Pith. sign in

REVIEW 2 major objections 4 minor 57 references

Uniform Sampling of Surfaces by Casting Rays

T0 review · 2 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Uniformly sampling rays through a bounding box and keeping every intersection with an implicit surface yields uniform white-noise surface samples.

desk verdict Solid, useful paper: the uniformity theorem is classical, but the practical ray-casting package for implicit surfaces is well-engineered, with the caveat that the guarantee is conditional on a valid Lipschitz bound. read the letter →

arxiv 2506.05268 v2 pith:GZT5JGFC submitted 2025-06-05 cs.GR

classification cs.GR MSC 65D1860D05
keywords uniformsamplingimplicitsurfacesraycastingspheretracingCauchy-Croftonformulawhitenoiseneuralblue
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 establishes that a classic geometric fact—random lines through a shape hit its surface in a pattern that mirrors surface area—can be turned into a practical, general-purpose sampler for implicit surfaces. It proves that casting uniformly distributed rays through a bounding box and recording every intersection with the zero level set produces white-noise samples that are area-proportional, in the strong sense of $k$-equidistribution, so no projection or mesh extraction is required. For signed-distance-like functions the intersection step is computed by sphere tracing with a known Lipschitz bound, and the paper demonstrates on neural and analytic implicits that this costs orders of magnitude fewer function evaluations than rejection sampling or marching cubes while matching or improving their uniformity. The payoff is a single primitive that feeds blue-noise generation, stratified sampling via sparse voxels, deformations of neural surfaces, and direct estimation of surface area, volume, and centroids.

What carries the argument

The carrying mechanism is the line-integral identity behind the Cauchy-Crofton formula: an integral of a function over a surface can be written as a constant times an integral over all oriented lines of the sum of the function's values at that line's intersections with the surface. Sampling the compact set of lines meeting the bounding box and returning every intersection lets each surface point be weighted by the number of sampled lines through it, and since each point lies on exactly one line per direction, a $k$-equidistributed line sequence produces a $k$-equidistributed point sequence. The practical tracer is a modified sphere march: with a known Lipschitz bound $\lambda$, step by $|f(p)|/\lambda$, record each zero crossing, then keep marching instead of stopping at the first hit.

What would settle it

Take an analytic torus inside $[-1,1]^3$, cast $M$ uniformly sampled rays using the exact continuous ray-torus intersection solver, bin the collected samples on a toroidal grid, and compare the per-bin proportions with the analytically known area distribution by total variation distance; if the distance does not converge to zero as $M\to\infty$, the claimed uniformity is false. The same test with an artificially underestimated Lipschitz bound in the sphere-tracing version should visibly miss thin features and inflate the total variation score.

Watch

Extended reading notes

Core claim

The central claim is that a surface sampler can be built directly from a ray-intersection subroutine: if oriented lines are sampled uniformly from the set of lines meeting a bounding box around the surface, and every intersection of each line with the surface is collected, then the resulting point sequence is $k$-equidistributed, hence uniform white noise on the surface. The paper proves this from the Cauchy-Crofton relation and shows that it holds for any surface representation with a ray tracer, including neural implicit functions when a modified sphere tracer marches with steps of size $|f(p)|/\lambda$. No mesh extraction and no projection step is needed, and sampled points lie exactly on the zero level set. The same framework yields surface area, volume, and centroid estimators from intersection and chord-length statistics, and the paper verifies uniformity on 114 mesh-based implicit surfaces against baselines of rejection sampling, Hamiltonian Monte Carlo, and marching-cubes extraction.

Load-bearing premise

The method needs a global Lipschitz bound for the implicit function that is a true upper bound everywhere in the box; when the supplied bound is too small, the tracing step can jump past a zero crossing and the sample distribution silently stops being uniform.

