Pith. sign in

REVIEW 4 major objections 5 minor 46 references

Locality-Aware Density Control for Efficient Gaussian-based Image Representation

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper argues that the bottleneck in 2D-Gaussian image representation is locality-unaware density control, and that fixing it with region-wise densification and similarity-driven merging yields consistent quality gains under equal Gaussi

desk verdict Well-engineered 2DGS density control paper, but the headline gains are confounded by a non-standard training schedule that is never controlled. read the letter →

arxiv 2607.17896 v1 pith:45BWVUEF submitted 2026-07-20 cs.CV

Jiacong Chen , Qingyu Mao , Xiandong Meng , Shuai Liu , Chao Li , Fanyang Meng , Youneng Bao , Yongsheng Liang This is my paper
classification cs.CV
keywords ImagerepresentationGaussiansplattingDensitycontrolRegion-wisedensificationmergingLocaldistortioncoherence2DGaussiansPSNR
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 how a fixed budget of 2D Gaussians is placed during optimization matters as much as how many are used. It identifies two failures in existing methods: under-reconstructed areas are densified pixel-by-pixel instead of as coherent regions, and over-reconstructed areas keep redundant neighboring Gaussians with nearly identical appearance. To fix this, it proposes LocoADC, a plug-and-play density control with Region-wise Gaussian Densification (RGD) and Similarity-Driven Gaussian Merging (SDGM), plus a local color consistency constraint that makes merging safe. Across Kodak, DIV2K×2, and CLIC, the framework improves the GI, EA-GI, and GI-Plus baselines under the same Gaussian budgets, including a 2.93 dB PSNR gain over GI on CLIC at 30k Gaussians. If correct, this means better image quality at equal storage and decoding cost, since the rasterization and decoding pipeline is unchanged.

What carries the argument

The mechanism is a dual locality model. In image space, local distortion coherence D(x) = ||mean of normalized distortion gradients over a w×w window||², multiplied by distortion E(x), forms the densification candidate set; spatially neighboring candidates are grouped into regions and one Gaussian is initialized per region. In Gaussian space, similarity-driven merging selects stable, overlapping, color-similar Gaussian pairs via Mahalanobis distance and a color threshold, ranks them by a Pair Discrepancy Score over position, color, and covariance, and merges via area-weighted position/color averaging with second-moment covariance matching under a support-area constraint. The supporting local

What would settle it

Run GI and GI-Plus through the full LocoADC schedule with RGD replaced by standard pixel-wise densification and SDGM disabled, keeping every other detail fixed (120k iterations, three stages, 5k operation intervals, same Gaussian budgets). If the PSNR gain over the published baselines largely persists, the locality-aware components are not the cause.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that reconstruction errors in Gaussian image representation are not isolated pixels but locally coherent patterns, and that spatially neighboring Gaussians converge to redundant similar shapes; exploiting both forms of locality yields a better use of a fixed Gaussian capacity. Concretely, LocoADC uses the distortion map's normalized gradient field to measure local distortion coherence, weights the distortion by this coherence, selects top candidates, groups neighboring candidates into regions, and allocates one new Gaussian per region. In parallel, it filters stable Gaussians, pairs spatially overlapping neighbors with similar colors, ranks pairs by

Load-bearing premise

That the PSNR improvements are caused by the locality-aware RGD/SDGM operations rather than by the altered training schedule alone; the paper does not run the baselines through the identical schedule without those operations.

Editorial extensions

If this is right

  • Under the same Gaussian budget, density control that respects locality can produce materially better reconstructions: the paper reports up to +2.93 dB PSNR (GI on CLIC at 30k Gaussians) and consistent gains across all three baselines and datasets.
  • Because LocoADC changes only density control and not the rasterization or decoding pipeline, the quality gains translate directly to existing Gaussian-based representations without extra decoding cost.
  • Merging redundant similar Gaussians is more effective than pruning small ones: the paper compares against size-based and learning-based 3DGS sparsification and reports higher PSNR with merging (32.68 dB vs 30.5 and 32.19 on Kodak).
  • The ablation indicates both components contribute: RGD alone adds about 0.34 dB over pixel-wise densification, and coherence-aware identification beats magnitude-only region-wise densification; SDGM plus color consistency adds further gains.
  • After convergence the method reaches a better PSNR-time trade-off than the baselines despite higher training cost, and its rate-distortion curve approaches learned compression methods on the reported benchmarks.

