Geometry Processing
Geometry processing develops algorithms to analyze, repair, and transform digital shapes, treating meshes much as signal processing treats sampled signals.
Definition
Geometry processing is the set of computational methods for filtering, simplifying, parameterizing, and otherwise transforming discrete surface representations, especially polygon meshes.
Scope
This topic covers mesh smoothing and denoising, simplification and level-of-detail generation, surface parameterization and texture mapping, remeshing and repair, and the discrete differential-geometric operators such as the discrete Laplacian that underpin these methods.
Core questions
- How can noise from a 3D scan be removed without destroying features?
- How is a detailed mesh simplified while preserving its shape?
- How is a surface flattened to the plane for texturing?
- How are continuous differential operators discretized on a mesh?
Key concepts
- Discrete Laplacian operator
- Mesh smoothing and denoising
- Mesh simplification
- Surface parameterization
- Remeshing and repair
- Level of detail
Key theories
- Signal-processing view of meshes
- Surface coordinates can be filtered like signals using the mesh Laplacian, enabling smoothing that suppresses high-frequency noise while a counteracting step prevents the shrinkage that naive smoothing causes.
- Quadric error mesh simplification
- Edge collapses are ordered by a quadric error metric that measures squared distance to the original surface, allowing aggressive simplification that preserves the overall shape and sharp features.
Clinical relevance
Geometry processing is essential for converting raw 3D scans into usable models, generating efficient level-of-detail assets for games, preparing watertight meshes for 3D printing, and analyzing anatomical surfaces in medical imaging.
History
Digital geometry processing emerged in the 1990s as 3D scanning produced large meshes needing cleanup; Taubin's signal-processing framing and Garland and Heckbert's quadric simplification became foundational techniques in a field that grew around the discrete Laplacian.
Key figures
- Gabriel Taubin
- Michael Garland
- Paul Heckbert
Related topics
Seminal works
- taubin1995
- garland1997
Frequently asked questions
- Why do 3D scans need processing before use?
- Raw scans are noisy, often have holes and excessive detail, and may not be watertight, so smoothing, hole-filling, and simplification are needed to make them usable for rendering, simulation, or printing.
- What is surface parameterization for?
- It assigns each surface point a coordinate in a flat domain, which is what makes it possible to wrap a 2D texture image onto a 3D model without excessive distortion.