Editorial extensions

If this is right

  • Any surface representation with a ray-intersection subroutine—implicit signed distance functions, unsigned distance fields, neural fields, and even Gaussian particle scenes—can be uniformly sampled without meshing or projection.
  • On neural implicit surfaces the sampler is orders of magnitude cheaper than rejection sampling or marching-cubes extraction, with comparable or better measured uniformity.
  • Because samples lie exactly on the surface and are area-proportional, the same ray-casting pass yields unbiased Monte Carlo estimates of surface area, enclosed volume, shell centroid, and volumetric centroid.
  • Subsampling the white-noise output produces blue noise, sparse voxel partitioning gives stratified low-variance sampling, and the same machinery handles offset surfaces, open boundaries, and non-manifold geometry.
  • The 2D restriction of the algorithm uniformly samples planar implicit outlines, which the paper connects to vector-graphics and font-generation tasks.

Reading between the lines

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

  • Because the theorem requires only $k$-equidistributed rays rather than statistically independent ones, deterministic low-discrepancy ray sets should inherit the uniformity guarantee, and the paper's convergence experiments already hint that such sets improve moment-estimation rates.
  • The intersection and chord statistics that feed the uniform samples also give an unbiased area and volume estimator whose cost scales with surface area, suggesting an unexplored use as a differentiable surrogate loss for shape optimization or reconstruction where ray generation is independent of surface parameters.
  • For neural fields whose global Lipschitz constant is only estimated, uniformity in practice depends on the supplied bound being a true upper bound; coupling sphere tracing with conservative interval bounds or rigorous Lipschitz estimation would make the numeric sampler match the theoretical guarantee.
  • The same line-intersection idea should extend to planar intersections of a surface to sample curves, or to higher-dimensional varieties, though average chord-length formulas imply efficiency falls as dimension increases.
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

2 major / 4 minor

Summary. The paper proposes a method for uniformly sampling points on implicit surfaces by sampling random rays through a bounding box and collecting every intersection of each ray with the zero level set. The authors justify uniformity via Cauchy-Crofton integral geometry, present a sphere-tracing modification to find all intersections under a Lipschitz bound, and validate the method against meshes, rejection sampling, Hamiltonian Monte Carlo, and marching-cubes-based sampling. They further demonstrate extensions to blue noise, stratified sampling via sparse voxels, moment estimation, curvature-based resampling, and neural implicit deformation.

Significance. The core observation is correct and useful: if the ray-intersection subroutine is exact and the ray distribution is the standard isotropic one, the collected points are uniform in the sense of equidistribution, following established Cauchy-Crofton theory. The paper contributes a self-contained proof of the 1-equidistribution of the resampling view in Appendix A, a practical sphere-marching variant to collect all intersections, and a broad set of applications. The experimental comparison on 114 meshes against a ground-truth uniform sampler is solid both in uniformity (total variation score) and in cost (number of implicit function evaluations). The main weaknesses are in the specification of the algorithm and the unqualified guarantee for neural implicits whose Lipschitz bounds are only estimated.

major comments (2)
  1. [Section 4.1-4.2, Algorithm 1 and Algorithm 2] The pseudocode does not implement the full-line intersection process required by the theory. In Algorithm 1, origins o_i are sampled in the plane through the origin normal to d_i; in Algorithm 2, tracing starts at t=0 at o_i and only advances t upward until t<length(ℓ_i). For a line whose intersection with the bounding box straddles that plane, only the forward half of the line is traced, so intersections on the backward half are never recorded. Moreover, the update p ← o_i + t d_i/λ together with t ← t + s makes the termination test dimensionally inconsistent with the parameterization of ℓ_i in Algorithm 1. Please specify the bounding-box chord interval [t_min, t_max] explicitly and trace the entire interval, or move the ray origins to a plane outside the bounding box; in either case, reconcile the step-size parameterization with the ray parameter used in the termination test.
  2. [Abstract, Section 4.2-4.3] The uniformity guarantee is stated without qualification, but Algorithm 2's sphere tracing is safe only if λ is a true upper Lipschitz bound of f. As the paper itself notes in Section 4.3, neural implicits such as SIREN and NeuS 'encourage SDF-ness but do not guarantee it and may have large global Lipschitz bounds.' For those headline examples λ is estimated rather than certified, and an underestimate can cause missed zero crossings and biased samples. Please scope the guarantee to exact ray intersection or to a certified Lipschitz bound, or provide empirical evidence on neural fields with known mesh ground truth that the estimated λ is sufficient in the tested regimes.
minor comments (4)
  1. [Section 6.4] The sentence 'subsample our uniform a samples as blue noise' appears to contain a typo; it should read 'subsample our uniform samples as blue noise.'
  2. [Appendix A] The proof relies on the limit ∑ k_i/M from [PPK16], yet the footnote states that the proof in [PPK16] has an error and omits a factor of 2/|L|. Since the corrected proof is said to follow the same argument, it would be helpful to include that corrected derivation explicitly rather than leaving the reader to reconstruct it.
  3. [Figure 12 and Section 5] The neural field examples in Figure 12 are qualitative; the quantitative TV evaluations in Section 5 are all on mesh-derived signed distance functions, where λ is known. A quantitative uniformity measurement on at least one neural implicit with an estimated Lipschitz bound would make the practical claim more convincing.
  4. [Algorithm 1] The text describes a '2√3-sided square'; this is the side length of the square, not its diagonal. The wording could be clarified to avoid ambiguity about the sampling domain.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the uniformity guarantee is derived from an external integral-geometry theorem and is not equivalent to the method's inputs.

