of Oulu（奥卢大学计算机科学与工程系）
The Role of Initialization in 3D Gaussian Splatting
Czech Technical University in Prague, Faculty of Electrical Engineering（布拉格捷克技术大学电子工程系）；Czech Technical University in Prague, Czech Institute of Informatics, Robotics and Cybernetics（布拉格捷克技术大学信息技术、机器人学与自动化研究所）
Segmentation-Driven Initialization for Sparse-view 3D Gaussian Splatting
Mid Sweden University（中瑞典大学）；Technical University of Berlin（柏林技术大学）；Hochschule für Technik und Wirtschaft Berlin（柏林技术与经济高等学院）；Department of Computer and Electrical Engineering, Mid Sweden University（中瑞典大学计算机与电气工程系）；Department of Telecommunication Systems, Technical University of Berlin（柏林技术大学电信系统系）
Physically Plausible Human-Object Rendering from Sparse Views via 3D Gaussian Splatting
College of Computer Science, Zhejiang University（浙江大学计算机科学与技术学院）；Department of Computer Science and Engineering, The Hong Kong University of Science and Technology（香港科技大学计算机科学与工程系）
SolidGS: Consolidating Gaussian Surfel Splatting for Sparse-View Surface Reconstruction
Texas A&M University（德克萨斯农工大学）；OPPO US Research Center（OPPO美国研究中心）；Nanyang Technological University（南洋理工大学）；Hong Kong University of Science and Technology（香港科技大学）；The University of Hong Kong（香港大学）
GeoSplat: A Deep Dive into Geometry-Constrained Gaussian Splatting
Department of Applied Mathematics and Theoretical Physics, University of Cambridge（应用数学与理论物理系，剑桥大学）；Institut Camille Jordan, Université Claude Bernard Lyon 1（让·Camille Jordan研究所，克莱蒙费朗斯大学）
SparseStreet: Sparse Gaussian Splatting for Real-Time Street Scene Simulation
Peking University（北京大学）；Chinese Academy of Sciences（中国科学院）；University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Autonomous Driving Development, NIO（蔚来自动驾驶开发）
AD-GS: Anatomical Density-guided 3D Gaussian Splatting for Sparse-View Dental CBCT Reconstruction
Beijing University of Posts and Telecommunications（北京邮电大学）；Peking University Third Hospital（北京大学第三医院）；National Institute of Informatics（信息学研究所）；Peking University（北京大学）
Learning Fine-Grained Geometry for Sparse-View Splatting via Cascade Depth Loss
School of Computer Science The University of Sydney Sydney, Australia；School of Computer Science University of Bristol Bristol, United Kingdom；School of Science Edith Cowan University Perth, Australia
LGDWT-GS: Local and Global Discrete Wavelet-Regularized 3D Gaussian Splatting for Sparse-View Scene Reconstruction
National Institute of Standards and Technology（国家标准与技术研究院）；Department of Physics, Colorado State University（科罗拉多州立大学物理系）；Electrical Engineering Department, University of Colorado（科罗拉多大学电气工程系）
SparseLGS: Sparse View Language Embedded Gaussian Splatting
School of Mathematical Science, University of Science and Technology of China（科学技术大学数学科学学院）；Meta；Apple Inc.（苹果公司）；Alpha Labs at Goertek（Goertek的Alpha实验室）
