Mesh optimization

Hugues HoppeT. DeRoseT. DuchampJ. McDonaldW. Stuetzle

article1993SIGGRAPH1,615 citations

Develops an energy-minimization framework that jointly optimizes mesh connectivity and vertex positions to accurately fit 3D point sets while reducing geometric complexity, recovering sharp features for both surface reconstruction and mesh simplification.

Listen

Generating accurate, low-complexity 3D geometric models from physical scans and simplifying dense polygonal meshes are critical challenges across computer graphics, design, and engineering. Traditional methods either restrict models to simple shapes, alter geometry cumulatively during simplification, or struggle to preserve sharp features such as corners and edges.

The article demonstrates an optimization framework that fits a triangular mesh to 3D point data by balancing geometric accuracy against model conciseness. The primary objective is to simultaneously optimize the number of vertices, their positions, and their connectivity while preserving the underlying topological structure.

The authors develop an energy minimization approach that combines three terms: distance error to the data points, a penalty for the total number of vertices, and a regularizing spring energy to prevent geometric instability. The optimization alternates between continuous refinement of vertex positions using conjugate gradients and discrete modifications of connectivity using three local topological operations: edge collapses, splits, and swaps. The evaluation tested the method on synthetic shapes and real-world laser scans ranging from thousands to over 16,000 data points.

The evaluation produced four key findings. First, the method achieved dramatic mesh reductions—often decreasing vertex counts by roughly 75% to over 90%—while maintaining high fidelity to original point sets. Second, the energy formulation naturally adapts mesh structure to local curvature, concentrating vertices in curved regions and elongating triangles across flatter areas. Third, the process reliably recovers sharp edges and corners, enabling automated segmentation of complex shapes into distinct smooth components via dihedral angle thresholds. Finally, incorporating a scheduled, decreasing spring regularization successfully prevented surface spikes and guided solutions to clean local minima.

These results show that directly minimizing global deviation from original data creates far superior simplified meshes compared to incremental decimation techniques. For organizations working with 3D data, this method significantly reduces computational and storage costs while maintaining the geometric integrity required for rendering and engineering analyses.

Moving forward, adopting teams should utilize the user-tunable representation parameter to manage the explicit trade-off between model compactness and geometric error according to project needs. Further technical development should focus on accelerating execution times through parallel computing, exploring alternative error metrics, and automating scanning paths to resolve data collection gaps such as self-shadowing.

The reported results carry high confidence across varied topologies, though findings remain bounded by the heuristic nature of local descent methods, which do not guarantee a global minimum. Practitioners should also note that execution times currently range up to 47 minutes per model on uniprocessor hardware, and handling severe data noise or scanning occlusions may still require manual preprocessing.

University of Washington data points optimization optimization problems mesh optimization mesh simplification surface reconstruction

hhoppe/Mesh-processing-library Hoppe et al (1993).pdf

Paper: Surface reconstruction from unorganized points, Hugues Hoppe et al. (1992). Reading this paper introduces the foundational pipeline for extracting initial simplicial surfaces from unorganized point clouds, which provides the starting topologies that mesh optimization refines.
Paper: Decimation of triangle meshes, William J. Schroeder et al. (1992). This work establishes standard local decimation criteria and vertex removal heuristics, establishing the baseline simplification paradigms that global energy minimization aims to surpass.
Paper: Free-form deformation of solid geometric models, Thomas W. Sederberg et al. (1986). It provides foundational principles for solid geometric deformations and control-point lattices that underpin smooth geometric fitting frameworks.

