Pith. sign in

REVIEW 5 major objections 7 minor 31 references

Power Diagram Enhanced Adaptive Isosurface Extraction from Signed Distance Fields

T0 review · 5 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that isosurface extraction from signed distance fields can be made adaptive by building a power diagram from SDF samples weighted by squared distances, reading the mesh from its dual regular Delaunay tetrahedralization…

desk verdict A promising isosurface extraction method that pairs power diagrams with regular Delaunay refinement, but the theoretical support is incomplete and the paper lacks code/data; it deserves peer review with major revisions. read the letter →

arxiv 2506.09579 v2 pith:XBRNHO7X submitted 2025-06-11 cs.CG

Pengfei Wang , Ziyang Zhang , Wensong Wang , Shuangmin Chen , Lin Lu , Shiqing Xin , Changhe Tu This is my paper
classification cs.CG MSC 68U0565D18
keywords signeddistancefieldsisosurfaceextractionpowerdiagramregularDelaunaytetrahedralizationadaptiverefinementsurfacereconstructionzerolevelsettopologypreservation
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

The paper sets out to show that isosurface extraction from a signed distance field can be made adaptive and high-fidelity by treating SDF samples as sites of a power diagram with squared distances as weights, then reading the surface off the diagram's dual regular Delaunay tetrahedralization. The proposed refinement loop inserts new samples precisely where the current mesh's face normals disagree with the field gradient, so the mesh converges to the zero level set while only a local patch of the complex changes per insertion. If the central claim holds, this gives an extractor that recovers thin structures, small cavities, and high-genus topology that uniform-grid methods and earlier sphere-based methods miss, at a computational cost that grows roughly linearly with the number of samples. The paper reports quantitative improvements, including a Chamfer distance of 2.68e-5 at $30^{3}$ resolution compared with 11.5e-5 for the closest prior method on fixed low-resolution grids.

What carries the argument

The key objects are the power diagram — a weighted Voronoi partition in which each sample owns the region where its squared distance minus its assigned weight is minimal — and its dual, the regular Delaunay tetrahedralization. The paper separates sites into two categories — inner samples together with outer projections, versus outer samples together with inner projections — so that every tetrahedron mixing the two categories splits as 2-2 or 1-3, and then applies edge-based dualization rules that turn each such tetrahedron into a triangular or quadrilateral surface patch. The deviation metric δ(△), defined as the integral over a face of the component of the SDF gradient orthogonal to the face normal, is the signal that decides where the next sample is placed: the face with the largest δ is refined by inserting the dual vertex of its tetrahedron and that vertex's surface projection. The locality of regular Delaunay insertion is what makes the refinement cheap, since each update touches only the affected neighborhood.

What would settle it

Take a signed distance field with a known analytic surface, run the method from a coarse uniform sample set, and inspect the output mesh for boundary edges, non-manifold edges, or triangles that cross an inner or outer tangent sphere; any such defect would show the dualization rules do not always give the claimed valid surface. A more direct check is to enumerate all assignments of the four site types to the vertices of a single tetrahedral cell and verify that the four edge-based dual vertices form one consistently oriented patch whose edges match the neighboring cells.

Watch

Extended reading notes

Core claim

The central discovery, stated on the paper's own terms, is that a power diagram constructed from SDF sample sites with squared distances as weights has a boundary which lies between the inner and outer tangent spheres without intersecting them, and that adding each sample's projection onto the zero level set as an extra site forces the diagram's boundary to become tangent to the surface at those projection points. Because direct power-diagram boundaries are geometrically unstable, the paper extracts the surface mesh from the regular Delaunay tetrahedralization dual to the power diagram: each tetrahedron whose vertices come from both site categories is dualized into a triangle or quadrilateral that represents the surface patch inside that tetrahedron, using per-edge rules that interpolate between the relevant site positions. The regular Delaunay complex supports localized incremental insertion, so adding a new sample only rewrites a small neighborhood of the complex and of the extracted mesh. The paper's second claimed discovery is that the integrated misalignment between the SDF gradient and the face normal is an effective proxy for reconstruction error, and that greedily sampling at the face with the largest such misalignment progressively corrects both geometry and topology until a budget or tolerance is reached. On the paper's evidence, this combination recovers fine details and complex topologies that prior methods miss, and does so with a per-insertion cost that remains roughly constant.

