Massively Parallel Multiview Stereopsis by Surface Normal Diffusion

By Silvano Galliani, Katrin Lasinger, Konrad Schindler. 2015

657 citations recorded by the source corpus.

Subfields: 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 Computer Vision and Pattern Recognition

Citations in the corpus

SurfaceNet: An End-to-End 3D Neural Network for Multiview Stereopsis (2017, methodology): “Figure 1: Reconstruction of the model 13 of the DTU dataset [1] in comparison to [4, 8, 27, 9].”
Multi-View Mesh Reconstruction with Neural Deferred Shading (2022, neutral): “[2, 6, 18, 19, 26, 68, 72, 74].”
Constraining dense hand surface tracking with elasticity (2020, methodology): “After capturing images, we perform PatchMatch-based multi-view stereo [Galliani et al. 2015] to obtain a 3D scan mesh for every frame.”
Learned Multi-patch Similarity (2017, neutral): “alues efﬁ- ciently. The other avoids exhaustive testing and instead relies on efﬁcient random sampling and propagation schemes like PatchMatch [2] to ﬁnd a good depth estimate at every position, e.g. [6]. For both strategies, efﬁcient GPU implementations exist. Also for an overview of contemporary m”
PatchmatchNet: Learned Multi-View Patchmatch Stereo (2020, methodology): “Hence, instead of propagating depth hypotheses naively from a static set of neighbors as done for Gipuma [16] and DeepPruner [11], we want to perform the propagation in an adaptive manner, which gathers hypotheses from the same surface.”
Pixelwise View Selection for Unstructured Multi-View Stereo (2016, methodology): “[15,46]), we generate a small set of additional plane hypotheses at each propagation step.”
PGSR: Planar-Based Gaussian Splatting for Efficient and High-Fidelity Surface Reconstruction (2024, neutral): “Traditional methods adhere to the universal multi-view stereo pipeline, whichcanberoughlycategorizedbasedontheinterme-diate representation they rely on, such as point cloud [31], [32], volume [33], depth map [34], [35], [36], etc.”
Just Look at the Image: Viewpoint-Specific Surface Normal Prediction for Improved Multi-View Reconstruction (2016, methodology): “Among the many available algorithms we choose the fast multi-view PatchMatch implementation of [18].”
Volume Rendering of Neural Implicit Surfaces (2021, background): “Classical multi-view stereo approaches are generally either depth-based [1, 31, 8, 7] or voxel-based [5, 3, 32].”
CNN based Patch Matching for Optical Flow with Thresholded Hinge Loss (2016, methodology): “ (GPU) + 16.5s patch matching (CPU) + 2s for up/downsampling and lowpass (CPU). The CPU parts of our approach likely can be signiﬁcantly sped up using GPU versions like a GPU based propagation scheme [2,13] for patch matching. This is contrary to PatchBatch where the GPU based CNN already takes the ”
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].”
Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching (2019, methodology): “Multi-View Stereo According to the comprehensive survey [5], works in traditional muti-view stereo can be roughly categorised into volumetric methods [13, 14, 18, 33], which estimate the relationship between each voxel and surfaces; point cloud based methods [6, 19], which directly process 3D points”
Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Pass (2025, neutral): “Traditional reconstruction pipelines, such as those based on Structure-from-Motion (SfM) [44] and Multi-View Stereo (MVS) [18], fundamentally rely on image pairs to reconstruct 3D geometry.”
Grounding Image Matching in 3D with MASt3R (2024, neutral): “…RRA@15 RTA@15 mAA(30) mAA(30) (a) Colmap+SG [21,72] 36 Camp [13] 0.835 0.554 0.695 Furu [31] 0.613 0.941 0.777 Tola [90] 0.342 1.190 0.766 Gipuma [32] 0.283 0.873 0.578 MVSNet [110] 0.396 0.527 0.462 CVP-MVSNet [109] 0.296 0.406 0.351 UCS-Net [17] 0.338 0.349 0.344 CER-MVS [55] 0.359 0.305 0.332…”
A Multi-view Stereo Benchmark with High-Resolution Images and Multi-camera Videos (2017, methodology): “PMVS and Gipuma produce very incomplete and noisy results while CMPMVS fails completely.”
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…”
Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction (2022, methodology): “Based on this intuition, we use the photometric consistency constraints in multi-view stereo (MVS) [8, 9, 34] to supervise our extracted implicit surface.”
Deep Stereo Using Adaptive Thin Volume Representation With Uncertainty Awareness (2019, methodology): “To reconstruct the final point cloud, we follow [14] to fuse the depth from multiple views; we use this fusion method for all our exper-”
Atlas: End-to-End 3D Scene Reconstruction from Posed Images (2020, methodology): “default point cloud fusion, while for the other methods we use the implementation of [19].”