Paper: Progressive meshes, Hugues Hoppe (1996). This work directly builds on the energy-minimization mesh optimization framework by adapting its edge-collapse operations into an invertible, continuous multi-resolution hierarchy.
Paper: Surface simplification using quadric error metrics, Michael Garland et al. (1997). It extends mesh simplification methodology by introducing quadric error metrics to drastically accelerate edge-contraction decisions while maintaining geometric fidelity.
Paper: A volumetric method for building complex models from range images, Brian Curless et al. (1996). This paper advances 3D surface generation by formulating a volumetric, signed-distance method for integrating multiple range images into seamless meshes without relying solely on explicit local mesh optimization.
Paper: Zippered polygon meshes from range images, Greg Turk et al. (1994). It provides an incremental multi-scan alignment and zippering pipeline for reconstructing complete surfaces from overlapping laser range images.
Paper: Implicit fairing of irregular meshes using diffusion and curvature flow, Mathieu Desbrun et al. (1999). It develops stable implicit integration and curvature-flow fairing techniques that offer a faster alternative to traditional explicit spring regularization on irregular meshes.
Paper: A signal processing approach to fair surface design, Gabriel Taubin (1995). It introduces a non-shrinking signal-processing filter for fairing polyhedral surfaces, addressing the shrinkage and computational overhead of iterative energy-based regularizations.
Paper: Reconstruction and representation of 3D objects with radial basis functions, J. C. Carr et al. (2001). It extends point-cloud surface fitting by employing polyharmonic radial basis functions and fast multipole evaluations for implicit reconstruction of large-scale noisy data.
Paper: Three-dimensional alpha shapes, Herbert Edelsbrunner et al. (1994). It establishes a formal combinatorial framework for multi-scale geometric shape reconstruction directly from point clouds via alpha complexes.

Abstract

We present a method for solving the following problem: Given a set of data points scattered in three dimensions and an initial triangular mesh M0, produce a mesh M, of the same topological type as M0, that fits the data well and has a small number of vertices. Our approach is to minimize an energy function that explicitly models the competing desires of conciseness of representation and fidelity to the data. We show that mesh optimization can be effectively used in at least two applications: surface reconstruction from unorganized points, and mesh simplification (the reduction of the number of vertices in an initially dense mesh of triangles).

Topics

data points optimization optimization problems reconstruction topological structure mesh topologies

Keywords

mesh optimization mesh simplification surface reconstruction geometric modeling triangular mesh surface fitting edge collapse vertex removal simulated annealing discrete optimization

1 Introduction
2 Mesh Representation
3 Definition of the Energy Function
4 Minimization of the Energy Function
4.1 Optimization for Fixed Simplicial Complex
4.1.1 Projection Subproblem
4.1.2 Linear Least Squares Subproblem
4.3 Exploiting Locality
4.3.1 Heuristics for Evaluating the Effect of Legal Moves
4.3.2 Legal Move Selection Strategy
4.4 Setting of the Spring Constant
5 Results
5.1 Surface Reconstruction
5.2 Mesh Simplification
5.3 Segmentation
5.4 Parameter Settings and Performance Statistics
6 Related Work
7 Summary and Future Work
References

Knowls

Knowl 1 — Mesh Optimization Energy Function
model/method
The mesh optimization problem seeks a triangular mesh $M = (K, V)$ of a given topological type that balances geometric fidelity to a set of 3D data points $X = \{x_1, \dots, x_n\} \subset \mathbb{R}^3$ with conciseness of representation. The mesh is parameterized by a simplicial complex $K$ (specifying vertex, edge, and face connectivity) and a set of vertex positions $V = \{v_1, \dots, v_m\} \subset \mathbb{R}^3$ . The objective is to minimize the energy function:

$E(K, V) = E_{\text{dist}}(K, V) + E_{\text{rep}}(K) + E_{\text{spring}}(K, V)$

where:
- $E_{\text{dist}}(K, V) = \sum_{i=1}^n d^2(x_i, \phi_V(|K|))$ is the distance energy, measuring the sum of squared Euclidean distances from each data point $x_i$ to the continuous geometric realization $\phi_V(|K|) \subset \mathbb{R}^3$ of the mesh.
- $E_{\text{rep}}(K) = c_{\text{rep}} m$ is the representation energy, penalizing mesh complexity linearly with the number of vertices $m = |V|$ , controlled by a user-specified weight $c_{\text{rep}} > 0$ .
- $E_{\text{spring}}(K, V) = \sum_{\{j, k\} \in K} \kappa \|v_j - v_k\|^2$ is the spring regularizing energy, placing a spring of rest length zero and spring constant $\kappa > 0$ on each edge of the mesh.
The spring energy guarantees the mathematical existence of an energy minimum and prevents degenerate spikes in regions lacking sample points. Unlike a curvature penalty, $E_{\text{spring}}$ does not penalize sharp dihedral angles between adjacent faces, thereby permitting the recovery of sharp edges and corners.
Knowl 2 — Nested Minimization Algorithm for Mesh Optimization
algorithm
Minimization of the mesh energy $E(K, V)$ is partitioned into two nested subproblems: an outer discrete optimization over simplicial complexes $K$ homeomorphic to an initial complex $K_0$ , and an inner continuous optimization over vertex positions $V$ for a fixed simplicial complex $K$ .
function OptimizeMesh(K_0, V_0, X, c_rep, kappa_schedule)
    Input: Initial complex K_0, initial vertex positions V_0, data points X, representation cost c_rep, decreasing schedule of spring constants kappa_schedule
    Output: Optimized mesh (K, V)
    K := K_0
    V := OptimizeVertexPositions(K_0, V_0, X, kappa_schedule[0])
    for kappa in kappa_schedule do
        CandidateSet := all edges in K
        while CandidateSet is not empty do
            Select and remove an edge e = {i, j} from CandidateSet
            for move_type in [edge_collapse, edge_swap, edge_split] do
                if IsLegalMove(e, move_type, K) then
                    (K_new, V_new, delta_E) := EvaluateLocalMove(e, move_type, K, V, X, c_rep, kappa)
                    if delta_E < 0 then
                        K := K_new
                        V := V_new
                        CandidateSet := CandidateSet union {all edges incident to the modified neighborhood in K}
                        break
                    end if
                end if
            end for
        end while
    end for
    return (K, V)
end function
The algorithm performs outer discrete transitions using elementary local topological operations (edge collapse, edge swap, edge split) selected via random descent from an active candidate set. For each proposed legal topological change, vertex positions in the local neighborhood are optimized, and the move is accepted if and only if it reduces total energy.
Knowl 3 — Continuous Vertex Optimization via Alternating Projections and Linear Least Squares
algorithm
For a fixed simplicial complex $K$ , minimizing $E(K, V)$ reduces to minimizing $E_{\text{dist}}(K, V) + E_{\text{spring}}(K, V)$ over $V$ . By explicitly introducing barycentric coordinates $B = \{b_1, \dots, b_n\}$ (where each $b_i \in |K| \subset \mathbb{R}^m$ parameterizes the projection of data point $x_i$ onto the mesh), the objective is redefined as:

$E(K, V, B) = \sum_{i=1}^n \|x_i - \phi_V(b_i)\|^2 + \sum_{\{j,k\} \in K} \kappa \|v_j - v_k\|^2$

Optimization proceeds by alternating two subproblems until convergence:
1. Projection Subproblem (Optimize $B$ given $V$ ): For each point $x_i$ , compute $b_i = \arg\min_{b \in |K|} \|x_i - \phi_V(b)\|^2$ . This is accelerated to $O(n)$ expected time using a spatial partitioning grid over mesh faces, together with temporal coherence (projecting $x_i$ only onto faces sharing a vertex with the face containing the projection from the prior iteration).
2. Linear Least Squares Subproblem (Optimize $V$ given $B$ ): With $B$ fixed, the problem uncouples into three independent linear least squares problems, one for each spatial coordinate $c \in \{1, 2, 3\}$ :
$\min_{v^c} \|A v^c - d^c\|^2$