Reading between the lines

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

  • A clean control experiment is missing in the paper: running the baselines through the exact three-stage LocoADC schedule (50k fit+densify, 20k merge-only, 50k fine-tune, operations every 5k iterations) with RGD and SDGM disabled. Until that is reported, part of the gain could be an artifact of the schedule itself.
  • The same locality principles could transfer to video or 3D Gaussian representations: wherever reconstruction errors are spatially coherent and neighboring primitives converge to similar appearance, region-wise densification and similarity-driven merging should free capacity; the paper lists these as future work rather than demonstrating them.
  • One observable side effect to watch: merging alone lowers MS-SSIM slightly (PSNR gains from 31.90 to 32.34 dB but MS-SSIM drops from 0.9614 to 0.9561), so aggressive merging trades a bit of fine-scale metric fidelity for capacity; the color-consistency constraint only partially compensates, suggesting a perceptual ceiling on how much redundancy can be safely merged.
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

4 major / 5 minor

Summary. The paper proposes LocoADC, a density-control framework for 2D Gaussian image representation, with two components: Region-wise Gaussian Densification (RGD), which allocates new Gaussians based on locally coherent distortion structures, and Similarity-Driven Gaussian Merging (SDGM), which merges redundant neighboring Gaussians under appearance/stability constraints, supported by a local color consistency regularizer. Experiments on Kodak, DIV2K×2, and CLIC show PSNR improvements over GI, EA-GI, and GI-Plus across several Gaussian budgets, including the headline 2.93 dB gain over GI on CLIC at a 30k Gaussian budget. The appendix includes component ablations, hyperparameter sensitivity, and comparisons with 3DGS sparsification strategies.

Significance. If the reported gains are attributable to the proposed locality-aware mechanisms, LocoADC would be a useful, plug-and-play improvement for Gaussian-based image representation, with potential implications for image compression. The paper has clear strengths: evaluation across three datasets, three baselines, multiple budgets, component ablations, a sensitivity analysis, pseudocode, and a public code link. However, the empirical attribution is currently weakened by a training-schedule confound and by unaddressed MS-SSIM trade-offs at low budgets. These issues are fixable but load-bearing for the central claim that the RGD/SDGM components themselves are responsible for the consistent improvements.

