Isotropic Remeshing with Inter-Angle Optimization

Chenlei Lv; Hanbing Zheng

REVIEW 3 major objections 5 minor 31 references

Isotropic Remeshing with Inter-Angle Optimization

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that angle-aware checks on split, collapse, and flip operations make isotropic remeshing both faster and more shape-faithful than the standard four-step strategy.

desk verdict Angle-gated remeshing is a modest but real improvement; the MLS surface-keeping mechanism doesn't match the equations and needs rewriting. read the letter →

arxiv 2507.13641 v2 pith:RK2JLVNO submitted 2025-07-18 cs.CG

Hanbing Zheng , Chenlei Lv This is my paper

classification cs.CG MSC 68U0565D18

keywords isotropicremeshinginter-angleoptimizationtriangularmeshMovingLeastSquaresobtuseanglesuppressionfour-steps

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the classic four-step strategy for isotropic remeshing—split, collapse, flip, and tangent smoothing—can be made faster and more stable by predicting, before each edit, how the surrounding triangle angles will change. The authors claim that blocking splits and flips that would create new obtuse angles, guarding collapses with boundary and vertex-degree conditions, and remapping vertex positions onto a Moving Least Squares surface derived from a denser point set prevents "fixing one issue while creating another" and avoids wasteful repeated editing. If this is right, remeshing would need fewer iterations, preserve shape more faithfully on sparse meshes, and still reach near-equilateral triangles. The reported experiments show an average maximum inter-angle of 113.8 degrees across a set of 12 benchmark models, versus 175.6, 169.6, and 179.7 degrees for three comparison methods, while average runtime is 10.89 seconds compared with 20.63 and 38.93 seconds for two of those methods.

What carries the argument

The load-bearing mechanism is the inter-angle optimization scheme applied to the split, collapse, and flip operations, together with Moving Least Squares up-sampling for the vertex-relocation step. For each operation, the algorithm estimates the shape of the triangles that would result from the edit and blocks the edit if it would create obtuse angles, disturb a boundary, or break a sharp feature. The angle checks couple the operations so that one edit does not undo the work of another, and the MLS upsampling supplies a denser reference surface so that tangent smoothing does not pull vertices off the original shape.

What would settle it

Take a sparse mesh with a known sharp crease, run the full remeshing pipeline, and compare the output against a high-resolution ground-truth scan: if the mean distance to the ground truth is no lower than that of an unconstrained four-step baseline, or if the maximum inter-angle of the output exceeds the reported 113.8-degree average on a set of such meshes, the geometric-consistency and angle-prediction claims would be undermined.

Watch

Extended reading notes

Core claim

The central claim is that controlling the three basic mesh-editing operations through local angle checks—rather than only edge-length checks—improves both the isotropic quality and the geometric fidelity of the output, and that the improvement is large enough to matter in practice. The split operation is blocked when any of the four adjacent angles in the face is obtuse, because splitting would create even larger obtuse angles; the collapse operation is blocked on boundary edges and when the resulting vertex degree would push the average inter-angle below roughly 60 degrees; and the flip operation is blocked when it would generate new obtuse angles or when the dihedral angle across the edge exceeds a 20-degree threshold that signals a sharp feature. For the vertex-relocation step, the paper inserts seven new points into each triangle, remaps them onto a Moving Least Squares surface, and then performs tangent smoothing by pulling vertices back toward this denser reference neighborhood. The paper claims this yields a better balance between geometric consistency and isotropy, with lower mean distance to the input and a much smaller maximum inter-angle.

Load-bearing premise

The central assumption is that the Moving Least Squares surface built from the sparse input mesh faithfully represents the true shape, so the up-sampled points used for pulling back vertex positions do not pull the mesh toward a wrong geometry.