Load-bearing premise

The load-bearing premise is that every tetrahedral cell whose corners mix the two categories of sample sites can always be converted into a surface patch, using the paper's listed edge rules, in a way that joins neighboring patches into a watertight mesh; the paper states this without proof.

Editorial extensions

If this is right

  • On fixed uniform grid inputs, the paper reports Chamfer distance 2.68e-5 at 30^3 resolution versus 11.5e-5 for the closest prior method, with the gap widening as resolution increases.
  • Because each inserted sample changes only a local neighborhood of the regular Delaunay complex, the time per insertion stays roughly constant and the total cost scales approximately linearly with the point count on the tested models.
  • The dual patch construction produces meshes with both triangles and quadrilaterals, avoiding the zigzag artifacts seen when power diagram boundaries are used directly.
  • Adding surface projections as extra sites makes the extracted mesh tangent to the true surface at those projections, which the paper argues is what preserves thin structures, sharp features, and high-genus topology.
  • The same adaptive loop works on neural SDFs such as SIREN and VolSDF, with batched insertion used to keep the number of network queries manageable.

Reading between the lines

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

  • A natural test the paper does not run is to degrade the gradient field, for example by adding controlled noise or using a neural SDF with non-unit gradients, and measure how reconstruction quality falls; the paper notes projections become inaccurate but does not quantify the breakpoint.
  • The watertightness of the extracted mesh rests on an unproved combinatorial claim about the dualization; proving that the dual patches agree on shared tetrahedral faces would upgrade the empirical results into a guarantee, and finding a counterexample would bound the method's validity.
  • The same weighted-site construction could be applied to fields that are not true signed distances, such as occupancy or density fields, if the projection step is replaced by a gradient-based closest-point estimate; the paper does not explore this setting.
  • The gradient-alignment metric can in principle score a flat, normal-aligned triangle highly even if it sits away from the zero level set; the paper's safeguard is that neighboring triangles will carry high energy, which could be stress-tested on a synthetic flat slab.
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.

Referee Report

5 major / 7 minor

Summary. The paper proposes an adaptive isosurface extraction method for signed distance fields that combines an incrementally constructed power diagram with regular Delaunay tetrahedralization. SDF sample points and their surface projections are used as weighted sites; the surface is extracted by processing tetrahedra with mixed site types using a set of dualization rules. A quality metric based on gradient alignment with triangle normals drives adaptive point insertion, and the process is claimed to refine the mesh locally until a user-specified tolerance or point budget is reached. Experiments compare the method against Marching Cubes, Dual Contouring, RFTS, RFTA, and McGrids, reporting improved Chamfer distances and better preservation of fine details and complex topologies, plus approximately linear scaling of insertion cost.

Significance. If the construction is correct, the paper offers a genuinely useful combination of ideas: using squared SDF values as power diagram weights, incorporating surface projections as sites, and exploiting the locality of regular Delaunay updates to make progressive refinement cheap. The reported quantitative gains in Tables I and II, especially for complex topologies, are substantial and would be of interest to the geometry processing community. However, the paper's central combinatorial step—the dualization of mixed tetrahedra into a watertight surface—is asserted rather than proved, and the theoretical result in Section IV.A does not cover the actual mesh produced by these rules. The empirical results are consistent with the claims, but they do not by themselves establish correctness of the extraction rules. The manuscript is promising but needs a rigorous treatment of the dualization step before the central claim can be accepted.

