Pith. sign in

REVIEW 2 major objections 5 minor 37 references

A single trained model of Fourier-boundary surfels can be rendered at continuous levels of detail by truncating coefficients at runtime.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 21:53 UTC pith:S3H7UFZG

load-bearing objection Solid planar-splatting paper: Fourier-boundary surfels give real runtime LoD by coefficient truncation, with clean ablations and SOTA-among-planar numbers. the 2 major comments →

arxiv 2603.19834 v3 pith:S3H7UFZG submitted 2026-03-20 cs.CV

Fourier Splatting: Generalized Fourier encoded primitives for scalable radiance fields

Mihnea-Bogdan Jurca , Bert Van hauwermeiren , Adrian Munteanu This is my paper
classification cs.CV
keywords novel view synthesisradiance fieldssplattingFourier primitiveslevel of detailplanar surfelsscalable rendering
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Existing real-time radiance-field methods improve quality only by adding more primitives and reduce quality only by pruning them. This paper claims that the primitive itself can be made scalable: each planar surfel is given a closed boundary described by a Fourier series whose coefficients control shape complexity. At the lowest order the primitive is a disc; each extra frequency lets it approximate more complex closed curves. Because the Fourier basis is orthogonal, discarding higher coefficients at render time yields a coarser but still valid shape without retraining. The authors show that, with a straight-through estimator to keep gradients alive outside the hard boundary and a densification routine that splits multi-lobed primitives, the representation reaches state-of-the-art quality among planar methods and stays competitive with volumetric ones, while quality degrades more gracefully under coefficient truncation than under primitive pruning.

Core claim

Fourier Splatting is the first inherently scalable primitive for real-time radiance fields: planar surfels whose boundaries are Fourier-encoded can be trained once and then rendered at any continuous level of detail simply by truncating the active frequency coefficients, decoupling reconstruction fidelity from primitive count.

What carries the argument

The Fourier-encoded boundary (Eq. 4): a closed polar curve whose radius is the modulus of a truncated Fourier polynomial of the polar angle, normalized so the curve never exceeds a learned circumradius. Truncation of higher coefficients is the LoD operator; a straight-through estimator and the HYDRA lobe-decomposition densifier keep the coefficients learnable under a hard power-window opacity.

Load-bearing premise

The straight-through estimator and the learned lobe-decomposition densifier must actually unlock complex, non-circular boundaries; if they fail, the method collapses back to ordinary disc primitives whose only scalability is pruning.

What would settle it

Train a full-K model, then measure image metrics and visual quality while successively truncating coefficients to K=1; if quality does not degrade more gracefully than an equal-budget Octree-GS or pruning baseline, or if the learned boundaries remain near-circular, the central scalability claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces Fourier Splatting, a planar surfel primitive whose boundary is parameterized by a Fourier series (Eq. 4) with amplitude normalization (Eq. 5) and a power-window opacity (Eq. 7). A single trained model can be rendered at continuous levels of detail by truncating coefficients at runtime. Optimization uses a straight-through estimator (Eq. 8) to supply exterior gradients and HYDRA, an MCMC-compatible densification procedure that either geometrically splits or MLP-decomposes multi-lobed primitives. On Mip-NeRF 360 and Tanks & Temples the method reports state-of-the-art metrics among planar primitives and competitive LPIPS versus volumetric baselines; ablations isolate the STE and HYDRA contributions, and qualitative/quantitative LoD curves illustrate graceful degradation under truncation.

Significance. If the empirical claims hold, the work supplies the first radiance-field primitive whose expressiveness itself is continuously scalable, decoupling fidelity from primitive count. This is a genuine conceptual advance over pruning- or hierarchy-based LoD schemes and is directly useful for bandwidth-constrained delivery. The paper supplies concrete supporting evidence: public-benchmark tables under a shared protocol, component ablations (Table 2), DTU surface metrics in the supplement, and explicit truncation experiments (Figs. 5 and 7). The Fourier orthonormality argument for truncation is classical and parameter-free, strengthening the scalability claim.