where $v^c \in \mathbb{R}^m$ is the coordinate vector of the vertices, $d^c \in \mathbb{R}^{n+e}$ contains the coordinate values of the $n$ data points followed by $e$ zeros (where $e$ is the number of edges), and $A$ is an $(n+e) \times m$ sparse design matrix. The first $n$ rows of $A$ contain the barycentric coordinates $b_i$ (at most 3 non-zero entries per row), and each of the trailing $e$ rows contains $\sqrt{\kappa}$ and $-\sqrt{\kappa}$ at the indices of an edge's endpoints. The sparse system contains $O(n+m)$ non-zero entries and is solved using the conjugate gradient method, requiring $O(n+m)$ time per iteration and typically reaching acceptable precision within 200 iterations even for $m \approx 10^4$ .
Knowl 4 — Topology-Preserving Legal Move Conditions for Simplicial Complex Modifications
theoretical result
To ensure that the discrete outer optimization explores only meshes homeomorphic to the starting simplicial complex $K_0$ , elementary modifications on a simplicial complex $K$ are restricted to legal moves:
1. Edge Split: Subdividing an edge $\{i, j\} \in K$ by inserting a new vertex is unconditionally a legal move because it never alters the topological type of $K$ .
2. Edge Collapse: Collapsing edge $\{i, j\} \in K$ into a single vertex $\{h\}$ is a legal move if and only if all three of the following conditions hold:
  - For every vertex $\{k\}$ adjacent to both $\{i\}$ and $\{j\}$ (i.e., $\{i, k\} \in K$ and $\{j, k\} \in K$ ), $\{i, j, k\}$ is a 2-simplex (face) of $K$ .
  - If both $\{i\}$ and $\{j\}$ are boundary vertices (vertices incident on an edge that belongs to only one face), then $\{i, j\}$ must itself be a boundary edge.
  - If neither $\{i\}$ nor $\{j\}$ is a boundary vertex, $K$ must contain strictly more than 4 vertices; if either $\{i\}$ or $\{j\}$ is a boundary vertex, $K$ must contain strictly more than 3 vertices.
3. Edge Swap: Swapping edge $\{i, j\} \in K$ shared by faces $\{i, j, k\}$ and $\{i, j, l\}$ for the opposing edge $\{k, l\}$ is a legal move if and only if $\{k, l\} \notin K$ .
Knowl 5 — Local Energy Evaluation and Candidate Set Heuristics for Mesh Transformations
model/method
Because solving a global continuous optimization after every proposed discrete legal move is computationally intractable, the outer optimization employs local evaluation heuristics and an active candidate set:
- Submesh Localization: For a proposed move on an edge $\{i, j\}$ , the change in energy $\Delta E$ is estimated pessimistically by extracting the local submesh and only the data points currently projecting onto that submesh. For an edge collapse into vertex $\{h\}$ , the submesh is $\text{star}(\{i\}; K) \cup \text{star}(\{j\}; K)$ ; for an edge split, it is $\text{star}(\{i, j\}; K)$ ; for an edge swap producing $\{k, l\}$ , the better of optimizing $v_k$ on $\text{star}(\{k\}; K')$ or $v_l$ on $\text{star}(\{l\}; K')$ is evaluated.
- Initialization Trials: For edge collapse, three local optimization starting positions for the replacement vertex $v_h$ are evaluated: $v_i$ , $v_j$ , and $\frac{1}{2}(v_i + v_j)$ , keeping the best outcome.
- Dihedral Angle Thresholding: To prevent self-intersecting embeddings without expensive global collision checks, a candidate move is rejected if the maximum dihedral angle across edges in the locally modified star exceeds a predefined threshold.