Editorial extensions

If this is right

On meshes similar to the tested benchmark set, the method produces remeshes with a much smaller maximum inter-angle, about 113.8 degrees on average versus 175.6, 169.6, and 179.7 degrees for the three compared baselines.
The average runtime is lower than the compared four-step remeshing variants, 10.89 seconds versus 20.63 and 38.93 seconds, and far lower than a centroidal Voronoi tessellation baseline at roughly 391 seconds.
Because the angle checks prevent conflicting edits, each iteration makes more progress toward isotropy, meaning fewer repetitive split-collapse-flip cycles are needed for convergence.
The MLS-based upsampling keeps the mean distance from the remeshed output to the input approximately as low as or lower than the baselines, while suppressing local distortion on sparser meshes.
Adjusting the target edge length still produces simplified or refined meshes that retain the isotropic property, which is useful for multi-resolution editing and for cleaning up meshes produced by generative 3D pipelines.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper's experiments, the same inter-angle principle could be applied to anisotropic remeshing by replacing the 60-degree target with a prescribed local angle field, rather than aiming only at equilateral triangles.
The fixed 20-degree dihedral threshold and the fixed step size are parameters; choosing them adaptively based on local feature scale might improve sharp-feature preservation on meshes with legitimate creases.
A controlled ablation that toggles each of the split, collapse, and flip checks on and off would reveal which check contributes most to the reported speedup and to the reduction in maximum inter-angle.
Because the MLS surface is constructed from the input mesh itself, its accuracy on extremely noisy or non-manifold inputs is an open question; stressing the pipeline with raw reconstruction outputs would test how far the geometric-consistency guarantee extends.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

First thing to know: this is a modest but real improvement to the four-steps isotropic remeshing pipeline. The unified inter-angle gating for split, collapse, and flip is new as a combination — earlier works used angle heuristics in isolation — and the experiments show a major reduction in maximum inter-angle (113.8° average vs ~170–180° for baselines) with comparable mean distance. That is a concrete, useful gain for graphics and geometry processing. The efficiency claim is plausible: on the three models where all timings are reported, the method beats VCG and AdaISO, and the angle gating is a sensible mechanism for avoiding wasted operations. The paper's writing is clear and the related work is well placed.

The real soft spot is the MLS contribution. Section 1 promises that the MLS solution 'prevents vertex optimization from deviating from the MLS surface,' but Section 4.2 never projects vertices onto that surface. Eq. 2 is ordinary tangent-plane smoothing; the MLS upsampling only changes the neighborhood weights in Eq. 1. So the claimed geometric consistency mechanism is missing from the implementation as described. The lower mean distance in Table 1 might come from the angle constraints alone, or from the denser neighborhood, but the paper doesn't show it either way. This is not a fatal flaw, but it is a correctness issue in the method description that needs fixing in revision.

Other soft spots are smaller: no ablations separating inter-angle gating from MLS upsampling, no error bars or statistical tests across the 12 SHREC models, a few missing baseline entries in Table 1, and only three models in the timing table. None of these undermine the core angle-gating contribution, but they make the efficiency and 'geometric consistency' claims weaker than the framing suggests.

Worth a serious referee? Yes. The angle-gating idea is a legitimate extension of a well-known pipeline, the empirical improvement in angle quality is dramatic, and the flaws are fixable with rephrasing and ablation experiments. A good referee would ask for the MLS claim to be rewritten or implemented, for ablations, and for fuller baseline reporting. I would not cite this in my own work in the next year, but I'd send it to review rather than desk reject.

Referee Report

3 major / 5 minor

Summary. The manuscript proposes an isotropic remeshing method built on the classical four-step split/collapse/flip/smoothing pipeline, adding inter-angle constraints to the first three operations and an MLS-based upsampling step to support the tangent-smoothing stage. It reports experiments on 12 SHREC models (Table 1) and runtime comparisons (Table 2) against VCG, RAR, AdaISO, and FCVT, claiming a substantially reduced maximum inter-angle (113.8 degrees on average versus 175.6/169.6/179.7 degrees for the baselines) and faster average runtime than VCG and AdaISO while keeping mean distance comparable.