major comments (2)
  1. §5.3 and Fig. 7 demonstrate monotonic quality loss under coefficient truncation for the authors’ own model, and Fig. 5 offers a qualitative side-by-side with Octree-GS. However, no quantitative rate–distortion comparison (PSNR/LPIPS versus bits or versus equal primitive count) is provided against any pruning- or hierarchy-based baseline. Without such a curve the claim of “more graceful” degradation remains only partially substantiated; a single equal-budget or equal-bitrate plot would make the central scalability advantage load-bearing rather than suggestive.
  2. Table 3 reports arithmetic operations per primitive but the manuscript never states measured FPS or peak memory on the same hardware used for the quality tables. Because the abstract and conclusion advertise a “versatile solution for bandwidth-constrained high-fidelity rendering,” the absence of wall-clock numbers leaves open whether the extra 8 ops per frequency (and the STE/HYDRA machinery) remain real-time once K > 1. A short runtime column or paragraph would close this gap.
minor comments (5)
  1. Several concatenated words appear in the supplied text (“fidelitystrictlytothenumber”, “density-drivenoptimization”, etc.); these are presumably paste artifacts but should be cleaned for the camera-ready version.
  2. Eq. (8) introduces β and γ without stating that they are frozen at 3.0 / 0.5 until the supplement; a one-sentence note in the main text would help reproducibility.
  3. The HYDRA MLP architecture (2×2048) and its training objective are described only in the supplement; a brief pointer in §4.3 would improve self-containment.
  4. Fig. 1 caption claims “progressively improved rendering quality” but the figure itself shows only primitive shapes; a small inset of the corresponding rendered patch would make the visual argument clearer.
  5. In the related-work discussion of surface primitives, BBSplat is cited for textured billboards; a short remark on how Fourier boundaries differ from learned alpha masks would sharpen the positioning.

Circularity Check

0 steps flagged

No significant circularity: empirical methods paper whose scalability claim follows from classical Fourier orthonormality and is validated on external held-out benchmarks.

full rationale

Fourier Splatting is a standard empirical computer-vision methods contribution. The core primitive (planar surfel whose boundary is a truncated Fourier series, Eqs. 4–5) is defined once; the claimed runtime LoD property is an immediate consequence of the classical fact that the Fourier basis is orthonormal, so the optimal L2 approximation of a K-term polynomial by a (K-1)-term polynomial is simple coefficient truncation (explicitly stated in Sec. 5.1). No parameter is fitted to a target metric and then re-presented as a prediction; the STE surrogate (Eq. 8) and HYDRA MLP are optimization devices whose efficacy is checked by ablation (Table 2) rather than assumed by definition. All quantitative claims (Table 1, Figs. 5–7) are measured against public held-out views and published baselines. Self-citations are ordinary community references (3DGS, 2DGS, MCMC densification, etc.) and do not form a load-bearing uniqueness or uniqueness-theorem chain. Consequently the derivation chain never reduces a claimed result to its own inputs by construction.

Axiom & Free-Parameter Ledger

6 free parameters · 4 axioms · 3 invented entities

The central claims rest on classical Fourier analysis of closed curves, the 2DGS surfel and ray-plane intersection machinery, the 3DGS-MCMC densification framework, and a collection of hand-chosen hyper-parameters and two invented algorithmic components (STE surrogate and HYDRA MLP). No new physical entities are postulated; the free parameters are ordinary optimization and architecture knobs.

