Multi-View Stereo: A Tutorial

By Yasutaka Furukawa, Carlos Hernández. 2015

838 citations recorded by the source corpus.

How researchers see this paper

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

In Ocean Engineering

Citations in the corpus

SurfaceNet: An End-to-End 3D Neural Network for Multiview Stereopsis (2017, background): “A more comprehensive overview of MVS approaches is given in the tutorial article [7].”
View Synthesis by Appearance Flow (2016, background): “In cases when more than one input view is available, modern multi-view stereo algorithms (see Furukawa and Hernandez [19] for an excellent tutorial) have demonstrated results of impressive visual quality.”
Space-time Neural Irradiance Fields for Free-Viewpoint Video (2020, background): “Existing multi-view stereo (MVS) methods (either geometric-based [60, 18] or learning-based [24, 72, 20, 37]) assume static scene and thus not suitable for dynamic videos as it often produces erroneous depth for moving objects or untextured regions.”
GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose (2018, background): “Traditional Scene Geometry Understanding Structurefrom-Motion (SfM) is a long standing problem which infers scene structure and camera motion jointly from potentially very large unordered image collections [13, 16].”
Immersive light field video with a layered mesh representation (2020, methodology): “One strategy is to use multi-view stereo [Furukawa and Hernandez 2015] and/or depth sensors to estimate the 3D depth of the scene in each direction, so that the panoramic imagery can be reprojected with image-based rendering to a novel viewpoint.”
Unsupervised Monocular Depth Estimation with Left-Right Consistency (2016, background): “There is a large body of work that focuses on depth estimation from images, either using pairs [46], several overlapping images captured from different viewpoints [14], temporal sequences [44], or assuming a fixed camera, static scene, and changing lighting [52, 2].”
DeepMVS: Learning Multi-view Stereopsis (2018, methodology): “1 MVS is one of the fundamental computer vision problems that have seen decades of research and it is a core component in numerous important applications, including 3D reconstruction, novel view synthesis, augmented reality, and medical imaging [9].”
BakedSDF: Meshing Neural SDFs for Real-Time View Synthesis (2023, methodology): “Most classical approaches for view synthesis use triangle meshes (typically reconstructed using a pipeline consisting of multi-view stereo [Furukawa and Hernández 2015; Schönberger et al. 2016], Poisson surface reconstruction [Kazhdan et al. 2006; Kazhdan and Hoppe 2013], and marching cubes…”
Transformation-Grounded Image Generation Network for Novel 3D View Synthesis (2017, methodology): “When multiple images are available, multi-view stereo algorithms [12] are applicable to explicitly reconstruct the 3D scene which can then be utilized to synthesize novel views.”
Tanks and temples (2017, methodology): “…visual computing, combining rigorous theory [Hartley and Zisserman 2000; Triggs et al. 2000], advanced computational methods [Agarwal et al. 2010; Furukawa and Hernández 2015; Wu et al. 2011], and a culture of open software development [Fuhrmann et al. 2015; Furukawa 2011; Moulon et al. 2016;…”
DUSt3R: Geometric 3D Vision Made Easy (2023, neutral): “The fully handcrafted [32, 34, 107, 147, 175], the more recent scene optimization based [31, 71, 76, 77, 143, 146, 148, 163], or learning based [40, 53, 65, 86, 161, 164, 180] approaches all depend on camera parameter estimates obtained via complex calibration procedures, either during the data…”
DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Cameras (2021, neutral): “Evaluating dense 3D reconstruction is typically considered in the domain of Multiview Stereo [19] and outside the scope of this work.”
Neural volumes (2019, background): “Point-and surface-based reconstruction techniques have a long history in computer vision (see [Furukawa and Hernández 2015] for a review).”
Towards Probabilistic Volumetric Reconstruction Using Ray Potentials (2015, background): “Please refer to [29, 8] for a more complete overview of multi-view stereo methods.”
VGGT: Visual Geometry Grounded Transformer (2025, neutral): “…from multiple overlapping images, typically assuming known camera parameters, which are often estimated with SfM. MVS methods can be divided into three categories: traditional handcrafted [26, 27, 64, 88], global optimization [25, 50, 91, 100], and learning-based meth-ods [30, 49, 55, 99, 106].”
Patches, Planes and Probabilities: A Non-Local Prior for Volumetric 3D Reconstruction (2016, background): “For a more complete review, we refer the reader to [15, 44].”
Digging Into Self-Supervised Monocular Depth Estimation (2018, methodology): “…by enforcing smoothness in the depth maps, and by computing photo-consistency on patches when solving for per-pixel depth via global optimization [12]. c d By constraining the network to perform image synthesis using an intermediary variable, in our case depth or disparity, we can then extract…”