Scribd
Upload
19 views

Lecture 19

1. Point clouds, triangle soups, and manifold/watertight meshes are common representations of surface data, with meshes providing more structure. 2. Triangle and quad meshes are common structures, with half-edge data structures enabling efficient neighbor queries. 3. Volume data can be represented through voxelization or tetrahedral meshes, though generating quality tetrahedral meshes from surfaces is challenging. 4. Constructive solid geometry uses boolean operations on simple primitives to model complex shapes with an exact representation.

Uploaded by

lelsyngami
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
19 views

Lecture 19

1. Point clouds, triangle soups, and manifold/watertight meshes are common representations of surface data, with meshes providing more structure. 2. Triangle and quad meshes are common structures, with half-edge data structures enabling efficient neighbor queries. 3. Volume data can be represented through voxelization or tetrahedral meshes, though generating quality tetrahedral meshes from surfaces is challenging. 4. Constructive solid geometry uses boolean operations on simple primitives to model complex shapes with an exact representation.

Uploaded by

lelsyngami
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 65

Mesh Data Structures and CSG

Representing Surface Data

point cloud

more structure
triangle soup

manifold mesh

watertight mesh
Point Clouds
Point Clouds
List of points
• may or may not include normals
Point Clouds
List of points
• may or may not include normals
• normals estimated by plane-fitting
Point Clouds
List of points
• may or may not include normals
• normals estimated by plane-fitting

Raw data from depth sensors


Triangle Soup
List of triangles
• each triangle has own verts
Triangle Soup
List of triangles
• each triangle has own verts

No notion of triangle neighbors


• (but can find nearby triangles)
Triangle Soup
List of triangles
• each triangle has own verts

No notion of triangle neighbors


• (but can find nearby triangles)

Why do we need neighbors anyway?


Manifold Mesh
Must satisfy three properties:
1. Every edge shared by one/two faces
Manifold Mesh
Must satisfy three properties:
1. Every edge shared by one/two faces

not manifold: has T junction


Manifold Mesh
Must satisfy three properties:
1. Every edge shared by one/two faces
2. Faces around verts are triangle fans

not manifold
Manifold Mesh
Must satisfy three properties:
1. Every edge shared by one/two faces
2. Faces around verts are triangle fans

Intuitively: mesh locally “looks like a


single surface”
Manifold Mesh
Must satisfy three properties:
1. Every edge shared by one/two faces
2. Faces around verts are triangle fans
3. Faces have consistent orientation

not
consistent consistent
Watertight Mesh
Manifold mesh that
• is single piece
• has no boundary
Watertight Mesh
Manifold mesh that
• is single piece
• has no boundary

Splits space into well-defined inside/out


• can be “filled with water” without leaks
Representing Surface Data

point cloud
Graphics Grand Challenge

more structure
real-word data often
triangle soup unstructured

graphics algorithms usually need


structured data
manifold mesh

watertight mesh
Representing Surface Data

point cloud

triangle soup
“mesh repair”
algorithms
manifold mesh

watertight mesh
Representing Surface Data

point cloud

triangle soup
“mesh repair”
algorithms
manifold mesh (rarely works)

watertight mesh
Representing Surface Data

point cloud

triangle soup
Poisson surface
reconstruction
manifold mesh

watertight mesh
Poisson Surface Reconstruction
Interpolates point cloud and normals
Triangle Mesh Data Structures
List of points & triangle indices
Triangle Mesh Data Structures
List of points & triangle indices
Pros:
• lightweight, compact
• native GPU data structure
• very common file data structure
Cons:
Triangle Mesh Data Structures
List of points & triangle indices
Pros:
• lightweight, compact
• native GPU data structure
• very common file data structure
Cons:
• neighbor queries slow
• finding boundaries slow
Half-edge Data Structure
Store mesh as set of half-edges
Half-edge Data Structure
Half-edge Data Structure
Store mesh as set of half-edges
Pros:
• easy to “walk around” faces, vertices
• all kinds of neighbor queries easy
Half-edge Data Structure
Store mesh as set of half-edges
Pros:
• easy to “walk around” faces, vertices
• all kinds of neighbor queries easy
Cons:
• large memory footprint
• tricky to implement (tons of pointers!)
Half-edge Data Structure
Store mesh as set of half-edges
Pros:
• easy to “walk around” faces, vertices
• all kinds of neighbor queries easy
Cons:
• large memory footprint
• tricky to implement (tons of pointers!)
• use existing libraries (e.g. OpenMesh)
Types of Manifold Meshes
triangle
quad

quad-dominant exotic (hexagonal, etc)


Triangles vs Quads

Triangles simpler
Quads more natural for flat & cylindrical
geometry
Arbitrary Quads Are Not Planar!