free parameters (6)
  • Number of Fourier frequencies K = 6
    Fixed to K=6 for all experiments; controls maximum shape complexity and parameter count (70 params/primitive).
  • STE hyperparameters β, γ = β=3.0, γ=0.5
    Hard-coded in the CUDA kernel to control the width and strength of the exterior gradient; chosen by hand, not derived.
  • Death and densification thresholds (τo, τI, Vmin, τg, screen-extent cutoffs) = τo=0.005, τI=0.04, Vmin=2, τg=0.0004, etc.
    Multiple numeric thresholds that decide which primitives die or split; listed in Tables 2–3 of the supplement and tuned per scene type.
  • Learning rates and loss weights (λ, λdist, λnormal, opacity LR, etc.) = various (see Tables 2–3)
    Scene-type-specific Adam schedules and regularizer weights that directly affect final PSNR/SSIM.
  • Per-scene primitive budgets Nmax = 2.1M–6.4M
    Hard caps (2.1 M–6.4 M) that determine capacity; taken from Triangle Splatting protocol but still free choices.
  • HYDRA MLP architecture and training objective = 2×2048, SSIM + regularization
    2-hidden-layer 2048-unit MLP trained to maximize SSIM of child vs. parent lobe; architecture and data mixture are design choices.
axioms (4)
  • standard math A closed planar curve can be represented to arbitrary accuracy by a Fourier series in polar angle, and truncation yields the optimal L2 approximation of lower order.
    Invoked in Sec. 4.1 and the frequency-schedule paragraph to justify both expressiveness and runtime LoD by coefficient dropping.
  • domain assumption Oriented planar surfels with ray-plane intersection and front-to-back alpha compositing (inherited from 2DGS) are a valid and efficient representation for radiance fields.
    Sec. 3.2 and 4; the entire method builds on this established surfel pipeline.
  • domain assumption MCMC densification that preserves the target distribution P(g) via mass-preserving opacity and scale updates remains valid when the kernel is a power window rather than a Gaussian.
    Sec. 4.3 derives the new integral factors (Eqs. 9–10) under this assumption.
  • ad hoc to paper A straight-through estimator that replaces the hard power window by a smooth surrogate only in the backward pass yields unbiased-enough gradients for the Fourier coefficients.
    Sec. 4.2, Eq. 8; the particular surrogate form and the restriction of exterior gradients to coefficients only are paper-specific design choices.
invented entities (3)
  • Fourier-encoded planar primitive (Fourier Splatting surfel) no independent evidence
    purpose: Provide an arbitrary closed boundary whose complexity is controlled by the number of retained frequencies, enabling inherent LoD.
    Core contribution of Sec. 4.1; no independent existence outside this representation.
  • HYDRA (Hybrid Decomposition for Rendering-Aware Preservation) no independent evidence
    purpose: Split complex multi-lobed primitives into simpler children while approximately preserving the MCMC target distribution.
    Sec. 4.3; learned MLP that maps parent Fourier coefficients and lobe bounds to child parameters.
  • STE power-window surrogate for exterior gradients no independent evidence
    purpose: Allow Fourier coefficients to receive gradient signal from pixels outside the hard boundary so that shapes can expand and become non-circular.
    Eq. 8 and Fig. 3; specific functional form is introduced for this paper.

pith-pipeline@v1.1.0-grok45 · 22856 in / 3647 out tokens · 39136 ms · 2026-07-13T21:53:48.733952+00:00 · methodology

0 comments
read the original abstract

Novel view synthesis has recently been revolutionized by 3D Gaussian Splatting (3DGS), which enables real-time rendering through explicit primitive rasterization. However, existing methods tie visual fidelity strictly to the number of primitives: quality downscaling is achieved only through pruning primitives. We propose the first inherently scalable primitive for radiance field rendering. Fourier Splatting employs scalable primitives with arbitrary closed shapes obtained by parameterizing planar surfels with Fourier encoded descriptors. This formulation allows a single trained model to be rendered at varying levels of detail simply by truncating Fourier coefficients at runtime. To facilitate stable optimization, we employ a straight-through estimator for gradient extension beyond the primitive boundary, and introduce HYDRA, a densification strategy that decomposes complex primitives into simpler constituents within the MCMC framework. Our method achieves state-of-the-art rendering quality among planar-primitive frameworks and comparable perceptual metrics compared to leading volumetric representations on standard benchmarks, providing a versatile solution for bandwidth-constrained high-fidelity rendering.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 6 canonical work pages · 1 internal anchor

  1. [1]

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Mip- NeRF360:UnboundedAnti-AliasedNeuralRadianceFields.pp.5470–5479(2022), https://openaccess.thecvf.com/content/CVPR2022/html/Barron_Mip- NeRF_ 360 _ Unbounded _ Anti - Aliased _ Neural _ Radiance _ Fields _ CVPR _ 2022 _ paper . html

    2022

  2. [2]

    Barron, J.T., Mildenhall, B., Verbin, D., Srinivasan, P.P., Hedman, P.: Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields. pp. 19697–19705 (2023),https: //openaccess.thecvf.com/content/ICCV2023/html/Barron_Zip- NeRF_Anti- Aliased_Grid-Based_Neural_Radiance_Fields_ICCV_2023_paper.html