full rationale

The derivation chain is not circular. The central uniformity claim (Sections 3.2 and Appendix A, Theorem 1) reduces, via the Cauchy-Crofton formula, to an external integral-geometry identity: an integral over the surface equals an integral over the line process summing over all intersections. The paper states this pedigree explicitly ('This relationship is well-known in mathematics', citing [PPK16] and [San76]) and supplies a self-contained proof in Appendix A using resampled importance sampling, with the requirement that all ray intersections are found. The cited [PPK16] is not authored by the present authors, so there is no load-bearing self-citation; indeed Appendix A corrects an error in [PPK16]'s proof, so the paper is not merely importing the result. No parameter is fitted to data and then renamed a prediction: lambda is a user-supplied Lipschitz bound, a precondition of sphere tracing, and uniformity is proven for the exact-intersection subroutine regardless of lambda. The practical caveat that neural fields only 'encourage SDF-ness but do not guarantee it and may have large global Lipschitz bounds' concerns whether sphere tracing actually recovers all intersections for that representation; it does not make the mathematical statement equivalent to its inputs. The empirical TV evaluations use a mesh-based ground-truth sampler independent of the method, providing external validation rather than circular confirmation.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The method does not introduce new physical or mathematical entities. Its central claim rests on classical integral geometry plus two practical assumptions: a known Lipschitz bound and the ability to find every ray intersection within a tolerance. The only hand-set number is the intersection tolerance epsilon.

free parameters (1)
  • intersection tolerance epsilon = 1e-4
    Algorithm 2 requires epsilon as the threshold for declaring a ray-surface intersection. It is a hand-set numerical tolerance, not fitted to data; it moves samples off the exact surface into a slab of width about 2*epsilon/lambda, so exact uniformity is an idealization.
assumptions (5)
  • standard math Cauchy-Crofton / integral geometry: the integral of a function over a surface S equals (1/(2 pi)) times the integral over all lines of the sum of the function over all intersections of each line with S
    Invoked in Section 3.2 and used in the Appendix A proof via the extension from [PPK16].
  • domain assumption Algorithm 1 samples from the uniform measure on the set of oriented lines intersecting the bounding box
    The uniformity guarantee depends on the ray distribution being the canonical line measure; the paper argues this is achieved by uniform direction sampling plus rejection sampling on the projected square.
  • domain assumption A valid global Lipschitz bound lambda of f is known
    Section 4.2 states the method requires a known Lipschitz bound so sphere tracing can step by |f(p)|/lambda without skipping zero crossings.
  • domain assumption The surface is contained in the user-provided bounding box
    Algorithm 1 samples rays that intersect the box; rays that miss the box are discarded, so the surface must lie inside the box for all samples to be captured.
  • standard math The surface is a sufficiently regular codimension-one set for the Cauchy-Crofton formula to apply
    The proof in Appendix A assumes finite-area, rectifiable surfaces; practical implicit surfaces in the paper satisfy this.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Uniform Sampling of Surfaces by Casting Rays." pith.science (2026). https://pith.science/paper/GZT5JGFC