major comments (5)
  1. [Section IV.A, Theorem 1] The theorem is stated as establishing that the power diagram boundary forms a valid surface, but the proof does not establish this. Equations (1)-(2) only show that a boundary point with equal power to one inner and one outer site cannot lie strictly inside either sphere, assuming inner and outer spheres are disjoint. The proof does not show that the power diagram boundary is a 2-manifold, that it separates space into inside and outside, or that every boundary facet lies between sites of opposite categories. The sentence "Since x lies on the power diagram boundary, there must exist an outer site p_j^+ with equal power distance" is unjustified unless the boundary is already known to separate inner from outer sites, which is part of what needs proving. The disjointness of inner and outer spheres is also an assumption that should be stated explicitly; it holds for globally 1-Lipschitz SDFs but not for arbitrary fields, including some neural SDFs used later in Section VI.H.
  2. [Section IV.C] The paper first describes "four distinct categories of sites" (Figure 3(c) and the text introducing P−, P+, P⊥−, P⊥+), then asserts that every tetrahedron containing different types of sites has only a 2-2 split or a 1-3 split. This conclusion is valid only if sites are partitioned into exactly two types. With four categories, a tetrahedron can contain one site from each category, a 1-1-1-1 configuration, and no extraction rule is provided for that case. The intended type system needs to be clarified, and a proof is needed that the claimed splits exhaust all configurations that occur in the regular Delaunay tetrahedralization.
  3. [Section IV.C, dualization rules] The dualization rules are asserted without a correctness or topology argument. The paper does not show that the dual vertices produced for a tetrahedron are non-degenerate, that the connecting rule "according to the cell's topology" is a well-defined triangulation rule for both 2-2 and 1-3 splits, or that adjacent tetrahedra produce compatible faces so that the resulting mesh is watertight. This is a load-bearing issue because the accuracy numbers in Tables I and II are measured on meshes produced by these rules. If some mixed tetrahedron yields a 1-1-1-1 configuration, or if the prescribed connections create inconsistent edges between neighboring tetrahedra, the reported improvements would not rest on a guaranteed-valid mesh. A proof or a precise citation to a standard dualization theorem for regular triangulations is required.
  4. [Section V-B, Eq. (3)] The deviation metric δ(△) measures only the tangential component of the SDF gradient. A triangle that is perfectly parallel to the zero level set but located at a distance from it receives δ = 0. The paper's defense, namely that projection points keep triangles near the surface and that adjacent triangles would have high energy, is heuristic and is not a formal guarantee that the adaptive refinement loop cannot stall on a misplaced patch. Since the stopping condition δ < ε is the basis for the claim that the mesh "sufficiently approximates" the surface, this gap should be addressed either by a formal argument or by explicitly presenting the metric as a heuristic rather than a convergence certificate.
  5. [Section VI.B, Table I] The comparison labeled "fixed grid inputs" may not be input-equivalent. The text says that all algorithms use identical grid point SDF values and gradient information, but the proposed method appears to add each grid point's surface projection as an additional site, giving it roughly twice the number of sites or additional surface-tangent information. If this is the case, the Chamfer distance advantage in Table I is partly attributable to extra input, not solely to the extraction framework. Please state the exact site counts and the input information used by each method in Table I and Figure 6, and, if projections are included, either rerun the comparison with matched information or clearly label the setting as using additional surface points.
minor comments (7)
  1. [Abstract] The phrase "outperforms sofa methods" appears to be a typo and should read "outperforms state-of-the-art methods."
  2. [Section IV.B] The sentence "with P− and P+ representing exterior and interior sampling points" reverses the convention established in Section IV.A, where P− denotes negative/interior points and P+ denotes positive/exterior points. Please correct this inconsistency.
  3. [Section IV.C] The rule for an edge between P− and P+ is well-defined as a vector-valued affine interpolation: v1*phi(v2) - v2*phi(v1) is a vector and phi(v2) - phi(v1) is a scalar. The dimensional inconsistency concern raised in the review does not apply to this formula as written.
  4. [Section IV.C] For a 2-2 split the dual vertices form a quadrilateral, but the paper does not specify how a non-planar quadrilateral is triangulated. Without this detail the mesh is not unambiguously defined for such cases.
  5. [Section VI.D] Table II's heading "CD·105" should be "CD·10^5". The paper should also state whether the Chamfer distance is symmetric and how the one million evaluation points are generated.
  6. [Section VI.E] The sentence "Figure 9 presents a analysis" should be "Figure 9 presents an analysis."
  7. [Section VI.H] The claim that neural SDF errors exhibit a "self-balancing property" is asserted without evidence or reference; this should be either supported or removed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method's outputs are not defined in terms of its reported metrics, no fitted constants are renamed as predictions, and the load-bearing geometric facts are either proved in the paper or cited from non-overlapping prior work.