    2023

  3. [3]

    Bengio, Y., Léonard, N., Courville, A.: Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation (Aug 2013).https: //doi.org/10.48550/arXiv.1308.3432,http://arxiv.org/abs/1308.3432, arXiv:1308.3432 [cs]

    Pith/arXiv arXiv doi:10.48550/arxiv.1308.3432 2013

  4. [4]

    In: ACM SIGGRAPH 2024 Conference Pa- pers

    Dai, P., Xu, J., Xie, W., Liu, X., Wang, H., Xu, W.: High-quality Surface Re- construction using Gaussian Surfels. In: ACM SIGGRAPH 2024 Conference Pa- pers. pp. 1–11. SIGGRAPH ’24, Association for Computing Machinery, New York, NY, USA (Jul 2024).https://doi.org/10.1145/3641519.3657441,https: //dl.acm.org/doi/10.1145/3641519.3657441

    doi:10.1145/3641519.3657441 2024

  5. [5]

    Hamdi, A., Melas-Kyriazi, L., Mai, J., Qian, G., Liu, R., Vondrick, C., Ghanem, B., Vedaldi, A.: GES : Generalized Exponential Splatting for Efficient Radiance Field Rendering. pp. 19812–19822 (2024),https://openaccess.thecvf.com/ content / CVPR2024 / html / Hamdi _ GES _ _Generalized _ Exponential _ Splatting _ for_Efficient_Radiance_Field_Rendering_CVPR...

    2024

  6. [6]

    2512.06818,http://arxiv.org/abs/2512.06818, arXiv:2512.06818 [cs]

    Held, J., Son, S., Vandeghen, R., Rebain, D., Gadelha, M., Zhou, Y., Cioppa, A., Lin, M.C., Droogenbroeck, M.V., Tagliasacchi, A.: MeshSplatting: Differentiable Rendering with Opaque Meshes (Dec 2025).https://doi.org/10.48550/arXiv. 2512.06818,http://arxiv.org/abs/2512.06818, arXiv:2512.06818 [cs]

    doi:10.48550/arxiv 2025

  7. [7]

    Held, J., Vandeghen, R., Deliege, A., Hamdi, A., Giancola, S., Cioppa, A., Vedaldi, A., Ghanem, B., Tagliasacchi, A., Droogenbroeck, M.V.: Triangle Splatting for Real-Time Radiance Field Rendering (May 2025).https://doi.org/10.48550/ arXiv.2505.19175,http://arxiv.org/abs/2505.19175, arXiv:2505.19175 [cs]

    Pith/arXiv arXiv 2025

  8. [8]

    Held, J., Vandeghen, R., Hamdi, A., Deliege, A., Cioppa, A., Giancola, S., Vedaldi, A., Ghanem, B., Van Droogenbroeck, M.: 3D Convex Splatting: Radiance Field Rendering with 3D Smooth Convexes. pp. 21360–21369 (2025),https:// openaccess.thecvf.com/content/CVPR2025/html/Held_3D_Convex_Splatting_ Radiance_Field_Rendering_with_3D_Smooth_Convexes_CVPR_2025_paper.html

    2025

  9. [9]

    In: ACM SIGGRAPH 2024 Confer- ence Papers

    Huang, B., Yu, Z., Chen, A., Geiger, A., Gao, S.: 2D Gaussian Splatting for Geometrically Accurate Radiance Fields. In: ACM SIGGRAPH 2024 Confer- ence Papers. pp. 1–11. SIGGRAPH ’24, Association for Computing Machinery, 16 Jurca et al. New York, NY, USA (Jul 2024).https://doi.org/10.1145/3641519.3657428, https://dl.acm.org/doi/10.1145/3641519.3657428

    doi:10.1145/3641519.3657428 2024

  10. [10]

    Huang, Y.H., Lin, M.X., Sun, Y.T., Yang, Z., Lyu, X., Cao, Y.P., Qi, X.: Deformable Radial Kernel Splatting. pp. 21513–21523 (2025),https : / / openaccess.thecvf.com/content/CVPR2025/html/Huang_Deformable_Radial_ Kernel_Splatting_CVPR_2025_paper.html