@misc{pith2026250605268,
  author       = {Pith},
  title        = {Pith review of: Uniform Sampling of Surfaces by Casting Rays},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GZT5JGFC}},
  note         = {Machine review of arXiv:2506.05268}
}
read the original abstract

Randomly sampling points on surfaces is an essential operation in geometry processing. This sampling is computationally straightforward on explicit meshes, but it is much more difficult on other shape representations, such as widely-used implicit surfaces. This work studies a simple and general scheme for sampling points on a surface, which is derived from a connection to the intersections of random rays with the surface. Concretely, given a subroutine to cast a ray against a surface and find all intersections, we can use that subroutine to uniformly sample white noise points on the surface. This approach is particularly effective in the context of implicit signed distance functions, where sphere marching allows us to efficiently cast rays and sample points, without needing to extract an intermediate mesh. We analyze the basic method to show that it guarantees uniformity, and find experimentally that it is significantly more efficient than alternative strategies on a variety of representations. Furthermore, we show extensions to blue noise sampling and stratified sampling, and applications to deform neural implicit surfaces as well as moment estimation.

Figures

Figures reproduced from arXiv: 2506.05268 by the authors.

Figure 1
Figure 1. Implicit surfaces represent the geometry of a shape as the zero level set of a function of 3D space. There are many varieties of implicit surfaces used in practice (left). We study a method for a fundamental surface operation: uniformly random point sampling. Any point on the continuous surface is equally likely to be chosen through our proposed process of sampling random rays in space and finding intersections with… view at source ↗
Figure 2
Figure 2. White noise sampling is often useful as a raw ingre￾dient to downstream applications, such as blue noise generators (e.g., [Yuk15, Bri07]. Our white noise samples on a neural im￾plicit [WLL∗ 21] (top left) are subsampled to a blue noise sample set [Yuk15]. For reference, we show the same process on a periodic square and its corresponding power spectrum images [SD11]. 2. Related Work Sampling surfaces is a core opera… view at source ↗
Figure 3
Figure 3. We visualize 500,000 samples using our method and all baseline methods on the level set of a neural implicit function [WLL∗ 21] learned from images of the ficus scene in the Blender dataset. Our method achieves similar or better sample quality as rejection sampling and sampling via marching cubes on a grid of resolution 40963 while being a factor of 57× and 296× cheaper in terms of function evaluations, respectively… view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: A neural field is learned to represent a smooth interpolation between two different chairs at t = 0 and t = 1 as in [LWJ∗ 22] (top). Our method uniformly samples the implicit surface of the interpolated shape at any point in the sequence (bottom). bounded sample counts…
Figure 6
Figure 6. Figure 6: Ray origin sampling is crucial for obtaining uniform ray samples. If origins are simply chosen as random points within the box (left), the rays will pass through the center of the box more frequently than the edges (middle left). Meanwhile, if the origins are chosen in…
Figure 5
Figure 5. Figure 5: An illustration of our uniform ray sampling algorithm as in Algorithm 1. Given the blue [−1,1] 3 cube around an implicitly￾defined teapot shape, we first sample a direction d⃗ i uniformly. Among all rays parallel to that direction, we only want the rays that intersect …
Figure 9
Figure 9. Figure 9: Sparse voxel structures can accelerate our method and act as a form of stratified sampling which reduces variance for es￾timations of shape quantities as described in Section 6.2. Here we show one instance of such a voxel structure with a voxel grid reso￾lution of 16, …
Figure 8
Figure 8. Figure 8: Our method only requires the surface to define a codimension-one shape as the level set of an implicit function. Non￾manifold junctures and open boundaries are no issue for our white noise sampling, such as this sailboat modeled as a cascade of ana￾lytic unsigned dista…
Figure 10
Figure 10. Figure 10: We compare our method with the specialized Hamil￾tonian Monte Carlo sampling algorithm proposed in [Chi22] and rejection sampling, by drawing an increasing number of samples from 5,000 to 50,000 on a subset of 8 different implicit functions defined by meshes from a da…
Figure 11
Figure 11. Figure 11: We ran our method on 114 different implicit surface functions from a dataset of meshes [MPZ14]. Here we plot the re￾lationship between ground truth surface area and the ratio between number of samples and the number of casted rays (i.e., average in￾tersection count). …
Figure 12
Figure 12. Figure 12: We apply our method to a variety of implicit functions. From left to right: a Möbius transformation of the “standard triple bubble” from [Sul12], a swept unsigned distance function resulting in a non-manifold shape, a surface of revolution generated by a cubic Bézier …
Figure 13
Figure 13. Figure 13: We compare our method with sampling a mesh extracted via marching cubes, from grids with resolutions ranging from 2003 to 10243 . With this particular shape, even though the uniformity of the sample set via marching cubes improves with increasing grid resolution, the …
Figure 14
Figure 14. Figure 14: White noise sampling can be straightforward with mesh surfaces, but offset surface sampling is often non-trivial unless con￾verted to an implicit surface. Our method can easily sample offset surfaces, along both positive and negative directions, in addition to the sur…
Figure 15
Figure 15. Figure 15: Our method can estimate various shape quantities like surface area, volume, surface centroids, and volumetric centroids, as demonstrated here for this PseudoSDF ShaderToy example of a burger as seen in Figure . ShaderToy PseudoSDF credit to ©Xor (CC BY-NC-SA 3.0) by t…
Figure 16
Figure 16. Figure 16: We plot the absolute surface area estimation error with samples acquired from sampled rays using a low-discrepancy se￾quence as proposed in [LYZ∗ 06, LYZ∗ 10], as well as uniformly sampled rays, as described in Section 4.1. We also plot asymptotes O(N 2 3 ) and O(N 1 …
Figure 17
Figure 17. Figure 17: We ran our method to sample points with 50,000 rays on 114 shapes from a dataset of meshes [MPZ14], and plotted the total number of evaluations required and the ground truth surface area. function evaluation cost of our method and the surface area of the shape being s…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

