Efficient Multi-View Reconstruction of Large-Scale Scenes using Interest Points, Delaunay Triangulation and Graph Cuts

By Patrick Labatut, Jean-Philippe Pons, Renaud Keriven. 2007

259 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning a Multi-View Stereo Machine (2017, methodology): “These depth maps can then be fused using a variety of different techniques [38, 8, 33, 59, 30].”
Hierarchical Surface Prediction for 3D Object Reconstruction (2017, methodology): “Multi-view approaches alleviate this, by the use of data adaptive discretization of the volume with Delaunay tedrahedrization [20], voxel block hashing [24] or octrees [5, 29].”
Multi-view reconstruction preserving weakly-supported surfaces (2011, methodology): “Recent approaches to multi-view reconstruction [21, 4, 22, 14, 5, 9, 8] attain a degree of accuracy and completeness comparable to laser scans [18, 20].”
Semantic 3D Reconstruction with Continuous Regularization and Ray Potentials Using a Visibility Consistency Constraint (2016, background): “Using a discrete graph formulation [18, 36] propose to model the free space between the camera center and the measured depth with pairwise potentials.”
R3LIVE: A Robust, Real-time, RGB-colored, LiDAR-Inertial-Visual tightly-coupled state Estimation and mapping package (2021, methodology): “For meshing, we make use of the Delaunay triangulation and graph cuts [28] implemented in CGAL [29]14.”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, methodology): “There is a variety of methods for dealing with this problem such as voxel hashing (Nießner et al. (2013)) or a data adaptive discretization of the space in the form of a Delaunay triangulation (Labatut et al. (2007)).”
BakedSDF: Meshing Neural SDFs for Real-Time View Synthesis (2023, methodology): “In Figure 5 we present a qualitative comparison of our mesh with those obtained using COLMAP and Pajdla 2011; Labatut et al. 2007] to obtain a binary segmentation of the scene and then forms a triangle mesh as the surface between these regions.”
Photometric Bundle Adjustment for Dense Multi-view 3D Modeling (2014, background): “Fig 5 shows an additional results of a statue in the Rietberg Museum, Zurich, where we initialize the camera calibration with structurefrom-motion [27, 26] and compute an initial mesh via [12].”
Learning Where to Classify in Multi-view Semantic Segmentation (2014, unknown): “Volumetric fusion is performed by tetrahedral partitioning of space over the obtained dense 3D point cloud, and by exploiting point-wise visibility information in a voting scheme [43, 44].”
Continuous Global Optimization in Multiview 3D Reconstruction (2007, methodology): “A possible remedy is to use an adaptive multi-resolution scheme (Sinha et al., 2007) or a sparse grid (Labatut et al., 2007), but these techniques can not give any guarantee about globality of the computed solution.”
Multi-View Stereo: A Tutorial (2015, methodology): “struct sparse 3D points, then performs 3D Delaunay triangulation of the points to discretize the space [129].”
Free View Synthesis (2020, methodology): “We utilize a Delaunaybased reconstruction [18,23] as it can tolerate a certain amount of outliers.”
Deep blending for free-viewpoint image-based rendering (2018, methodology): “Approaches based on Delaunay tetrahedralization (e.g., [Jancosek and Pajdla 2011; Labatut et al. 2007; RealityCapture 2016]) are able to generate impressive 3D models from photos, even in the presence of traditionally hard cases such as large textureless regions.”
Multi-Label Semantic 3D Reconstruction Using Voxel Blocks (2016, methodology): “It is possible for instance to use a data adaptive discretization of the space in form of a Delaunay triangulation [15], an octree data structure to store the geometry [18] or voxel hashing [21].”
Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction (2022, methodology): “In the past years, traditional methods [13, 15, 28, 34] have adopted a multi-step pipeline to achieve impressive reconstruction results.”
Large-Scale Semantic 3D Reconstruction: An Adaptive Multi-resolution Model for Multi-class Volumetric Labeling (2016, background): “[27] deviate from a regular partitioning of the volume,”
Massively Parallel Multiview Stereopsis by Surface Normal Diffusion (2015, methodology): “We see surface fitting as a subsequent step that is largely independent of the matching – in fact the most popular approaches [21, 25, 28] are rather agnostic about the preceding matcher, and we found the widely used Poisson method [25] to work well for our point clouds.”