major comments (4)
  1. [§4.1, Tables 1–2] The training schedule is not controlled between LocoADC and the baselines. LocoADC uses a staged schedule (50k iterations of fitting+densification, 20k merge-only, 50k fine-tuning with continued densification; operations every 5k), whereas baseline results are obtained using their official schedules for the same total 120k iterations. Thus the reported gains, including the 2.93 dB CLIC number, may partly reflect the extra merge-only and fine-tuning phases rather than RGD/SDGM. The ablations in Table 2 also change the schedule when a component is removed (e.g., the 20k merge-only stage is present only when merging is active). Please add a schedule-only control in which each baseline, or at least GI-Plus, runs the identical staged schedule without RGD/SDGM/color loss, and report those results alongside Table 1 and Table 2.
  2. [Table 1, §4.4] MS-SSIM drops at low Gaussian budgets are not discussed. For example, on Kodak at 5k, GI goes from 0.9632 to 0.9560 with LocoADC, and EA-GI from 0.9614 to 0.9551; at 10k on Kodak, GI drops from 0.9824 to 0.9794. The text states LocoADC 'remain[s] competitive in MS-SSIM,' which is not supported. The ablation in Table 7 shows that merging alone reduces MS-SSIM from 0.9640 (RGD only) to 0.9561, with color consistency only partially recovering it to 0.9596. The paper should explicitly report and analyze this PSNR–MS-SSIM trade-off, and qualify the 'consistent improvement' claim accordingly.
  3. [Eq. (11), Algorithm 2] The gradient-stability filter appears internally inconsistent. The text says 'we first filter out unstable Gaussians according to their accumulated optimization gradients,' explaining that Gaussians with large gradients are under active correction and may incur large errors after merging. But the stated condition, g_i / max_k g_k < 0.5, would select low-gradient (stable) Gaussians for removal if 'filter out' is interpreted literally, which is the opposite of the stated intent. Please clarify whether the threshold is intended to remove high-gradient Gaussians (i.e., retain those with relative gradient below 0.5) or to remove low-gradient ones, and correct Eq. (11) and Algorithm 2 to match the code. This affects the reproducibility and correctness of SDGM.
  4. [§4.4, Table 2] The component ablation is not a factorial design that isolates each mechanism while holding the optimization trajectory fixed. Adding RGD changes both the densification criterion and the densification schedule; adding merging changes both the sparsification operation and the 20k merge-only phase. Without a control that uses the same schedule and operation timing while toggling only the RGD/SDGM decisions, the attribution of the observed gains to the locality-aware criteria is not established. This is closely related to the schedule-confound concern above and should be addressed in the same set of experiments.
minor comments (5)
  1. [§4.4] The text refers to 'Fig. 9(a)' and 'Fig. 9(b)' when discussing the local color consistency constraint, but the displayed figure is captioned 'Figure 7: Effectiveness of local color consistency constraint.' The figure references in the appendix's Fig. 9 are also confusing because that figure shows rate-distortion/time-quality plots. Please renumber or fix the cross-references.
  2. [§3.2, Eq. (8)] The notation D(x) for local distortion coherence in Eq. (7)–(8) is easily confused with the densification operator D(G_t) in Eq. (4). Consider using a different symbol, e.g., C(x) or H(x), for the coherence map.
  3. [§4.1] For a fair comparison, the statement that all methods use 120,000 iterations is insufficient. Please specify the exact schedules, learning rates, and other hyperparameters used for each baseline, and report the number of seeds/runs and standard deviations for the main tables.
  4. [Appendix A, Algorithm 1] The grouping of 'spatially neighboring pixels' in RGD is not formalized. Please provide the connectivity/radius criterion used to form regional subsets, as this affects reproduction and the number of Gaussians allocated per region.
  5. [Table 6] The gradient-cutoff sensitivity row shows MS-SSIM varying from 0.9622 (at 0.1) to 0.9582 (at 0.9), a larger spread than PSNR. This trade-off is not discussed and should be acknowledged, especially since the default 0.5 gives the lowest MS-SSIM among tested cutoff values.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical method paper with no derivation that reduces to its inputs.

full rationale

This paper does not claim a first-principles derivation or a predictive law; it proposes two algorithmic components (RGD and SDGM) plus a color-consistency regularizer and evaluates them on standard benchmarks. No equation in the paper is defined in terms of the quantity it is claimed to predict: the densification candidates are computed from the current reconstruction error, the merging criterion uses Mahalanobis overlap and color difference, and final PSNR is measured against ground-truth images on external test sets. The gains are not statistically forced by construction, because the PSNR target is not used as a fitting objective for any parameter that is then reported as a prediction. Self-citations appear only as baselines (e.g., EA-GI, which shares authors with this paper) and do not carry the argument: the central comparison against GI and GI-Plus, which are not self-citations, provides independent support. There is no imported uniqueness theorem, no ansatz justified only by the authors' prior work, and no renamed known result presented as a new derivation. The reviewer-flagged concern that the staged training schedule (50k fitting + 20k merge-only + 50k fine-tuning) is not held fixed across baselines is a legitimate ablation/correctness issue, but it is a confound about attribution of improvement, not a circularity: the method's outputs are not equivalent to its inputs by definition. The appendix's own limitation statements (performance gap to neural representations, limited scope) further confirm the paper's claims are empirical and bounded rather than self-validating. Therefore, no circular step is present and the circularity score is 0.