57 extracted references · 49 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry.original add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 i...

  3. [3]

    : Controlling neural level sets

    Atzmon M., Haim N., Yariv L., Israelov O., Maron H., Lipman Y. : Controlling neural level sets. Advances in Neural Information Processing Systems 32 (2019)

  4. [4]

    : Random points on an algebraic manifold

    Breiding P., Marigliano O. : Random points on an algebraic manifold. SIAM Journal on Mathematics of Data Science 2, 3 (2020), 683--704

  5. [5]

    : Fast poisson disk sampling in arbitrary dimensions

    Bridson R. : Fast poisson disk sampling in arbitrary dimensions. SIGGRAPH sketches 10, 1 (2007), 1

  6. [6]

    : Uniform Sampling over Level Sets

    Chiu E. : Uniform Sampling over Level Sets. Master's thesis, Massachusetts Institute of Technology, 2022

  7. [7]

    : Learning gradient fields for shape generation

    Cai R., Yang G., Averbuch-Elor H., Hao Z., Belongie S., Snavely N., Hariharan B. : Learning gradient fields for shape generation. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part III 16 (2020), Springer, pp. 364--381

  8. [8]

    : Improving neural implicit surfaces geometry with patch warping

    Darmon F., Bascle B., Devaux J.-C., Monasse P., Aubry M. : Improving neural implicit surfaces geometry with patch warping. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022), pp. 6260--6269

Show all 57 references
  1. [9]

    A., Henning R., Johnson R

    Detwiler J. A., Henning R., Johnson R. A., Marino M. G. : A generic surface sampler for monte carlo simulations. IEEE Transactions on Nuclear Science 55, 4 (Aug. 2008), 2329–2333. URL: http://dx.doi.org/10.1109/TNS.2008.2001063, https://doi.org/10.1109/tns.2008.2001063 doi:10....

  2. [10]

    V., Laurynėnas A., Ivanauskas F

    Daugmaudis J. V., Laurynėnas A., Ivanauskas F. : The influence of bounding surface on the precision of the cauchy–crofton method. Lietuvos matematikos rinkinys 51, proc. LMS (Dec. 2010), 250–255. URL: https://www.journals.vu.lt/LMR/article/view/17823, https://doi.org/10.15388/...