Arbitrarily triangulate to render them…


Subdivision
Input: coarse control mesh

Output: finer mesh with smoother details


Linear Subdivision
Split faces 1:4

Insert verts at
edge midpts

Adds faces,
but doesn’t
change shape
Nonlinear Subdivision
Everybody has pet subdivision method
Most popular:
• triangle meshes: Loop subdivision
Nonlinear Subdivision
Everybody has pet subdivision code
Most popular:
• triangle meshes: Loop subdivision
• quad meshes: Catmull-Clark
Catmull-Clark Subdivision
Rules for adding new points and replacing
old points

Works best for regular quad meshes


Regular vs Irregular Vertices
Regular vertices have four edges

Irregular vertices have 3 or 5+ edges


• also called extraordinary vertices
Catmull-Clark Subdivision
Rules for adding new points and replacing
old points

Many schemes for handling irregular verts


Subdivision: Complications
Dealing with irregular vertices
Dealing with creases
• some edges shouldn’t be smoothed…
Subdivision: Complications
Dealing with irregular vertices
Dealing with creases
• some edges shouldn’t be smoothed…
Dealing with boundaries

In general, allowing finer-grained control


of subdivision process
Subdivision Surface
Smooth surface at limit of subdivision

Fundamental building block of Pixar’s


Renderman engine
Other Mesh Operations
Decimation Quadrangulation

Remeshing
Smoothing
Other Mesh Operations
Graphics subfield: geometry processing
• uses sophisticated theory from linear
algebra, differential geometry, etc.

In practice: several good packages


• CGAL – general purpose
• OpenMesh – halfedge, subdivision
Volume Data
Who cares about volumes?
• just render outer skin?
Volume Data
Who cares about volumes?
• just render outer skin?

Translucent object (colored glass, fog,…)


Volume Data
Who cares about volumes?
• just render outer skin?

Translucent object (colored glass, fog,…)


Physical simulations & deformations
• fracture
Volume Data
Who cares about volumes?
• just render outer skin?

Translucent object (colored glass, fog,…)


Physical simulations & deformations
• fracture
Modeling primitive for cutting & sculpting
Representing Volumes
Voxelization: 3D rasterization
Representing Volumes
Voxelization: 3D rasterization

really easy… but boundary chunky


Representing Volumes
Tetrahedral (tet) mesh: 3D analogue of
triangle mesh
• each tet “hyperface”
has 4 faces, 6 edges,
4 verts
Representing Volumes
Tetrahedral (tet) mesh: 3D analogue of
triangle mesh
• each tet “hyperface”
has 4 faces, 6 edges,
4 verts
• bdry is watertight
triangle mesh
Generating 2D Mesh From Curve
Can be done by triangulation
Generating 2D Mesh From Curve
Can be done by triangulation

Greedy “ear-cutting” always works


Generating 3D Mesh From Surface
“Tetrahedralization” not always possible!
Shoenhardt Polytope
Generating 3D Mesh From Surface
“Tetrahedralization” not always possible!
Shoenhardt Polytope
Must add extra
inner points
• “Steiner pts”
Generating 3D Mesh From Surface
“Tetrahedralization” not always possible!
Shoenhardt Polytope
Must add extra
inner points
• “Steiner pts”

Algorithm is complex and bug-prone


• popular library: TetGen
Representing Volumes
Hexahedral mesh: 3D version of quad
mesh
• “hexahedron” = 6 sides (cube)
Representing Volumes
Hexahedral mesh: 3D version of quad
mesh
• “hexahedron” = 6 sides (cube)

No general hexahedralization
algorithm exists (!)
Constructive Solid Geometry
Start with simple buildings blocks
• sphere, cubes, cylinders, …

Build complicated
objects using tree
of operations
CSG Operations

union
CSG Operations

union intersection
CSG Operations

union intersection set difference


CSG Tree
Pros:
• geometry is exact
• simple representation for
visually-complex shapes
CSG Tree
Pros:
• geometry is exact
• simple representation for
visually-complex shapes
Cons:
• some shapes difficult
to model
Used in many game engines (e.g. Unreal)

You might also like