- Active Candidate Set Strategy: All edges are initially placed in a candidate set. In each iteration, an edge is removed and evaluated in fixed precedence: edge collapse, then edge swap, then edge split. If any transformation successfully decreases total energy, it is applied, and all neighboring edges in the modified submesh are added back to the candidate set.
Knowl 6 — Regularization Annealing Schedule and Coordinate Normalization
model/method
The regularizing spring constant $\kappa$ in the spring energy $E_{\text{spring}}(K, V) = \sum_{\{j,k\} \in K} \kappa \|v_j - v_k\|^2$ is adjusted according to a scheduled continuation method (annealing schedule):
- The optimization algorithm is executed in successive stages with monotonically decreasing values of $\kappa$ : $\kappa \in \{10^{-2}, 10^{-3}, 10^{-4}, 10^{-8}\}$ .
- At large values of $\kappa$ ( $10^{-2}$ ), the optimization establishes global geometric convergence and removes redundant vertices while preventing self-intersections or geometric folds.
- At small values of $\kappa$ ( $10^{-8}$ ), the regularizing forces vanish, enabling the mesh to closely conform to the data points and recover sharp edges and corners without artificial flattening.
To ensure scale invariance under Euclidean transformations and uniform scaling, data points $X$ and initial mesh vertices $V_0$ are pre-scaled uniformly into a unit cube $[0, 1]^3$ prior to optimization, and the inverse transformation is applied to the final optimized mesh.
Knowl 7 — Mesh Simplification by Surface Point Sampling and Energy Minimization
model/method
Dense triangular meshes are simplified while strictly bounding geometric error by reformulating simplification as mesh optimization against a sampled point set:
1. Point Sampling Protocol: Points $X$ are extracted from an initial high-resolution mesh $M_0$ through three complementary mechanisms: uniform random sampling over the surface area of all faces, adding all original vertex positions of $M_0$ , and sampling additional points along boundary edges to preserve boundary curves.
2. Optimization Execution: The initial mesh $M_0$ and sample points $X$ are passed as input to the mesh optimization algorithm under a chosen representation cost $c_{\text{rep}}$ .
Because the objective function minimizes Euclidean distance deviation from the original surface sample points:
- Vertex density naturally concentrates in regions of high Gaussian curvature and becomes sparse across planar regions.
- Triangle edges automatically elongate along principal directions of minimum curvature.
- Mesh edges and vertices align precisely along sharp creases and corners present in the original geometry.
Knowl 8 — Mesh Feature Recovery and Dual-Graph Surface Segmentation
model/method
Because the energy formulation does not penalize sharp dihedral angles, the optimized mesh naturally aligns edges with surface creases. This property is exploited to automatically segment the mesh into smooth surface patches:
1. Construct a dual adjacency graph where each node represents a triangular face of the optimized mesh.
2. Connect two nodes with an edge if and only if their corresponding faces share a mesh edge and their dihedral angle is smaller than a chosen threshold angle.
3. Compute the connected components of the dual graph. Each connected component corresponds to a distinct, smoothly curved surface region bounded by sharp features.
4. For smooth surface rendering, vertex normals are estimated by averaging the normals of incident faces belonging exclusively to the same connected component, preserving sharp creases across segment boundaries while providing smooth Gouraud/Phong shading within patches.