    2025

  11. [11]

    In: 2014 IEEE Conference on Computer Vision and Pattern Recognition

    Jensen, R., Dahl, A., Vogiatzis, G., Tola, E., Aanæs, H.: Large Scale Multi-view Stereopsis Evaluation. In: 2014 IEEE Conference on Computer Vision and Pattern Recognition. pp. 406–413 (Jun 2014).https://doi.org/10.1109/CVPR.2014.59, https://ieeexplore.ieee.org/document/6909453, iSSN: 1063-6919

    work page doi:10.1109/cvpr.2014.59 2014

  12. [12]

    Jurca, M.B., Royen, R., Giosan, I., Munteanu, A.: RT-GS2: Real-Time Generaliz- able Semantic Segmentation for 3D Gaussian Representations of Radiance Fields (Aug 2024).https://doi.org/10.48550/arXiv.2405.18033,http://arxiv.org/ abs/2405.18033, arXiv:2405.18033 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2405.18033 2024

  13. [13]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkuehler, T., Drettakis, G.: 3D Gaussian Splatting for Real-Time Radiance Field Rendering. ACM Trans. Graph.42(4), 139:1–139:14 (Jul 2023).https://doi.org/10.1145/3592433,https://dl.acm.org/doi/10. 1145/3592433

    doi:10.1145/3592433 2023

  14. [14]

    Advances in Neural Information Processing Systems37, 80965–80986 (Dec 2024)

    Kheradmand, S., Rebain, D., Sharma, G., Sun, W., Tseng, Y.C., Isack, H., Kar, A., Tagliasacchi, A., Yi, K.M.: 3D Gaussian Splatting as Markov Chain Monte Carlo. Advances in Neural Information Processing Systems37, 80965–80986 (Dec 2024). https : / / doi . org / 10 . 52202 / 079017 - 2573,https : / / proceedings . neurips . cc / paper _ files / paper / 202...

    2024

  15. [15]

    ACM Trans

    Knapitsch, A., Park, J., Zhou, Q.Y., Koltun, V.: Tanks and temples: benchmark- ing large-scale scene reconstruction. ACM Trans. Graph.36(4), 78:1–78:13 (Jul 2017).https://doi.org/10.1145/3072959.3073599,https://dl.acm.org/doi/ 10.1145/3072959.3073599

    doi:10.1145/3072959.3073599 2017

  16. [16]

    Kulhanek, J., Rakotosaona, M.J., Manhardt, F., Tsalicoglou, C., Niemeyer, M., Sattler, T., Peng, S., Tombari, F.: LODGE: Level-of-Detail Large-Scale Gaus- sian Splatting with Efficient Rendering (Oct 2025).https://doi.org/10.48550/ arXiv.2505.23158,http://arxiv.org/abs/2505.23158, arXiv:2505.23158 [cs]

    arXiv 2025

  17. [17]

    Li, H., Liu, J., Sznaier, M., Camps, O.: 3D-HGS: 3D Half-Gaussian Splatting. pp. 10996–11005 (2025),https://openaccess.thecvf.com/content/CVPR2025/html/ Li_3D-HGS_3D_Half-Gaussian_Splatting_CVPR_2025_paper.html

    2025

  18. [18]

    In: Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers

    Liu, R., Sun, D., Chen, M., Wang, Y., Feng, A.: Deformable Beta Splatting. In: Proceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers. pp. 1–11. SIGGRAPH Conference Pa- pers ’25, Association for Computing Machinery, New York, NY, USA (Jul 2025). https://doi.org/10.1145/3721238.3730716,https:/...

    doi:10.1145/3721238.3730716 2025

  19. [19]

    Lu, T., Yu, M., Xu, L., Xiangli, Y., Wang, L., Lin, D., Dai, B.: Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering. pp. 20654–20664 (2024), https://openaccess.thecvf.com/content/CVPR2024/html/Lu_Scaffold- GS_ Structured_3D_Gaussians_for_View- Adaptive_Rendering_CVPR_2024_paper. html

    2024

  20. [20]

    ACM65(1), 99–106 (Dec 2021).https://doi.org/10.1145/3503250,https:// dl.acm.org/doi/10.1145/3503250 Fourier Splatting 17