Significance. If the reported results hold, the proposed angle-filtered editing strategy is a useful heuristic contribution to a well-studied remeshing problem: the maximum-angle improvement is large, the method is evaluated on a public benchmark, and the paper includes a code release. These are concrete strengths. The central claim is falsifiable and the experimental setup is straightforward to reproduce. However, the current evidence does not yet establish the causal role of the MLS-based component, and the quantitative reporting has gaps that need to be addressed before the claims can be taken at face value.

major comments (3)

[Section 4.2, Eq. (2)] The claimed surface-constraint mechanism is not present in the described update. Equation (2) computes p_i' = p_i + lambda (I - n_i n_i^T) (p_i' - p_i), which moves p_i along the tangent plane at p_i; it does not project the updated vertex onto the MLS surface or onto the upsampled point set. The text only says that the neighborhood used by the 'pulling back' function is defined on the upsampled point set. Since the second contribution and the geometric-consistency claim rest on this mechanism, the authors must either add an explicit projection step and demonstrate that output vertices stay near the input surface, or revise the contribution and the causal attribution accordingly.
[Section 5, Tables 1 and 2] The quantitative support is incomplete. RAR and AdaISO have missing entries for several of the 12 models (e.g., T1022, T1078, T112, and T1134 in Table 1), and Table 2 reports runtimes for only T0, T14, and T28; the 'Avg' row therefore averages different model subsets for different columns, and the paper does not state how missing values are handled. In addition, all numbers are single-run measurements with no error bars or statistical tests, and the claimed mean-distance advantage (Table 1: Ours 0.0016 versus 0.0017 for each baseline) is within one unit of the last reported digit. The authors should report complete per-model data or clearly state the subset used, and provide multiple runs or a statistical comparison.
[Sections 4 and 5] No ablation separates the two proposed components. The maximum-angle improvement could come entirely from the angle filters in split, collapse, and flip, while the MLS upsampling might contribute little or nothing, or vice versa; the runtime gain could likewise be due to the filters alone. A small ablation on the same 12 models with variants 'without MLS upsampling' and 'without angle filters' would make the central attribution load-bearing and is necessary to support the claim that geometric consistency keeping is responsible for the reported quality.

minor comments (5)

[Section 4.2, Eq. (1) and Eq. (2)] The symbol p_i' is reused for both the weighted centroid in Eq. (1) and the updated vertex in Eq. (2); using distinct symbols such as c_i and p_i^{new} would remove ambiguity.
[Table 1, T1155 row] The AdaISO entry for T1155 reports theta_max = 79.9 degrees, while all other AdaISO entries are near 180 degrees; this appears to be a typo and should be checked.
[Section 5, metric definition] The definition of theta_avg as 'pi/3 - sum |theta_i - pi/3|' is not normalized or averaged over the mesh; clarify how the sum is computed and how the reported values around 52-53 degrees should be interpreted.
[Section 4.2] The seven-point insertion procedure is described only verbally; a small pseudocode block or figure would remove ambiguity about which midpoints are inserted and how the MLS remapping is applied.
[Section 4.1 and 4.2] The default thresholds (epsilon = 20 degrees for the dihedral-angle flip check, lambda = 0.5 for tangent smoothing) are stated without any sensitivity study; a brief paragraph on stability with respect to these parameters would strengthen the method description.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the experimental claims are validated against external baselines, and all thresholds are explicit hand-chosen heuristics rather than fitted parameters.

full rationale

The paper does not derive its reported quality or runtime numbers from a fitted target. The split, collapse, and flip criteria (4/3l, 4/5l, the 20-degree dihedral threshold, and lambda=0.5) are stated as fixed heuristic choices, and the evaluation compares against external baselines (VCG [4], RAR [9], FCVT [8], AdaISO [21]) on SHREC models. No quantity measured as an outcome (theta_max, theta_avg, Md, Hd, runtime) is used as an input to calibrate the method, so there is no fitted-input-called-prediction or self-definitional step. The self-citations [20,21] appear only as related-work context and as a comparison baseline; they do not carry the load of the main claim. One substantive concern is a possible mismatch between the advertised MLS geometric-consistency mechanism and Eq. 2, which only projects a displacement onto the tangent plane and does not explicitly reproject vertices onto the MLS surface; however, that is a correctness or evidence gap, not a circularity. Because the central claims are tested against external benchmarks and no derivation reduces to its own inputs, the circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