Knowl 9 — Quantitative Reconstruction and Simplification Performance Under Varied Representation Costs

data/table

The mesh optimization algorithm was evaluated across surface reconstruction tasks (from synthetic sampled points and laser range scanner data) and mesh simplification tasks on a DEC Alpha workstation.

Model	$m$ (vertices)	Faces	$n$ (points)	$c_{\text{rep}}$	$\kappa$	$E_{\text{dist}}$	$E$	Time (min)
Sampled object (Phase 1 $M_0$ )	1572	3152	4102	-	-	$8.57 \times 10^{-2}$	-	-
Sampled object (fixed $K_0$ )	1572	3152	4102	$10^{-5}$	$10^{-2}$	$8.04 \times 10^{-4}$	$4.84 \times 10^{-2}$	1.5
Sampled object (opt stage 1)	508	1024	4102	$10^{-5}$	$10^{-2}$	$6.84 \times 10^{-4}$	$3.62 \times 10^{-2}$	(+3.0)
Sampled object (opt stage 2)	270	548	4102	$10^{-5}$	$10^{-3}$	$6.08 \times 10^{-4}$	$6.94 \times 10^{-3}$	(+2.2)
Sampled object (final)	163	334	4102	$10^{-5}$	varied	$4.86 \times 10^{-4}$	$2.12 \times 10^{-3}$	17.0
Distributor cap (Phase 1 $M_0$ )	9220	18272	12745	-	-	$6.41 \times 10^{-2}$	-	-
Distributor cap (final)	690	1348	12745	$10^{-5}$	varied	$4.23 \times 10^{-3}$	$1.18 \times 10^{-2}$	47.0
Golf club head (Phase 1 $M_0$ )	4059	8073	16864	-	-	$2.20 \times 10^{-2}$	-	-
Golf club head (final)	262	515	16864	$10^{-5}$	varied	$2.19 \times 10^{-3}$	$4.95 \times 10^{-3}$	44.5
Minimal surface ( $M_0$ )	2032	3832	-	-	-	-	-	-
Minimal surface (simplified 1)	487	916	6752	$10^{-5}$	varied	$1.86 \times 10^{-3}$	$8.05 \times 10^{-3}$	9.9
Minimal surface (simplified 2)	239	432	6752	$10^{-4}$	varied	$9.19 \times 10^{-3}$	$4.39 \times 10^{-2}$	10.2

In all cases, "varied" indicates the standard schedule $\kappa \in \{10^{-2}, 10^{-3}, 10^{-4}, 10^{-8}\}$ . The results show an order-of-magnitude reduction in vertex count (e.g., from 1572 to 163 vertices for the sampled object; 9220 to 690 for the distributor cap; 4059 to 262 for the golf club head) while reducing distance energy $E_{\text{dist}}$ by up to two orders of magnitude relative to the initial topological reconstruction $M_0$ . Increasing the complexity penalty $c_{\text{rep}}$ from $10^{-5}$ to $10^{-4}$ on the minimal surface model produces a more compact mesh (239 vs. 487 vertices) at the cost of higher geometric error.

Coverage note — No substantial contributed material was omitted. The knowls fully capture the energy formulation, continuous and discrete optimization algorithms, topological legality criteria, locality and annealing heuristics, mesh simplification, surface segmentation, and empirical data.

References

1.Ruud M. Bolle and Baba C. Vemuri. On three-dimensional surface reconstruction methods. IEEE PAMI, 13(1):1–13, January 1991.
2.T. DeRose, H. Hoppe, T. Duchamp, J. McDonald, and W. Stuetzle. Fitting of surfaces to scattered data. SPIE, 1830:212–220, 1992.
3.Gene Golub and Charles Van Loan. Matrix Computations. John Hopkins University Press, 2nd edition, 1989.
4.Ardeshir Goshtasby. Surface reconstruction from scattered measurements. SPIE, 1830:247–256, 1992.
5.H. Hoppe, T. DeRose, T. Duchamp, J. McDonald, and W. Stuetzle. Surface reconstruction from unorganized points. Computer Graphics (SIGGRAPH ’92 Proceedings), 26(2):71–78, July 1992.
6.H. Hoppe, T. DeRose, T. Duchamp, J. McDonald, and W. Stuetzle. Mesh optimization. TR 93-01-01, Dept. of Computer Science and Engineering, University of Washington, January 1993.
7.J.L. Mallet. Discrete smooth interpolation in geometric modeling. CAD, 24(4):178–191, April 1992.
8.Samuel Marin and Philip Smith. Parametric approximation of data using ODR splines. GMR 7057, General Motors Research Laboratories, May 1990.
9.J.V. Miller, D.E. Breen, W.E. Lorensen, R.M. O’Bara, and M.J. Wozny. Geometrically deformed models: A method for extracting closed geometric models from volume data. Computer Graphics (SIGGRAPH ’91 Proceedings), 25(4):217–226, July 1991.
10.William Schroeder, Jonathan Zarge, and William Lorensen. Decimation of triangle meshes. Computer Graphics (SIGGRAPH ’92 Proceedings), 26(2):65–70, July 1992.
11.R. B. Schudy and D. H. Ballard. Model detection of cardiac chambers in ultrasound images. Technical Report 12, Computer Science Department, University of Rochester, 1978.
12.R. B. Schudy and D. H. Ballard. Towards an anatomical model of heart motion as seen in 4-d cardiac ultrasound data. In Proceedings of the 6th Conference on Computer Applications in Radiology and Computer-Aided Analysis of Radiological Images, 1979.
13.Stan Sclaroff and Alex Pentland. Generalized implicit functions for computer graphics. Computer Graphics (SIGGRAPH ’91 Proceedings), 25(4):247–250, July 1991.
14.E. H. Spanier. Algebraic Topology. McGraw-Hill, New York, 1966.
15.Greg Turk. Re-tiling polygonal surfaces. Computer Graphics (SIGGRAPH ’92 Proceedings), 26(2):55–64, July 1992.
16.G. Wyvill, C. McPheeters, and B. Wyvill. Data structures for soft objects. The Visual Computer, 2(4):227–234, August 1986.

