cdcseacave/openMVS
OpenMVS
openMVS ist eine Multi-View-Stereo-Bibliothek und Photogrammetrie-Pipeline, die für die 3D-Szenenrekonstruktion verwendet wird. Sie transformiert Structure-from-Motion-Daten – insbesondere Kameraposen und Sparse Point Clouds – in detaillierte 3D-Modelle, die aus Dense Point Clouds und texturierten Meshes bestehen.
Das Projekt bietet eine Abfolge von Verarbeitungsschritten, um Punktwolken zu verdichten, 3D-Oberflächen-Meshes zu generieren und fotorealistische Texturen anzuwenden. Es verwendet Multi-View-Textur-Blending, um akkurate Farben auf die rekonstruierte Geometrie abzubilden, und setzt iterative Verfeinerung ein, um Mesh-Details zu optimieren.
Das System beinhaltet Funktionen zum Importieren externer Structure-from-Motion-Szenen, zum Definieren von Interessenbereichen via Bounding-Box-Filterung und zum Visualisieren von 3D-Szenen. Es unterstützt zudem das Extrahieren von Keyframes aus Videos und den Export von Geometrie in Standardformate wie PLY, OBJ und GLB.
Features
- Neural Scene Reconstructions - Creates detailed 3D models of real-world environments from multi-view image datasets.
- 3D Reconstruction Pipelines - Provides a complete pipeline for converting sets of 2D photographs into accurate 3D digital surface meshes.
- Dense Reconstruction Backends - Generates dense 3D point clouds from sparse scenes using Patch-Match or Semi-Global Matching algorithms.
- Multi-View Stereo Generators - Utilizes multiple calibrated views of a scene to generate dense 3D geometry.
- Point-to-Mesh Conversions - Converts dense point clouds into continuous surface meshes with iterative geometry refinement.
- Point Cloud Surface Extraction - Extracts 3D surface meshes that best fit the geometry of an input dense point cloud.
- Point Cloud Generation - Generates high-density 3D point clouds from sparse data using Patch-Match or Semi-Global Matching.
- 3D Mesh Generators - Programmatically creates 3D polygonal meshes from input point clouds and refines them for structural detail.
- 3D Mesh Reconstruction Tools - Provides a complete toolset for converting sparse point clouds into refined 3D surface meshes.
- 3D Mesh Texturing Tools - Implements a process for blending pixel data from multiple camera views to texture 3D geometry.
- Point Cloud Generators - Generates detailed 3D point clouds from raw image frames using depth map estimation.
- Photogrammetric Texture Mapping - Implements automatic mapping of colors from calibrated photographs onto reconstructed 3D meshes.
- Constrained Delaunay Triangulations - Implements constrained Delaunay triangulations to generate continuous 3D surface meshes from dense point clouds.
- Geometric Error Minimization - Optimizes surface geometry using gradient-descent to minimize the distance between the mesh and the dense point cloud.
- Iterative Geometry Refinement - Employs iterative geometry refinement to improve the fidelity of 3D meshes through optimization and smoothing.
- Textured Mesh Optimizations - Computes sharp and accurate color textures to map onto 3D mesh surfaces via joint optimization.
- Photorealistic Mesh Texturing - Maps original image data onto 3D meshes to create realistic colored surfaces.
- Multi-Texture Blending - Implements multi-view texture blending to map accurate colors onto reconstructed geometry using weighted pixel data.
- Image-to-3D Texture Engines - Provides a pipeline to transfer 2D image textures onto reconstructed 3D meshes for photorealistic surfaces.
- Point Cloud Densification - Converts sparse point clouds into complete and accurate dense point clouds for detailed scene representation.
- Semi-Global Matching - Computes depth maps using Semi-Global Matching to optimize global energy functions across image pixels.
- Stereo Depth Estimation - Estimates depth maps using Patch-Match stereo matching to find the best geometric fit across multiple images.
- Structure from Motion Pipelines - Implements a full pipeline for calculating camera poses and sparse point clouds from unordered photos.
- Mesh - Computes and applies accurate textures to color 3D mesh surfaces based on input images.
- 3D Scene Importers - Provides utilities to import camera poses and sparse point clouds from external SfM solvers into its project format.
- 3D Scene Visualization - Provides an interactive viewer for rendering project files, depth maps, and standard 3D geometry.
- Camera Pose Estimators - Calculates camera orientations and positions from unordered images using pinhole or spherical camera models.
- Mesh Detail Reconstruction - Provides capabilities to recover and enhance fine geometric details on reconstructed 3D meshes.
- Geometry Data Exporters - Writes reconstructed point clouds and meshes to standard formats including PLY, OBJ, and GLB.
- Photogrammetry Data Pipelines - Ships a workflow for converting raw image captures into structured poses and point clouds for 3D modeling.
- Video Frame Extraction - Extracts stable and well-spaced keyframes from video or spherical footage for use in 3D reconstruction.
- Pinhole Camera Models - Utilizes pinhole camera models to project 3D world coordinates onto 2D image planes for triangulation.
- Structure From Motion - Library for multi-view stereo reconstruction.
- Multi View Stereo - Library for dense point cloud and mesh generation.