All parameters are algorithmic thresholds or step sizes set by hand or inherited from prior work. No parameters are fit to the SHREC data, and no new physical or mathematical entities are introduced.

free parameters (4)

flip dihedral angle threshold epsilon = 20 degrees (default)
Used in Section 4.1 Flip Scheme to block flips when the angle between face normals exceeds epsilon. Chosen by hand with no sensitivity study reported.
tangent smoothing step lambda = 0.5 (default)
Controls vertex update step size in Equation 2. Chosen by hand with no sensitivity study reported.
collapse degree judgment threshold
Collapse is blocked when the resulting vertex degree would drive the average inter-angle below about 60 degrees, but the exact degree cutoff is not stated, leaving an implicit hand-tuned parameter.
split and collapse length thresholds = 4/3 and 4/5 of target edge length
Inherited from Botsch and Kobbelt [4] and used without re-tuning in Section 3. They directly control when edit operations trigger.

assumptions (4)

domain assumption The input mesh is a manifold triangle mesh with well-defined edges, faces, and boundary.
The degree judgment and boundary conditions in Section 4.1 assume consistent local connectivity. Non-manifold inputs would break the collapse and flip checks.
domain assumption The MLS surface constructed from the original sparse mesh faithfully represents the underlying shape.
Section 4.2 replaces tangent-plane projection with neighborhoods defined on the MLS-upsampled point set. If the point set surface drifts, geometric consistency is not preserved.
domain assumption Standard four-step operations converge to an isotropic triangulation when given the new angle constraints.
The paper provides no convergence proof for the modified operations. It assumes the heuristic censoring does not prevent reaching the target edge length and degree distribution.
ad hoc to paper Local angle-based checks predict global convergence behavior.
The core design principle in Section 4.1 is that avoiding immediate obtuse angles and degree violations avoids fixing one issue while creating another. This local-to-global transfer is asserted, not proved.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Isotropic Remeshing with Inter-Angle Optimization." pith.science (2026). https://pith.science/paper/RK2JLVNO

@misc{pith2026250713641,
  author       = {Pith},
  title        = {Pith review of: Isotropic Remeshing with Inter-Angle Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RK2JLVNO}},
  note         = {Machine review of arXiv:2507.13641}
}

read the original abstract

As an important metric for mesh quality evaluation, the isotropy property holds significant value for applications such as texture UV-mapping, physical simulation, and discrete geometric analysis. Classical isotropy remeshing methods adjust vertices and edge lengths, which exhibit certain limitations in terms of input data sensitivity, geometric consistency control, and convergence speed. In this paper, we propose an improved isotropy remeshing solution with inter-angle optimization during mesh editing to enhance shape control capability and accelerate convergence. The advantage of the solution lies in its ability to predict the impact of edge length adjustments on subsequent optimization by monitoring angle transformations. It avoids inefficient editing that may cause performance fluctuations, thereby improving efficiency. Experiments demonstrate that the proposed method effectively improves the overall efficiency of mesh optimization.

Figures

Figures reproduced from arXiv: 2507.13641 by the authors.

**Figure 1.** Inter-angle optimization for split scheme. Angles ̸ α and ̸ β control the split operation for p1p2, which prevent the formation of ̸ α‘ and ̸ β ′ . Split Scheme. Controlling split operation is simpler because adding edges to existing faces for splitting does not change the original geometry. Nevertheless, the implementation of angular surveillance remains imperative. The reason is that performing a split operation … view at source ↗

**Figure 2.** Inter-angle optimization for collapse scheme. (a)∼(d) show the boundary condition; (e)∼(h) show the degree judgment. vertex-based degree analysis may lead to abnormal degree assignments for the new vertex, consequently increasing the difficulty of subsequent flip operations. From the perspective of inter-angle view, an excessively high vertex degree implies that the average inter-angle associated with that vertex … view at source ↗