  3. [11]

    Franklin J. N. : Deterministic simulation of random processes. Mathematics of Computation 17, 81 (1963), 28--59

  4. [12]

    : Segment tracing using local lipschitz bounds

    Galin E., Gu \'e rin E., Paris A., Peytavie A. : Segment tracing using local lipschitz bounds. In Computer Graphics Forum (2020), vol. 39, Wiley Online Library, pp. 545--554

  5. [13]

    : Ray tracing harmonic functions

    Gillespie M., Yang D., Botsch M., Crane K. : Ray tracing harmonic functions. ACM Trans. Graph 43, 4 (2024)

  6. [14]

    Hart J. C. : Sphere tracing: A geometric method for the antialiased ray tracing of implicit surfaces. The Visual Computer 12, 10 (1996), 527--545

  7. [15]

    L., Huang H

    Huang S.-S., Chen G., Heng C. L., Huang H. : Neuralindicator: implicit surface reconstruction from neural indicator priors. In Forty-first International Conference on Machine Learning (2024)

  8. [16]

    : Diffcd: A symmetric differentiable chamfer distance for neural implicit surface fitting

    H \"a renstam-Nielsen L., Sang L., Saroha A., Araslanov N., Cremers D. : Diffcd: A symmetric differentiable chamfer distance for neural implicit surface fitting. In European Conference on Computer Vision (2024), Springer, pp. 432--447

  9. [17]

    Khosravifard A., Hematiyan M. R. : A new method for meshless integration in 2d and 3d galerkin meshfree methods. Engineering Analysis with Boundary Elements 34, 1 (2010), 30--40

  10. [18]

    : Bssrdf importance sampling

    King A., Kulla C., Conty A., Fajardo M. : Bssrdf importance sampling. In ACM SIGGRAPH 2013 Talks. 2013, pp. 1--1

  11. [19]

    : 3d gaussian splatting for real-time radiance field rendering

    Kerbl B., Kopanas G., Leimk \"u hler T., Drettakis G. : 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph. 42, 4 (2023), 139--1

  12. [20]

    Knuth D. E. : The METAFONT book, vol. C of Computers and Typesetting. Addison-Wesley, Reading, Massachusetts, 1986

  13. [21]

    E., Cline H

    Lorensen W. E., Cline H. E. : Marching cubes: A high resolution 3d surface construction algorithm. In Seminal graphics: pioneering efforts that shaped the field. 1998, pp. 347--353

  14. [22]

    : Sampling with mollified interaction energy descent

    Li L., Liu Q., Korba A., Yurochkin M., Solomon J. : Sampling with mollified interaction energy descent. arXiv preprint arXiv:2210.13400 (2022)

  15. [23]

    D., Williams F., Jacobson A., Fidler S., Litany O

    Liu H.-T. D., Williams F., Jacobson A., Fidler S., Litany O. : Learning smooth neural functions via lipschitz regularization. In ACM SIGGRAPH 2022 Conference Proceedings (2022), pp. 1--13

  16. [24]

    R., Bowyer A

    Li X., Wang W., Martin R. R., Bowyer A. : Using low-discrepancy sequences and the crofton formula to compute surface areas of geometric models. Computer-Aided Design 35, 9 (2003), 771--782

  17. [25]

    : A quasi-monte carlo method for computing areas of point-sampled surfaces

    Liu Y.-S., Yong J.-H., Zhang H., Yan D.-M., Sun J.-G. : A quasi-monte carlo method for computing areas of point-sampled surfaces. Computer-Aided Design 38, 1 (2006), 55--68

  18. [26]

    : Surface area estimation of digitized 3d objects using quasi-monte carlo methods

    Liu Y.-S., Yi J., Zhang H., Zheng G.-Q., Paul J.-C. : Surface area estimation of digitized 3d objects using quasi-monte carlo methods. Pattern Recognition 43, 11 (2010), 3900--3909

  19. [27]

    : On the generalization of the average chord length

    Mazzolo A. : On the generalization of the average chord length. Annals of Nuclear Energy 35, 3 (2008), 503--506. URL: https://www.sciencedirect.com/science/article/pii/S0306454907001843, https://doi.org/https://doi.org/10.1016/j.anucene.2007.06.026 doi:https://doi.org/10.1016/...