full rationale

I walked the derivation chain: (1) the power-diagram construction and Theorem 1 are self-contained; the theorem is proved from the non-intersection of inner and outer SDF spheres, and the proof does not presuppose the extraction result. (2) The actual surface is extracted from the regular Delaunay tetrahedralization via the dualization rules in Section IV.C. These rules are algorithmic choices, not quantities fitted to the benchmark tables; the reported Chamfer distances, normals, and F1 scores are measured on the produced meshes, not optimized by construction. (3) Projection points are obtained from SDF gradients, and the adaptive insertion strategy uses a gradient-alignment deviation metric; neither is calibrated against Tables I or II. The stopping criteria (epsilon, k_max, initial grid) are user-selected parameters, not fitted to the reported accuracy. (4) External references to RFTS/RFTA/McGrids are by non-overlapping authors and are used for comparison or for the standard property that distance-defined spheres are tangent to the surface; there is no self-citation chain that carries the central claim. I did note two unproven premises: the assertion that every mixed tetrahedron has only a 2-2 or 1-3 split, and the gap between Theorem 1 (power diagram boundary) and the regular-Delaunay-derived mesh. These are correctness risks, not circularity, because the method's output is not definitionally equivalent to its inputs and no prediction reduces to a fitted value. Hence the appropriate circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The algorithm introduces no new physical entities or fitted constants. Its parameters are user-controlled stopping thresholds. The main unproven assumptions are the correctness of the dualization rules and the efficiency of incremental Delaunay updates.

free parameters (3)
  • epsilon (accuracy tolerance) = not reported
    Stopping criterion in Algorithm Step 4. User-chosen threshold, not fitted to data.
  • k_max (max point count) = varies (e.g., 40,000, 50,000, 60,000)
    Stopping criterion. Selected per experiment; not fitted to optimize reported metrics.
  • initial grid resolution = e.g., 8^3, 15^3, 20^3
    Initial sampling density for the regular Delaunay. Varied in ablations; not fitted.
assumptions (5)
  • domain assumption Inner and outer tangent spheres of exact SDF samples do not intersect.
    Used in the proof of Theorem 1 (Section IV.A). True for exact SDF where sphere radius equals distance to the surface; may fail for approximate neural SDFs.
  • domain assumption SDF gradients are available and unit-length, pointing along the surface normal.
    Required for projection points (Section IV.B) and the gradient-alignment deviation metric (Section V.B). Neural SDFs do not strictly satisfy this, as the paper admits in Section VI.H.
  • standard math The power diagram of weighted sites has a dual regular Delaunay tetrahedralization that can be updated locally.
    Standard computational geometry result invoked in Sections IV.C and IV.D.
  • ad hoc to paper The dualization rules in Section IV.C produce a mesh that approximates the zero level set and is topologically faithful.
    Stated without proof. No argument that all tetrahedra reduce to the 2-2 or 1-3 splits, nor that connecting dual vertices yields a watertight manifold.
  • domain assumption Localized Delaunay updates give approximately linear overall runtime.
    Claimed in Section VI.E based on empirical timing; no worst-case or expected-case complexity analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Power Diagram Enhanced Adaptive Isosurface Extraction from Signed Distance Fields." pith.science (2026). https://pith.science/paper/XBRNHO7X

@misc{pith2026250609579,
  author       = {Pith},
  title        = {Pith review of: Power Diagram Enhanced Adaptive Isosurface Extraction from Signed Distance Fields},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XBRNHO7X}},
  note         = {Machine review of arXiv:2506.09579}
}
read the original abstract

Extracting high-fidelity mesh surfaces from Signed Distance Fields has become a fundamental operation in geometry processing. Despite significant progress over the past decades, key challenges remain namely, how to automatically capture the intricate geometric and topological structures encoded in the zero level set of SDFs. In this paper, we present a novel isosurface extraction algorithm that introduces two key innovations: 1. An incrementally constructed power diagram through the addition of sample points, which enables repeated updates to the extracted surface via its dual regular Delaunay tetrahedralization; and 2. An adaptive point insertion strategy that identifies regions exhibiting the greatest discrepancy between the current mesh and the underlying continuous surface. As the teaser figure shows, our framework progressively refines the extracted mesh with minimal computational cost until it sufficiently approximates the underlying surface. Experimental results demonstrate that our approach outperforms sofa methods, particularly for models with intricate geometric variations and complex topologies.

Figures

Figures reproduced from arXiv: 2506.09579 by the authors.