**Figure 3.** Inter-angle optimization for flip scheme. (a)∼(c) show the inter-angle constraint; (e)∼(f) show the shape control. Flip Scheme. Beyond the primary objective of degree optimization, flip operation must account the inter-angle constraint, which is similar to the split scheme. An instance is shown in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Some remeshing results by different methods [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Inter-angle distribution histograms of re-meshed models (T0, T14, T28) [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Remeshing results by our method with different multi-parameters. Mp: multiparameter value; V: vertex number; F: face number [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

**Figure 7.** Figure 7: Time cost curves for different methods with related multi-parameters. ones based on a larger test dataset. In addition, we employ a CVT-based solution (FCVT) [8] to be a reference in [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 8.** Figure 8: Remeshing results based on 3D mesh generation framework. Left: generated meshes; right: remeshing results. ever, they suffer from poor isotropic quality and excessive mesh volume. Our method can effectively enhance the isotropic quality of meshes while simultaneously …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 30 canonical work pages

[1]

In: Proc

Alexa, M., Behr, J., Cohen-Or, D., Fleishman, S., Levin, D., Silva, C.T.: Point set surfaces. In: Proc. Visualization. pp. 21–29. IEEE (2001)

work page 2001
[2]

In: Proc

Alliez, P., Cohen-Steiner, D., Devillers, O., Lévy, B., Desbrun, M.: Anisotropic polygonal remeshing. In: Proc. SIGGRAPH, pp. 485–493 (2003)

work page 2003
[3]

Journal of computer and system sciences48(3), 384–409 (1994)

Bern, M., Eppstein, D., Gilbert, J.: Provably good mesh generation. Journal of computer and system sciences48(3), 384–409 (1994)

work page 1994
[4]

In: Proc

Botsch, M., Kobbelt, L.: A remeshing approach to multiresolution modeling. In: Proc. Eurographics Symp. on Geometry Processing. pp. 185–192 (2004)

work page 2004
[5]

Bronstein, A., Bronstein, M., Castellani, U., Falcidieno, B., Fusiello, A., Godil, A., Guibas, L., Kokkinos, I., Lian, Z., Ovsjanikov, M., et al.: Shrec 2010: robust large-scale shape retrieval benchmark. Proc. 3DOR5(4), 1–8 (2010)

work page 2010
[6]

In: Scarano, V., Chiara, R.D., Erra, U

Cignoni, P., Callieri, M., Corsini, M., Dellepiane, M., Ganovelli, F., Ranzuglia, G.: MeshLab: an Open-Source Mesh Processing Tool. In: Scarano, V., Chiara, R.D., Erra, U. (eds.) Eurographics Italian Chapter Conference. pp. 129–136. The Eurographics Association (2008)

work page 2008
[7]

Journal of Computational Physics262, 358–378 (2014)

Dapogny, C., Dobrzynski, C., Frey, P.: Three-dimensional adaptive domain remesh- ing, implicit domain meshing, and applications to free and moving boundary prob- lems. Journal of Computational Physics262, 358–378 (2014)

work page 2014
[8]

In: Computer Graphics Forum

Du, X., Liu, X., Yan, D.M., Jiang, C., Ye, J., Zhang, H.: Field-aligned isotropic surface remeshing. In: Computer Graphics Forum. vol. 37, pp. 343–357. Wiley Online Library (2018)

work page 2018

Show all 31 references

[9]

In: Proc

Dunyach, M., Vanderhaeghe, D., Barthe, L., Botsch, M.: Adaptive remeshing for real-time mesh deformation. In: Proc. Eurographics. pp. 29–32. The Eurographics Association (2013)

2013
[10]

ACM Trans

Goes, F.d., Memari, P., Mullen, P., Desbrun, M.: Weighted triangulations for ge- ometry processing. ACM Trans. on Graphics33(3), 28:1–28:13 (2014)

2014
[11]