Assumptions & free parameters 9 free parameters · 7 assumptions · 0 invented entities

The method introduces no new physical entities. It relies on standard rendering equations, hand-chosen hyperparameters, and two empirical premises about locality (error continuity and Gaussian similarity). The most fragile free parameter is the training schedule, which is not isolated from the algorithmic contribution.

free parameters (9)
  • window size w = 7
    Used in Eq. (7) to compute local distortion coherence; sensitivity analysis in Tab. 6 shows stability across 3-11.
  • candidate number K = not specified in main text
    Number of top pixels selected in Eq. (8) for densification; default not stated.
  • Mahalanobis overlap threshold tau_M = 5.0
    Used in Eq. (12) to define Gaussian adjacency; sensitivity analysis in Tab. 6.
  • color similarity threshold tau_color = 10/255
    Used in Eq. (13) to retain similar color pairs; sensitivity analysis shows trade-off with MS-SSIM.
  • merging constraint ratio eta = 1.5
    Used in Eq. (17) to prevent overly loose merging; sensitivity analysis in Tab. 6.
  • color consistency weight lambda_color = 0.01
    Used in Eq. (10); sensitivity analysis in Tab. 6 shows mild variation.
  • gradient cutoff for stable Gaussians = 0.5
    Used in Eq. (11) to filter unstable Gaussians before merging; sensitivity analysis in Tab. 6.
  • training schedule = 50k/20k/50k
    Staged schedule: 50k fit+densify, 20k merge-only, 50k fine-tune; not ablated against baselines.
  • densification/merging interval = 5000 iterations
    Both operations run every 5k iterations.
assumptions (7)
  • domain assumption The GI accumulated blending rasterization (Eq. 1-2) is used as the rendering model.
    Central premise borrowed from [38]; LocoADC modifies density control only, not the rendering equation.
  • domain assumption Reconstruction errors are locally continuous, so high-error pixels form coherent regions.
    Motivated by Fig. 2(a); the entire RGD strategy rests on this empirical premise.
  • domain assumption Sobel-based distortion gradient direction is a reliable measure of local distortion coherence.
    Eqs. (6)-(7) define coherence from the Sobel gradient; this is an ad hoc modeling choice, validated only indirectly via ablations.
  • domain assumption Spatially nearby Gaussians that overlap and have similar colors can be merged with negligible fidelity loss.
    Core premise of SDGM; the ablation (Tab. 7) shows merging alone lowers MS-SSIM, so the assumption holds only approximately.
  • domain assumption Area-weighted second-moment matching (Eq. 16) preserves the covered area of the original pair.
    Used to initialize the merged Gaussian; axiomatic in this paper.
  • ad hoc to paper The support-area constraint (Eq. 17) with eta=1.5 prevents quality degradation from over-merging.
    The threshold is hand-chosen; sensitivity analysis shows larger eta reduces MS-SSIM.
  • domain assumption Optimizing 120k iterations with the given schedule is a fair comparison across methods.
    Baselines are run for the same iteration count, but not necessarily with the same staged schedule; this is a key unverified assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Locality-Aware Density Control for Efficient Gaussian-based Image Representation." pith.science (2026). https://pith.science/paper/45BWVUEF

@misc{pith2026260717896,
  author       = {Pith},
  title        = {Pith review of: Locality-Aware Density Control for Efficient Gaussian-based Image Representation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/45BWVUEF}},
  note         = {Machine review of arXiv:2607.17896}
}
read the original abstract

2D Gaussian Splatting is an attractive direction for image representation due to its explicit formulation, fast rasterization, and favorable decoding efficiency. The representation quality of this paradigm depends on the proper allocation of Gaussian capacity to the demanding regions. However, existing methods fail to allocate Gaussian capacity efficiently during optimization: under-reconstructed content is often refined in a fragmented pixel-wise manner, while neighboring optimized Gaussians with similar attributes are redundantly retained. This inefficiency motivates the need for a density control framework that jointly addresses insufficient allocation in under-reconstructed regions and redundant allocation in over-reconstructed regions. Our key insight is that this framework should exploit two complementary forms of locality: the local continuity of reconstruction errors in image space for improved Gaussian allocation, and the local similarity of neighboring Gaussians in Gaussian space for redundant elimination. Based on this insight, we propose Locality-Aware Density Control (LocoADC), a plug-and-play framework that improves Gaussian capacity utilization through Region-wise Gaussian Densification (RGD) and Similarity-Driven Gaussian Merging (SDGM) strategies, together with a local color consistency constraint for more reliable merging. Extensive experiments on diverse datasets show that LocoADC consistently improves multiple baselines by enabling more effective local Gaussian allocation, including a 2.93 dB PSNR gain over GI on the CLIC dataset under the same 30k Gaussian budget. Code is available at: \textit{https://github.com/ChenJiaCong-1005/LocoADC}.