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.:NeRF:representingscenesasneuralradiancefieldsforviewsynthesis.Commun. ACM65(1), 99–106 (Dec 2021).https://doi.org/10.1145/3503250,https:// dl.acm.org/doi/10.1145/3503250 Fourier Splatting 17

    doi:10.1145/3503250 2021

  21. [21]

    ACM Trans

    Müller, T., Evans, A., Schied, C., Keller, A.: Instant neural graphics primitives with a multiresolution hash encoding. ACM Trans. Graph.41(4), 102:1–102:15 (Jul 2022).https://doi.org/10.1145/3528223.3530127,https://dl.acm.org/ doi/10.1145/3528223.3530127

    doi:10.1145/3528223.3530127 2022

  22. [22]

    Qin, M., Li, W., Zhou, J., Wang, H., Pfister, H.: LangSplat: 3D Language Gaussian Splatting. pp. 20051–20060 (2024),https://openaccess.thecvf.com/content/ CVPR2024/html/Qin_LangSplat_3D_Language_Gaussian_Splatting_CVPR_2024_ paper.html

    2024

  23. [23]

    IEEE Transactions on Pattern Analysis and Machine Intelligence pp

    Ren, K., Jiang, L., Lu, T., Yu, M., Xu, L., Ni, Z., Dai, B.: Octree-GS: To- wards Consistent Real-time Rendering with LOD-Structured 3D Gaussians. IEEE Transactions on Pattern Analysis and Machine Intelligence pp. 1–15 (2025). https://doi.org/10.1109/TPAMI.2025.3568201,https://ieeexplore.ieee. org/abstract/document/10993308

    doi:10.1109/tpami.2025.3568201 2025

  24. [24]

    Schonberger, J.L., Frahm, J.M.: Structure-From-Motion Revisited. pp. 4104– 4113 (2016),https : / / openaccess . thecvf . com / content _ cvpr _ 2016 / html / Schonberger_Structure-From-Motion_Revisited_CVPR_2016_paper.html

    2016

  25. [25]

    Shen, J., Qian, Y., Zhan, X.: LOD-GS: Achieving Levels of Detail using Scalable Gaussian Soup. pp. 671–680 (2025),https://openaccess.thecvf.com/content/ CVPR2025/html/Shen_LOD- GS_Achieving_Levels_of_Detail_using_Scalable_ Gaussian_Soup_CVPR_2025_paper.html

    2025

  26. [26]

    Svitov, D., Morerio, P., Agapito, L., Del Bue, A.: BillBoard Splatting (BB- Splat): Learnable Textured Primitives for Novel View Synthesis. pp. 25029–25039 (2025),https : / / openaccess . thecvf . com / content / ICCV2025 / html / Svitov _ BillBoard_Splatting_BBSplat_Learnable_Textured_Primitives_for_Novel_ View_Synthesis_ICCV_2025_paper.html

    2025

  27. [27]

    48550/arXiv.2309.16653,http://arxiv.org/abs/2309.16653, arXiv:2309.16653 [cs]

    Tang,J.,Ren,J.,Zhou,H.,Liu,Z.,Zeng,G.:DreamGaussian:GenerativeGaussian Splatting for Efficient 3D Content Creation (Mar 2024).https://doi.org/10. 48550/arXiv.2309.16653,http://arxiv.org/abs/2309.16653, arXiv:2309.16653 [cs]

    Pith/arXiv arXiv 2024

  28. [28]

    IEEE Transactions on Image Processing 13(4), 600–612 (Apr 2004).https://doi.org/10.1109/TIP.2003.819861,https: //ieeexplore.ieee.org/abstract/document/1284395

    Wang, Z., Bovik, A., Sheikh, H., Simoncelli, E.: Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing 13(4), 600–612 (Apr 2004).https://doi.org/10.1109/TIP.2003.819861,https: //ieeexplore.ieee.org/abstract/document/1284395

    doi:10.1109/tip.2003.819861 2004

  29. [29]