In: Proc

He, X., Lv, C., Huang, P., Huang, H.: Windpoly: Polygonal mesh reconstruction via winding numbers. In: Proc. Euro. Conf. on Computer Vision. pp. 294–311. Springer (2024)

2024
[12]

Computer-Aided Design144, 103166.1–103166.12 (2022)

Hou, W., Zong, C., Wang, P., Xin, S., Chen, S., Liu, G., Tu, C., Wang, W.: Sdf-rvd: Restricted voronoi diagram on signed distance field. Computer-Aided Design144, 103166.1–103166.12 (2022)

2022
[13]

IEEE Trans

Hu, B.Y., Ye, C., Su, J.P., Liu, L.: Manifold-constrained geometric optimization via local parameterizations. IEEE Trans. Visualization & Computer Graphics29(2), 1318–1329 (2021)

2021
[14]

Engineering with Computers26(4), 363–376 (2010)

Jiao, X., Colombi, A., Ni, X., Hart, J.: Anisotropic mesh adaptation for evolving triangulated surfaces. Engineering with Computers26(4), 363–376 (2010)

2010
[15]

ACM Trans

Kazhdan, M., Hoppe, H.: Screened poisson surface reconstruction. ACM Trans. on Graphics 32(3), 29:1–29:13 (2013) 12 H. Zheng and C. Lv

2013
[16]

ACM Trans

Lévy, B., Liu, Y.: L p centroidal voronoi tessellation and its applications. ACM Trans. on Graphics29(4), 119:1–119:11 (2010)

2010
[17]

arXiv preprint arXiv:2408.10198 (2024)

Liu, M., Zeng, C., Wei, X., Shi, R., Chen, L., Xu, C., Zhang, M., Wang, Z., Zhang, X., Liu, I., et al.: Meshformer: High-quality mesh generation with 3d-guided re- construction model. arXiv preprint arXiv:2408.10198 (2024)

2024 arXiv
[18]

ACM Trans

Liu, Y.J., Xu, C.X., Fan, D., He, Y.: Efficient construction and simplification of delaunay meshes. ACM Trans. on Graphics34(6), 174:1–174:13 (2015)

2015
[19]

IEEE Trans

Lv, C., Lin, W., Zhao, B.: Voxel structure-based mesh reconstruction from a 3d point cloud. IEEE Trans. on Multimedia24, 1815–1829 (2021)

2021
[20]

IEEE Trans

Lv, C., Lin, W., Zhao, B.: Intrinsic and isotropic resampling for 3d point clouds. IEEE Trans. Pattern Analysis & Machine Intelligence45(3), 3274–3291 (2022)

2022
[21]

IEEE Trans

Lv, C., Lin, W., Zheng, J.: Adaptively isotropic remeshing based on curvature smoothed field. IEEE Trans. Visualization & Computer Graphics 30(7), 3196– 3209 (2024)

2024
[22]

ACM Trans

Lv, C., Wu, Z., Wang, X., Zhou, M.: 3d facial similarity measurement and its application in facial organization. ACM Trans. on Multimedia Computing, Com- munications, and Applications16(3), 82:1–82:20 (2020)

2020
[23]

Pattern Recognition88, 458–469 (2019)

Lv, C., Wu, Z., Wang, X., Zhou, M., Toh, K.A.: Nasal similarity measure of 3d faces based on curve shape space. Pattern Recognition88, 458–469 (2019)

2019
[24]

IEEE Trans

Wang, Y., Yan, D.M., Liu, X., Tang, C., Guo, J., Zhang, X., Wonka, P.: Isotropic surface remeshing without large and small angles. IEEE Trans. Visualization & Computer Graphics 25(7), 2430–2442 (2018)

2018
[25]

In: Computer Graphics Forum

Xu, Q.C., Yan, D.M., Li, W., Yang, Y.L.: Anisotropic surface remeshing without obtuse angles. In: Computer Graphics Forum. vol. 38, pp. 755–763. Wiley Online Library (2019)

2019
[26]