Citation

MLA

Hoppe, H., et al. “Mesh Optimization”. Proceedings of the 20th Annual Conference on Computer Graphics and Interactive Techniques, 1993, pp. 19–26, https://doi.org/10.1145/166117.166119.

APA

Hoppe, H., DeRose, T., Duchamp, T., McDonald, J., & Stuetzle, W. (1993). Mesh optimization. Proceedings of the 20th Annual Conference on Computer Graphics and Interactive Techniques, 19–26. https://doi.org/10.1145/166117.166119

Chicago

Hoppe, H., T. DeRose, T. Duchamp, J. McDonald, and W. Stuetzle. 1993. “Mesh Optimization”. Proceedings of the 20th Annual Conference on Computer Graphics and Interactive Techniques, 19–26. https://doi.org/10.1145/166117.166119.

Harvard

Hoppe, H. et al. (1993) “Mesh optimization”, Proceedings of the 20th annual conference on Computer graphics and interactive techniques. ACM, pp. 19–26. Available at: https://doi.org/10.1145/166117.166119.

Vancouver

1. Hoppe H, DeRose T, Duchamp T, McDonald J, Stuetzle W (1993) Mesh optimization. In: Proceedings of the 20th annual conference on Computer graphics and interactive techniques. ACM, pp 19–26

BibTeX

@inproceedings{Hoppe_1993, series={SIGGRAPH93}, title={Mesh optimization}, url={http://dx.doi.org/10.1145/166117.166119}, DOI={10.1145/166117.166119}, booktitle={Proceedings of the 20th annual conference on Computer graphics and interactive techniques}, publisher={ACM}, author={Hoppe, Hugues and DeRose, Tony and Duchamp, Tom and McDonald, John and Stuetzle, Werner}, year={1993}, month=Sept, pages={19–26}, collection={SIGGRAPH93} }

Metadata:Crossref

Source Code

This paper has an official code repository available. Click below to access the source code.

View Repository

Access the Paper

This paper is available from its original source. Click below to access the PDF.

Open PDF

Mesh optimization

Abstract

Topics

Keywords

Table of Contents

Knowls

Knowl 1 — Mesh Optimization Energy Function

Knowl 2 — Nested Minimization Algorithm for Mesh Optimization

Knowl 3 — Continuous Vertex Optimization via Alternating Projections and Linear Least Squares

Knowl 4 — Topology-Preserving Legal Move Conditions for Simplicial Complex Modifications

Knowl 5 — Local Energy Evaluation and Candidate Set Heuristics for Mesh Transformations

Knowl 6 — Regularization Annealing Schedule and Coordinate Normalization