Figures

Figures reproduced from arXiv: 2607.17896 by the authors.

Figure 1
Figure 1. (a) Visual comparisons show that our method pro [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Locality of Gaussian-based image representation. (a) [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the proposed LocoADC framework. Our LocoADC improves Gaussian allocation from two aspects: in [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparison under different numbers of Gaussians on the Kodak and DIV2K [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison of our approach with GI, EA-GI, and GI-Plus on the Kodak, DIV2K [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Locality analysis in local regions. For each region, we compare the Gaussian density maps, the number of allocated [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Effectiveness of local color consistency constraint. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Detailed training process of the proposed LocoADC framework. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Rate-distortion (left) and Time-PSNR (right) comparisons. [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Qualitative ablation of the RGD strategy. Com [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Qualitative results on Kodak, DIV2k ×2 and CLIC datasets [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 5 linked inside Pith

  1. [1]

    Eirikur Agustsson and Radu Timofte. 2017. Ntire 2017 challenge on single image super-resolution: Dataset and study. InProceedings of the IEEE conference on computer vision and pattern recognition workshops. 126–135

    2017

  2. [2]

    Johannes Ballé, Valero Laparra, and Eero P Simoncelli. 2016. End-to-end optimized image compression.arXiv preprint arXiv:1611.01704(2016)

    arXiv 2016

  3. [3]

    Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick John- ston. 2018. Variational image compression with a scale hyperprior.arXiv preprint arXiv:1802.01436(2018)

    arXiv 2018

  4. [4]

    Youneng Bao, Wen Tan, Chuanmin Jia, Mu Li, Yongsheng Liang, and Yonghong Tian. 2025. ShiftLIC: Lightweight Learned Image Compression With Spatial- Channel Shift Operations.IEEE Trans. Circuits Syst. Video Technol.35, 9 (2025), 9428–9442

    2025

  5. [5]

    Jiacong Chen, Qingyu Mao, Shuai Liu, Chao Li, Jierun Lin, Xiandong Meng, Fanyang Meng, and Yongsheng Liang. 2026. Entropy-Aware Image Representa- tion via 2D Gaussian Splatting.Signal Processing(2026), 110521

    2026

  6. [6]

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto. 2020. Learned image compression with discretized gaussian mixture likelihoods and attention modules. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 7939–7948

    2020

  7. [7]

    Renato J Cintra and Fábio M Bayer. 2011. A DCT approximation for image compression.IEEE Signal Processing Letters18, 10 (2011), 579–582

    2011

  8. [8]

    Mikolaj Czerkawski, Javier Cardona, Robert Atkinson, Craig Michie, Ivan An- donovic, Carmine Clemente, and Christos Tachtatzis. 2024. Neural knitworks: Patched neural implicit representation networks.Pattern Recognition151 (2024), 110378

    2024

Show all 46 references
  1. [9]

    Emilien Dupont, Adam Goliński, Milad Alizadeh, Yee Whye Teh, and Arnaud Doucet. 2021. Coin: Compression with implicit neural representations.arXiv preprint arXiv:2103.03123(2021)

    2021 arXiv

  2. [10]

    Zhiwen Fan, Kevin Wang, Kairun Wen, Zehao Zhu, Dejia Xu, Zhangyang Wang, et al. 2024. Lightgaussian: Unbounded 3d gaussian compression with 15x reduc- tion and 200+ fps.Advances in neural information processing systems37 (2024), 140138–140158

    2024

  3. [11]

    Abdullah Hamdi, Luke Melas-Kyriazi, Jinjie Mai, Guocheng Qian, Ruoshi Liu, Carl Vondrick, Bernard Ghanem, and Andrea Vedaldi. 2024. Ges: Generalized exponential splatting for efficient radiance field rendering. InProceedings of the IEEE/CVF Conference on Computer Vision and Pa...

