- Home
- mesh smoothing
- CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
4.8 (113) · $ 28.99 · In stock
Mesh smoothing Idea: Filter out high frequency noise (common in scanners).
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing
Follow in slide show (F5) mode for the explanatory animations. Asst. Prof. Yusuf Sahillioğlu. Computer Eng. Dept, , Turkey.
Do it in parallel, i.e., use original coordinates although they might have been updated previously.
Mesh smoothing Illustration in 1D:
Mesh smoothing Observation: close curve converges to a single point
Mesh smoothing Illustration in 2D: Same as for curves (1D).
Mesh smoothing Observation: close mesh, e.g., sphere, converges to a single point.
Repeated iterations of Laplacian smoothing shrinks the mesh.
This is introduced by the Mesh Fairing paper by Taubin in
In short, mesh quality improvement. In contrast to mesh repair (next class), the input of remeshing algorithms is usually assumed to be a manifold triangle mesh. Mesh quality: sampling density, regularity, and shape of mesh elements.
Remeshing Mesh elements: triangle vs. quadrangle.
edge directions not messy.
Favoring triangle meshes. Four points or more may not be on the same plane, but three points always are (ignoring collinearity). This has the interesting property that scalar values vary linearly over the surface of the triangle. This, in turn, means most if not all of what s needed to shade, texture map, and depth filter a triangle can be calculated using linear interpolation which can be done extremely fast in specialized hardware. Triangles are the simplest* primitive, so algorithms dealing with triangles can be heavily optimized, e.g., fast point-in-triangle test. * every object can be split into triangles but a triangle cannot be split into anything else than triangles.
Quad to tri conversion Tri to quad conversion
The shape of isotropic elements is locally uniform in all directions. Ideally, a triangle/quadrangle is isotropic if it is close to equilateral/square.
The shape of isotropic elements is locally uniform in all directions. Ideally, a triangle/quadrangle is isotropic if it is close to equilateral/square. Isotropic elements: roundness ~ 1. (favored in numerical apps, FEM). Roundness: ratio of circumcircle radius to the length of the shortest edge.
Smaller elements are assigned to areas w/ high curvature.
Valence close to 6; ~equal edge lengths.
surface-based. Param-based: map to 2D domain, do the remeshing (2d problem), lift up.
Delaunay triangulation: maximize the min angle = no point in circumcircle of a triangle.
surface-based. surface-based: work directly on the mesh embedded in 3D.
Beware of illegal edge collapses: i) Normal flip after collapse! ii) intersection of 1-ring neighborhood of i and j contains 3+ vertices!
Beware of illegal edge collapses: i) A heuristic while removing short edges: Collapse into the vert w/ higher valence. Works ‘cos high-valence verts stay fixed and every collapse reduces # adjacent short edges. Moving a high valence vertex v, however ( case), can lead to an unbounded accumulation of edge collapses since new short edges can become adjacent to v.
Beware of illegal edge splits: i) Infinite-loop problem if you split shorter edges first (top row)!
Beware of illegal edge flips: i) edge is adjacent to 2 tris whose union is not a convex quadrilateral! convex if no projection (of the 4th vert) is inside the tri (defined by the other 3 verts) //4th vert is projected to the plane defined by the other 3.
A sequence of edge collapses, aka mesh decimation:
A sequence of edge collapses, aka mesh decimation: Isosurface extraction by Marching Cubes over-tessellates (left).
Curvature factor is introduced as coplanar surfaces can be represented using fewer polygons than areas w/ a high curvature. Good collapses: Bad collapses:
Curvature factor is introduced as coplanar surfaces can be represented using fewer polygons than areas w/ a high curvature. Collapse cost of edge (u,v): Tu is the set of triangles that contain the vertex u and Tuv is the set of triangles that share the edge (u,v). Cost is length (||u-v||) multiplied by a curvature factor (< 1). Curvature factor computed by comparing the dot products of all involved face normals to find the largest angle b/w 2 faces.
Error quadric: based on the observation that in the original model each vertex is the solution of the intersection of a set of planes. Such a set of planes is associatd w/ each vertex as supporting planes. The error at the vertex w.r.t. this set is the sum of squared distances to its supporting planes. Hence this error helps preserving the original details in the decimated model. Edge-length metric: Error quadric metric:
Error quadric in 1D (supporting planes supporting lines).
Error quadric derivation. Error quadric Kp can be used to find the squared distance of any point in space to its supporting planes.
Error quadric visualization.
Algorithm: Collapse cost of edge (u,v): compute the optimal contraction target vertex for this edge (for simplicity u collapses to v). The error at this proposed new vertex (v) becomes the cost of collapsing this edge. Place all edges in a min-heap keyed on collapse costs. Iteratively remove the edge w/ min cost, collapse it, update the costs of all involved edges. Detail: original algorithm uses ‘vertex pairs’ = edges + 2-close-vertices. Detail: original algorithm collapses u to a point p* that minimizes error Kp*. Surface Simplification Using Quadric Error Metrics.
Normal orientation correction: naïve algo (neighbor triangles have similar normals) vs. simple algo ( Linear mesh interpolation vs. nonlinear mesh interpolation. Mesh decimation with curvature metric (slide 31) vs. error quadric metric (32) Mesh segmentation using paper: Consistent mesh partitioning and skeletonisation using the shape diameter function. Mesh skeleton extraction using the same paper. Hole filling using Deformation model in Microsoft KinEtre.
Lecture 11: Digital Geometry Processing (CMU 15-462/662)
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
CENG 789 – Digital Geometry Processing - online presentation
PPT - Digital Image Processing Chapter 4: Image Enhancement in the Frequency Domain 22 June 2005 PowerPoint Presentation - ID:9266381
A comprehensive survey on 3D face recognition methods - ScienceDirect
Finite Element Analysis-Applications in Mechanical Engineering, PDF, Numerical Analysis
10.1007/978 3 319 06136 8, PDF, Viscoelasticity
Addgene: pMD2.G Citations
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
A Nitsche-eXtended Finite Element Method for Distributed Optimal Control Problems of Elliptic Interface Equations
Digital Image Processing: Image Enhancement in the Spatial Domain
CENG 789 – Digital Geometry Processing 05- Smoothing and Remeshing - ppt video online download
CENG589 - Digital Geometry Processing
PDF) BRAIN, DECISION MAKING AND MENTAL HEALTH SPRINGER, VOL 12