  20. [28]

    : Instant neural graphics primitives with a multiresolution hash encoding

    M \"u ller T., Evans A., Schied C., Keller A. : Instant neural graphics primitives with a multiresolution hash encoding. ACM transactions on graphics (TOG) 41, 4 (2022), 1--15

  21. [29]

    : 3d gaussian ray tracing: Fast tracing of particle scenes

    Moenne-Loccoz N., Mirzaei A., Perel O., de Lutio R., Martinez Esturo J., State G., Fidler S., Sharp N., Gojcic Z. : 3d gaussian ray tracing: Fast tracing of particle scenes. ACM Transactions on Graphics (TOG) 43, 6 (2024), 1--19

  22. [30]

    : Robust field-aligned global parametrization

    Myles A., Pietroni N., Zorin D. : Robust field-aligned global parametrization. ACM Trans. Graph. 33, 4 (2014), 135--1

  23. [31]

    C., Alexa M., Gross M

    \"O ztireli A. C., Alexa M., Gross M. : Spectral sampling of manifolds. ACM Transactions on Graphics (TOG) 29, 6 (2010), 1--8

  24. [32]

    J., Florence P., Straub J., Newcombe R., Lovegrove S

    Park J. J., Florence P., Straub J., Newcombe R., Lovegrove S. : Deepsdf: Learning continuous signed distance functions for shape representation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (2019), pp. 165--174

  25. [33]

    : Pointclouds: Distributing points uniformly on a surface

    Palais R., Palais B., Karcher H. : Pointclouds: Distributing points uniformly on a surface. arXiv preprint arXiv:1611.04690 (2016)

  26. [34]

    Quinn J. A. : Low-discrepancy point sampling of 2D manifolds for visual computing. Cardiff University (United Kingdom), 2009

  27. [35]

    Rubin D. B. : The calculation of posterior distributions by data augmentation: Comment: A noniterative sampling/importance resampling alternative to the data augmentation algorithm for creating a few imputations when fractions of missing information are modest: The sir algorit...

  28. [36]

    : Point sampling with uniformly distributed lines

    Rovira J., Wonka P., Castro F., Sbert M. : Point sampling with uniformly distributed lines. In Proceedings Eurographics/IEEE VGTC Symposium Point-Based Graphics, 2005. (2005), IEEE, pp. 109--118

  29. [37]

    Reddy P., Zhang Z., Wang Z., Fisher M., Jin H., Mitra N. J. : A multi-implicit neural representation for fonts. In Advances in Neural Information Processing Systems 34: Annual Conference on Neural Information Processing Systems 2021, NeurIPS 2021, December 6-14, 2021, virtual ...

  30. [38]

    Santal \'o L. A. : Integral geometry and geometric probability. 1976

  31. [39]

    : Accurate spectral analysis of two-dimensional point sets

    Schl \" o mer T., Deussen O. : Accurate spectral analysis of two-dimensional point sets. J. Graphics, GPU, & Game Tools 15 , 3 (2011), 152--160. URL: https://doi.org/10.1080/2151237X.2011.609773, https://doi.org/10.1080/2151237X.2011.609773 doi:10.1080/2151237X.2011.609773

  32. [40]

    Stol J., De Figueiredo L. H. : Self-validated numerical methods and applications. In Monograph for 21st Brazilian Mathematics Colloquium, IMPA, Rio de Janeiro. Citeseer (1997), vol. 5, Citeseer, p. 122

  33. [41]

    Schaufler G., Jensen H. W. : Ray tracing point sampled geometry. In Rendering Techniques 2000: Proceedings of the Eurographics Workshop in Brno, Czech Republic, June 26--28, 2000 11 (2000), Springer, pp. 319--328

  34. [42]

    : Stochastic poisson surface reconstruction

    Sell \' a n S., Jacobson A. : Stochastic poisson surface reconstruction. ACM Trans. Graph. 41 , 6 (2022), 227:1--227:12. URL: https://doi.org/10.1145/3550454.3555441, https://doi.org/10.1145/3550454.3555441 doi:10.1145/3550454.3555441