In: Computer Graphics Forum

Yan, D.M., Guo, J., Jia, X., Zhang, X., Wonka, P.: Blue-noise remeshing with farthest point optimization. In: Computer Graphics Forum. vol. 33, pp. 167–176. Wiley Online Library (2014)

2014
[27]

In: Computer Graphics Forum

Yan, D.M., Lévy, B., Liu, Y., Sun, F., Wang, W.: Isotropic remeshing with fast and exact computation of restricted voronoi diagram. In: Computer Graphics Forum. vol. 28, pp. 1445–1454. Wiley Online Library (2009)

2009
[28]

arXiv preprint arXiv:1907.00523 (2019)

Ye, Z., Yi, R., Yu, M., Liu, Y.J., He, Y.: Geodesic centroidal voronoi tessellations: Theories, algorithms and applications. arXiv preprint arXiv:1907.00523 (2019)

2019 arXiv
[29]

ACM Trans

Yi, R., Liu, Y.J., He, Y.: Delaunay mesh simplification with differential evolution. ACM Trans. on Graphics37(6), 263:1–263:12 (2018)

2018
[30]

ACM Trans

Zhang, R., Pan, S., Lv, C., Gong, M., Huang, H.: Architectural co-lod generation. ACM Trans. on Graphics (Proc. SIGGRAPH Asia)43(6), 193:1–193:16 (2024)

2024
[31]

In: Symposium on Interactive 3D Graphics and Games

Zheng, J., Tan, T.S.: Computing centroidal voronoi tessellation using the gpu. In: Symposium on Interactive 3D Graphics and Games. pp. 1–9 (2020)

2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

In: Proc

Alexa, M., Behr, J., Cohen-Or, D., Fleishman, S., Levin, D., Silva, C.T.: Point set surfaces. In: Proc. Visualization. pp. 21–29. IEEE (2001)

work page 2001

[2] [2]

In: Proc

Alliez, P., Cohen-Steiner, D., Devillers, O., Lévy, B., Desbrun, M.: Anisotropic polygonal remeshing. In: Proc. SIGGRAPH, pp. 485–493 (2003)

work page 2003

[3] [3]

Journal of computer and system sciences48(3), 384–409 (1994)

Bern, M., Eppstein, D., Gilbert, J.: Provably good mesh generation. Journal of computer and system sciences48(3), 384–409 (1994)

work page 1994

[4] [4]

In: Proc

Botsch, M., Kobbelt, L.: A remeshing approach to multiresolution modeling. In: Proc. Eurographics Symp. on Geometry Processing. pp. 185–192 (2004)

work page 2004

[5] [5]

Bronstein, A., Bronstein, M., Castellani, U., Falcidieno, B., Fusiello, A., Godil, A., Guibas, L., Kokkinos, I., Lian, Z., Ovsjanikov, M., et al.: Shrec 2010: robust large-scale shape retrieval benchmark. Proc. 3DOR5(4), 1–8 (2010)

work page 2010

[6] [6]

In: Scarano, V., Chiara, R.D., Erra, U

Cignoni, P., Callieri, M., Corsini, M., Dellepiane, M., Ganovelli, F., Ranzuglia, G.: MeshLab: an Open-Source Mesh Processing Tool. In: Scarano, V., Chiara, R.D., Erra, U. (eds.) Eurographics Italian Chapter Conference. pp. 129–136. The Eurographics Association (2008)

work page 2008

[7] [7]

Journal of Computational Physics262, 358–378 (2014)

Dapogny, C., Dobrzynski, C., Frey, P.: Three-dimensional adaptive domain remesh- ing, implicit domain meshing, and applications to free and moving boundary prob- lems. Journal of Computational Physics262, 358–378 (2014)

work page 2014

[8] [8]

In: Computer Graphics Forum

Du, X., Liu, X., Yan, D.M., Jiang, C., Ye, J., Zhang, H.: Field-aligned isotropic surface remeshing. In: Computer Graphics Forum. vol. 37, pp. 343–357. Wiley Online Library (2018)

work page 2018