Figure 1
Figure 1. The workflow of our isosurface extraction. Starting from an initial coarse mesh surface, our method iteratively refines [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Compared with existing isosurface extraction ap [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. (a) Each sampling site generates a sphere with its center [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Given a signed distance field with associated gradients, we initialize the surface by constructing a regular Delaunay [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: When a face is well-aligned with the zero-level set, [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Comparison with state-of-the-art methods on three low-genus inputs. Results are generated using uniform [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Comparison with state-of-the-art methods on two high-genus inputs. Previous methods and Ours [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Performance scaling of our algorithm on two com [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Intermediate stages of our algorithm demonstrated on two classic models with high genus and rich surface details. The [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: Comparison of results using (a) sampling points only [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]
Figure 13
Figure 13. Figure 13: Neural SDF extraction in two representative scenarios: [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 29 canonical work pages

  1. [1]

    S.F.F. Gibson. Using distance maps for accurate surface representation in sampled volumes. In IEEE Symposium on V olumeVisualization (Cat. No.989EX300), pages 23–30, 1998

    work page 1998

  2. [2]

    Frisken, Ronald N

    Sarah F. Frisken, Ronald N. Perry, Alyn P. Rockwood, and Thouis R. Jones. Adaptively Sampled Distance Fields: A General Representation of Shape forComputer Graphics. Association for Computing Machinery, New York, NY , USA, 1 edition, 2023

    work page 2023

  3. [3]

    Jones, J.A

    M.W. Jones, J.A. Baerentzen, and M. Sramek. 3d distance fields: a sur- vey of techniques and applications. IEEE Transactions on Visualization and Computer Graphics, 12(4):581–599, 2006

    work page 2006

  4. [4]

    Level set methods and fast marching methods:

    Marko Suba ˇsi´c. Level set methods and fast marching methods:. Journal of Computing and Information Technology (cit@srce.hr); V ol.11No.1, 11, 03 2003

    work page 2003

  5. [5]

    Stanley Osher and R. Fedkiw. The Level Set Methods and Dynamic Implicit Surfaces, volume 57, pages xiv+273. 05 2004

    work page 2004

  6. [6]

    Malladi and J.A

    R. Malladi and J.A. Sethian. Level set and fast marching methods in image processing and computer vision. In Proceedings of 3rd IEEE International Conference on Image Processing, volume 1, pages 489– 492 vol.1, 1996

    work page 1996

  7. [7]

    Constructive solid geometry on neural signed distance fields

    Zo ¨e Marschner, Silvia Sell ´an, Hsueh-Ti Derek Liu, and Alec Jacobson. Constructive solid geometry on neural signed distance fields. In SIGGRAPH Asia 2023 Conference Papers, SA ’23, New York, NY , USA, 2023. Association for Computing Machinery

    work page 2023

  8. [8]

    Deepsdf: Learning continuous signed distance func- tions for shape representation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. Deepsdf: Learning continuous signed distance func- tions for shape representation. In The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2019

    work page 2019

Show all 31 references
  1. [9]

    Neural geometric level of detail: Real-time rendering with implicit 3D shapes

    Towaki Takikawa, Joey Litalien, Kangxue Yin, Karsten Kreis, Charles Loop, Derek Nowrouzezahrai, Alec Jacobson, Morgan McGuire, and Sanja Fidler. Neural geometric level of detail: Real-time rendering with implicit 3D shapes. 2021

    2021

  2. [10]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Ko- mura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. arXiv preprint arXiv:2106.10689, 2021

    2021 arXiv

  3. [11]

    Lorensen and Harvey E

    William E. Lorensen and Harvey E. Cline. Marching cubes: A high resolution 3d surface construction algorithm. In Proceedings of the 14th Annual Conference on Computer Graphics and Interactive Techniques, SIGGRAPH ’87, page 163–169, New York, NY , USA, 1987. Association for Com...

    1987

  4. [12]

    Neural marching cubes

    Zhiqin Chen and Hao Zhang. Neural marching cubes. CoRR, abs/2106.11272, 2021

    2021 arXiv

  5. [13]

    An efficient method of triangulating equi- valued surfaces by using tetrahedral cells

    Akio Doi and Akio Koide. An efficient method of triangulating equi- valued surfaces by using tetrahedral cells. IEICE TRANSACTIONS on Information and Systems, 74(1):214–224, 1991

    1991

  6. [14]

    Dual contouring of hermite data

    Tao Ju, Frank Losasso, Scott Schaefer, and Joe Warren. Dual contouring of hermite data. ACM Trans. Graph., 21(3):339–346, July 2002

    2002

  7. [15]

    Deep marching cubes: Learning explicit surface representations

    Yiyi Liao, Simon Donne, and Andreas Geiger. Deep marching cubes: Learning explicit surface representations. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2916– 2925, 2018

    2018

  8. [16]

    Neural dual contouring

    Zhiqin Chen, Andrea Tagliasacchi, Thomas Funkhouser, and Hao Zhang. Neural dual contouring. ACM Transactions on Graphics (Special Issue of SIGGRAPH), 41(4), 2022

    2022

  9. [17]

    Richter, Beno ˆıt Guil- lard, Timur Bagautdinov, Pierre Baque, and Pascal Fua

    Edoardo Remelli, Artem Lukoianov, Stephan R. Richter, Beno ˆıt Guil- lard, Timur Bagautdinov, Pierre Baque, and Pascal Fua. Meshsdf: differ- entiable iso-surface extraction. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS ’20,...