  35. [43]

    : Implicit neural representations with periodic activation functions

    Sitzmann V., Martel J., Bergman A., Lindell D., Wetzstein G. : Implicit neural representations with periodic activation functions. Advances in neural information processing systems 33 (2020), 7462--7473

  36. [44]

    Sullivan J. M. : Pleasing shapes for topological objects. In Mathematics and Modern Art: Proceedings of the First ESMA Conference, held in Paris, July 19-22, 2010 (2012), Springer, pp. 153--165

  37. [45]

    Talbot J. F. : Importance resampling for global illumination. Brigham Young University, 2005

  38. [46]

    Trusty T., Chen H., Levin D. I. : The shape matching element method: Direct animation of curved surface models. ACM Transactions on Graphics (TOG) 40, 4 (2021), 1--14

  39. [47]

    : Vecfusion: Vector font generation with diffusion

    Thamizharasan V., Liu D., Agarwal S., Fisher M., Gharbi M., Wang O., Jacobson A., Kalogerakis E. : Vecfusion: Vector font generation with diffusion. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024 (2024), IEEE , ...

  40. [48]

    : Computation of intersection volume using discrete quadrature algorithm

    Tong J., Li Q., Peng W., Fang Y., Wu Y. : Computation of intersection volume using discrete quadrature algorithm. Optik 158 (2018), 725--736

  41. [49]

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

    Wang P., Liu L., Liu Y., Theobalt C., Komura T., Wang W. : Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. arXiv preprint arXiv:2106.10689 (2021)

  42. [50]

    : Neural fields in visual computing and beyond

    Xie Y., Takikawa T., Saito S., Litany O., Yan S., Khan N., Tombari F., Tompkin J., Sitzmann V., Sridhar S. : Neural fields in visual computing and beyond. CoRR abs/2111.11426 (2021). URL: https://arxiv.org/abs/2111.11426, http://arxiv.org/abs/2111.11426 arXiv:2111.11426

  43. [51]

    : Geometry processing with neural fields

    Yang G., Belongie S., Hariharan B., Koltun V. : Geometry processing with neural fields. Advances in Neural Information Processing Systems 34 (2021), 22483--22497

  44. [52]

    : Omnisvg: A unified scalable vector graphics generation model, 2025

    Yang Y., Cheng W., Chen S., Zeng X., Zhang J., Wang L., Yu G., Ma X., Jiang Y.-G. : Omnisvg: A unified scalable vector graphics generation model, 2025. URL: https://arxiv.org/abs/2504.06263, http://arxiv.org/abs/2504.06263 arXiv:2504.06263

  45. [53]

    C., Denny J., Lindsey A., Thomas S., Amato N

    Yeh H.-Y. C., Denny J., Lindsey A., Thomas S., Amato N. M. : Umaprm: Uniformly sampling the medial axis. In 2014 IEEE International Conference on Robotics and Automation (ICRA) (2014), IEEE, pp. 5798--5803

  46. [54]

    : Volume rendering of neural implicit surfaces

    Yariv L., Gu J., Kasten Y., Lipman Y. : Volume rendering of neural implicit surfaces. Advances in Neural Information Processing Systems 34 (2021), 4805--4815

  47. [55]

    : Sample elimination for generating poisson disk sample sets

    Yuksel C. : Sample elimination for generating poisson disk sample sets. In Computer Graphics Forum (2015), vol. 34, Wiley Online Library, pp. 25--32

  48. [56]

    : Iso-points: Optimizing neural implicit surfaces with hybrid representations

    Yifan W., Wu S., Oztireli C., Sorkine-Hornung O. : Iso-points: Optimizing neural implicit surfaces with hybrid representations. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2021), pp. 374--383

  49. [57]

    : Unbiased sampling and meshing of isosurfaces

    Yan D.-M., Wallner J., Wonka P. : Unbiased sampling and meshing of isosurfaces. IEEE transactions on visualization and computer graphics 20, 11 (2014), 1579--1589

Pith tools

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