    In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G

    Ye, M., Danelljan, M., Yu, F., Ke, L.: Gaussian Grouping: Segment and Edit Anything in 3D Scenes. In: Leonardis, A., Ricci, E., Roth, S., Russakovsky, O., Sattler, T., Varol, G. (eds.) Computer Vision – ECCV 2024. pp. 162–179. Springer Nature Switzerland, Cham (2025).https://doi.org/10.1007/978-3-031-73397- 0_10

    work page doi:10.1007/978-3-031-73397- 2024

  30. [30]

    In: Proceedings of the 32nd ACM International Conference on Multimedia

    Ye, Z., Li, W., Liu, S., Qiao, P., Dou, Y.: AbsGS: Recovering Fine Details in 3D Gaussian Splatting. In: Proceedings of the 32nd ACM International Conference on Multimedia. pp. 1053–1061. MM ’24, Association for Computing Machinery, New York, NY, USA (Oct 2024).https://doi.org/10.1145/3664647.3681361, https://dl.acm.org/doi/10.1145/3664647.3681361

    work page doi:10.1145/3664647.3681361 2024

  31. [31]

    Yi, T., Fang, J., Wang, J., Wu, G., Xie, L., Zhang, X., Liu, W., Tian, Q., Wang, X.: GaussianDreamer: Fast Generation from Text to 3D Gaussians by Bridg- ing 2D and 3D Diffusion Models. pp. 6796–6807 (2024),https://openaccess. thecvf.com/content/CVPR2024/html/Yi_GaussianDreamer_Fast_Generation_ from_Text_to_3D_Gaussians_by_Bridging_CVPR_2024_paper.html

    2024

  32. [32]

    ACM Trans

    Yu, Z., Sattler, T., Geiger, A.: Gaussian Opacity Fields: Efficient Adaptive Surface Reconstruction in Unbounded Scenes. ACM Trans. Graph.43(6), 271:1–271:13 18 Jurca et al. (Nov 2024).https://doi.org/10.1145/3687937,https://dl.acm.org/doi/10. 1145/3687937

    doi:10.1145/3687937 2024

  33. [33]

    ACM Trans

    Zhang, B., Fang, C., Shrestha, R., Liang, Y., Long, X.X., Tan, P.: RaDe-GS: Ras- terizing Depth in Gaussian Splatting. ACM Trans. Graph.45(2), 19:1–19:14 (Mar 2026).https://doi.org/10.1145/3789201,https://dl.acm.org/doi/10.1145/ 3789201

    work page doi:10.1145/3789201 2026

  34. [34]

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The Unreasonable Ef- fectiveness of Deep Features as a Perceptual Metric. pp. 586–595 (2018),https:// openaccess.thecvf.com/content_cvpr_2018/html/Zhang_The_Unreasonable_ Effectiveness_CVPR_2018_paper.html

    2018

  35. [35]

    Zhang, Z., Huang, B., Jiang, H., Zhou, L., Xiang, X., Shen, S.: Quadratic Gaussian Splatting: High Quality Surface Reconstruction with Second-order Geometric Primitives. pp. 28260–28270 (2025),https://openaccess.thecvf. com / content / ICCV2025 / html / Zhang _ Quadratic _ Gaussian _ Splatting _ High _ Quality_Surface_Reconstruction_with_Second-order_Geom...

    2025

  36. [36]

    org/abs/2508.05343, arXiv:2508.05343 [cs]

    Zhou, J., Huang, Y., Dai, W., Zou, J., Zheng, Z., Kan, N., Li, C., Xiong, H.: 3DGabSplat: 3D Gabor Splatting for Frequency-adaptive Radiance Field Render- ing (Aug 2025).https://doi.org/10.48550/arXiv.2508.05343,http://arxiv. org/abs/2508.05343, arXiv:2508.05343 [cs]

    Pith/arXiv arXiv doi:10.48550/arxiv.2508.05343 2025

  37. [37]

    In: Proceedings of the 28th annual conference on Computer graphics and interactive techniques

    Zwicker, M., Pfister, H., van Baar, J., Gross, M.: Surface splatting. In: Proceedings of the 28th annual conference on Computer graphics and interactive techniques. pp. 371–378. SIGGRAPH ’01, Association for Computing Machinery, New York, NY, USA (Aug 2001).https://doi.org/10.1145/383259.383300,https://dl. acm.org/doi/10.1145/383259.383300 Fourier Splatti...

    work page doi:10.1145/383259.383300 2001