    2020

  10. [18]

    Reach for the spheres: Tangency-aware surface reconstruction of sdfs, 2023

    Silvia Sell ´an, Christopher Batty, and Oded Stein. Reach for the spheres: Tangency-aware surface reconstruction of sdfs, 2023

    2023

  11. [19]

    Reach for the arcs: Reconstructing surfaces from sdfs via tangent points

    Silvia Sell ´an, Yingying Ren, Christopher Batty, and Oded Stein. Reach for the arcs: Reconstructing surfaces from sdfs via tangent points. In SIGGRAPH 2024 Conference Papers, 2024

    2024

  12. [20]

    Mcgrids: Monte carlo-driven adaptive grids for iso-surface extraction, 2024

    Daxuan Renınst, Hezi Shiınst, Jianmin Zheng, and Jianfei Cai. Mcgrids: Monte carlo-driven adaptive grids for iso-surface extraction, 2024

    2024

  13. [21]

    An efficient method of triangulating equi- valued surfaces by using tetrahedral cells

    Akio Doi and Akio Koide. An efficient method of triangulating equi- valued surfaces by using tetrahedral cells. IEICE Transactions on Information and Systems, 74(1):214–224, 1991

    1991

  14. [22]

    G.M. Nielson. Dual marching cubes. In IEEE Visualization 2004, pages 489–496, 2004

    2004

  15. [23]

    Unconstrained Isosurface Extraction on Arbitrary Octrees

    Michael Kazhdan, Allison Klein, Ketan Dalal, and Hugues Hoppe. Unconstrained Isosurface Extraction on Arbitrary Octrees. In Alexan- der Belyaev and Michael Garland, editors, Geometry Processing. The Eurographics Association, 2007

    2007

  16. [24]

    Self-adaptive marching cubes 3d surface reconstruction method for 3d-gis spatial raster volume data

    Tao Jun, Liu Xu, Wang Yongming, and Chen Jianjie. Self-adaptive marching cubes 3d surface reconstruction method for 3d-gis spatial raster volume data. In 2012 International Conference on Image Analysis and Signal Processing, pages 1–4, 2012

    2012

  17. [25]

    Adaptive marching cubes

    Renben Shu, Chen Zhou, and Mohan S Kankanhalli. Adaptive marching cubes. The Visual Computer, 11:202–217, 1995

    1995

  18. [26]

    V oromesh: Learning watertight surface meshes with voronoi diagrams, 2023

    Nissim Maruani, Roman Klokov, Maks Ovsjanikov, Pierre Alliez, and Mathieu Desbrun. V oromesh: Learning watertight surface meshes with voronoi diagrams, 2023

    2023

  19. [27]

    Hilton, A.J

    A. Hilton, A.J. Stoddart, J. Illingworth, and T. Windeatt. Marching trian- gles: range image fusion for complex object modelling. In Proceedings of 3rd IEEE International Conference on Image Processing, volume 2, pages 381–384 vol.2, 1996

    1996

  20. [28]

    Paul Chew

    L. Paul Chew. Guaranteed-quality mesh generation for curved surfaces. In Proceedings of the Ninth Annual Symposium on Computational Geometry, SCG ’93, page 274–280, New York, NY , USA, 1993. Association for Computing Machinery

    1993

  21. [29]

    On volumetric shape reconstruction from implicit forms

    Li Wang, Franck H ´etroy-Wheeler, and Edmond Boyer. On volumetric shape reconstruction from implicit forms. In European Conference on Computer Vision, 2016

    2016

  22. [30]

    Lorensen and Harvey E

    William E. Lorensen and Harvey E. Cline. Marching cubes: A high resolution 3d surface construction algorithm. SIGGRAPH Comput. Graph., 21(4):163–169, August 1987

    1987

  23. [31]

    gptyoolbox: A python geometry processing toolbox, 2023

    Silvia Sell ´an, Oded Stein, et al. gptyoolbox: A python geometry processing toolbox, 2023. https://gpytoolbox.org/

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.