    2024

  4. [12]

    Dailan He, Ziming Yang, Weikun Peng, Rui Ma, Hongwei Qin, and Yan Wang

  5. [13]

    Shoukang Hu, Tao Hu, and Ziwei Liu. 2024. Gauhuman: Articulated gaussian splatting from monocular human videos. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 20418–20431

    2024

  6. [14]

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis

  7. [15]

    Kodak. 1999. Kodak lossless true color image suite. In https://r0k.us/graphics/kodak/. 11

    1999

  8. [16]

    Joo Chan Lee, Daniel Rho, Xiangyu Sun, Jong Hwan Ko, and Eunbyung Park

  9. [17]

    Linfei Li, Lin Zhang, Zhong Wang, and Ying Shen. 2026. SmartSplat: Feature- Smart Gaussians for Scalable Compression of Ultra-High-Resolution Images. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 40. 6315–6323

    2026

  10. [18]

    Tiantian Li, Xinjie Zhang, Xingtong Ge, Tongda Xu, Dailan He, Jun Zhang, and Yan Wang. 2026. GaussianImage++: Boosted Image Representation and Compression with 2D Gaussian Splatting. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 40. 6442–6449

    2026

  11. [19]

    Huanxiong Liang, Yunuo Chen, Yicheng Pan, Sixian Wang, Jincheng Dai, Guo Lu, and Wenjun Zhang. 2025. Structure-Guided Allocation of 2D Gaussians for Image Representation and Compression.arXiv preprint arXiv:2512.24018(2025)

    2025

  12. [20]

    Xiangrui Liu, Xinju Wu, Pingping Zhang, Shiqi Wang, Zhu Li, and Sam Kwong

  13. [21]

    Panagiotis Papantonakis, Georgios Kopanas, Bernhard Kerbl, Alexandre Lanvin, and George Drettakis. 2024. Reducing the memory footprint of 3d gaussian splatting.Proceedings of the ACM on Computer Graphics and Interactive Techniques 7, 1 (2024), 1–17

    2024

  14. [22]

    Vincent Sitzmann, Julien Martel, Alexander Bergman, David Lindell, and Gordon Wetzstein. 2020. Implicit neural representations with periodic activation func- tions.Advances in neural information processing systems33 (2020), 7462–7473

    2020

  15. [23]

    Yangyang Song and Xiaozhen Xie. 2025. Triple-layer representation of low rank and group sparsity for hyperspectral image denoising.Signal Processing233 (2025), 109960

    2025

  16. [24]

    InProceedings of the 32nd ACM International Conference on Multimedia

    Compgs: Efficient 3d scene representation via compressed gaussian splat- ting. InProceedings of the 32nd ACM International Conference on Multimedia. 2936–2944

  17. [25]

    Matthew Tancik, Pratul Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan Barron, and Ren Ng

  18. [26]

    David S Taubman, Michael W Marcellin, and Majid Rabbani. 2002. JPEG2000: Image compression fundamentals, standards and practice.Journal of Electronic Imaging11, 2 (2002), 286–287

    2002

  19. [27]

    George Toderici, Wenzhe Shi, Radu Timofte, Lucas Theis, Johannes Balle, Eirikur Agustsson, Nick Johnston, and Fabian Mentzer. 2020. Workshop and challenge on learned image compression (clic2020). InCVPR

    2020

  20. [28]

    Yannick Strümpler, Janis Postels, Ren Yang, Luc Van Gool, and Federico Tombari

  21. [29]

    InEuropean Con- ference on Computer Vision

    Implicit neural representations for image compression. InEuropean Con- ference on Computer Vision. Springer, 74–91

  22. [30]

    Hao Wang, Ashish Bastola, Chaoyi Zhou, Wenhui Zhu, Xiwen Chen, Xuanzhao Dong, Siyu Huang, and Abolfazl Razi. 2026. Fast 2DGS: Efficient Image Rep- resentation with Deep Gaussian Prior. InProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 1184–1193

    2026

  23. [31]

    Henan Wang, Hanxin Zhu, Tianyu He, Runsen Feng, Jiajun Deng, Jiang Bian, and Zhibo Chen. 2024. End-to-end rate-distortion optimized 3d gaussian repre- sentation. InEuropean Conference on Computer Vision. Springer, 76–92

    2024

  24. [32]

    Zhou Wang, Eero P Simoncelli, and Alan C Bovik. 2003. Multiscale structural similarity for image quality assessment. InThe thrity-seventh asilomar conference on signals, systems & computers, 2003, Vol. 2. Ieee, 1398–1402

    2003

  25. [33]

    Zongxin Ye, Wenyu Li, Sidun Liu, Peng Qiao, and Yong Dou. 2024. Absgs: Recovering fine details in 3d gaussian splatting. InProceedings of the 32nd ACM international conference on multimedia. 1053–1061

    2024

  26. [34]

    Joanna Waczyńska, Tomasz Szczepanik, Piotr Borycki, Sławomir Tadeja, Thomas Bohné, and Przemysław Spurek. 2024. Mirage: Editable 2d images using gaussian splatting.arXiv preprint arXiv:2410.01521(2024)

    2024 arXiv

  27. [35]

    Gregory K Wallace. 1991. The JPEG still picture compression standard.Commun. ACM34, 4 (1991), 30–44

    1991

  28. [36]

    Weixiang Zhang, Shuzhao Xie, Chengwei Ren, Shijia Ge, Mingzi Wang, and Zhi Wang. 2025. Enhancing implicit neural representations via symmetric power transformation. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 10157–10165

    2025

  29. [37]

    Weixiang Zhang, Shuzhao Xie, Chengwei Ren, Siyi Xie, Chen Tang, Shijia Ge, Mingzi Wang, and Zhi Wang. 2025. Evos: Efficient implicit neural training via evolutionary selector. InProceedings of the Computer Vision and Pattern Recognition Conference. 30472–30482

    2025

  30. [38]

    Xinjie Zhang, Xingtong Ge, Tongda Xu, Dailan He, Yan Wang, Hongwei Qin, Guo Lu, Jing Geng, and Jun Zhang. 2024. Gaussianimage: 1000 fps image repre- sentation and compression by 2d gaussian splatting. InEuropean Conference on Computer Vision. Springer, 327–345

    2024

  31. [39]

    Zheng Zhang, Wenbo Hu, Yixing Lao, Tong He, and Hengshuang Zhao. 2024. Pixel-gs: Density control with pixel-aware gradient for 3d gaussian splatting. In European Conference on Computer Vision. Springer, 326–342

    2024

  32. [40]

    Zehao Yu, Torsten Sattler, and Andreas Geiger. 2024. Gaussian opacity fields: Efficient adaptive surface reconstruction in unbounded scenes.ACM Transactions on Graphics (ToG)43, 6 (2024), 1–13

    2024

  33. [41]

    Zhaojie Zeng, Yuesong Wang, Chao Yang, Tao Guan, and Lili Ju. 2025. Instant GaussianImage: A Generalizable and Self-Adaptive Image Representation via 2D Gaussian Splatting.arXiv preprint arXiv:2506.23479(2025)

    2025 arXiv

  34. [46]

    Lingting Zhu, Guying Lin, Jinnan Chen, Xinjie Zhang, Zhenchao Jin, Zhao Wang, and Lequan Yu. 2025. Large images are Gaussians: High-quality large image representation with levels of 2D Gaussian splatting. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 39...

    2025

  35. [2020]

    Fourier features let networks learn high frequency functions in low di- mensional domains.Advances in neural information processing systems33 (2020), 7537–7547

    2020

  36. [2022]

    InProceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Elic: Efficient learned image compression with unevenly grouped space- channel contextual adaptive coding. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition. 5718–5727

  37. [2023]

    Graph.42, 4 (2023), 139–1

    3D Gaussian splatting for real-time radiance field rendering.ACM Trans. Graph.42, 4 (2023), 139–1

    2023

  38. [2024]

    InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Compact 3d gaussian representation for radiance field. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 21719–21728

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.