Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Compression itself can be the forward diffusion, so one reverse network handles all bit rates in two steps.

desk verdict A genuinely useful multi-rate generative codec with a plausible and promising empirical design, but the central score identity is asserted without proof and is false for uniform quantization noise, so the theoretical claims and the SOTA numbers should be treated as provisional. read the letter →

arxiv 2505.20984 v1 pith:7DVTIKLT submitted 2025-05-27 eess.IV cs.CV

Minghao Han , Weiyi You , Jinhua Zhang , Leheng Zhang , Ce Zhu , Shuhang Gu This is my paper
classification eess.IVcs.CV
keywords generativeimagecompressionscore-baseddiffusionstochasticdifferentialequationsrate-variablequantizationscalinglearnedreverseperceptualreconstruction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that the lossy-compression pipeline itself can serve as the forward corruption process of a diffusion model, so that a single reverse network can undo compression at any chosen bit rate and produce photo-realistic images. In this framing, rate-variable quantization is modeled as adding uniform noise of width $q(t)$, and the task of reversing that noise is treated as a score-matching problem. The authors train a denoiser to estimate the score of the rate-variable feature distribution and use it in an Euler reverse sampler, adding a small stochastic term to improve perceptual quality. If the claim is right, generative image compression no longer needs a separate denoising pass after decoding; the reverse network is the generative model, and only two reverse steps are needed for any rate.

What carries the argument

The central object is the rate-variable compression forward process: scaling a latent $y$ by $q$ before quantization and rescaling after produces $y + U(-0.5, 0.5)\cdot q$, so the quantization scale $q(t)$ plays the role of diffusion time. The reverse machinery is the ODE $dx = -(dq/dt)\nabla \log p(x; q(t))\,dt$, discretized by Euler's method, with the score replaced by the trained denoiser through $\nabla \log p(x; q(t)) \approx (D_\theta(x, q) - x)/q$. This substitution is what carries the argument: it converts the compression-rate schedule into a reverse-time trajectory from compressed to original features, and the stochastic sampler in Eq. (12) with $\alpha_t = \beta\sqrt{q_t - q_{\min}}$ injects controlled randomness to improve perceptual fidelity.

What would settle it

Estimate $\nabla \log p(x; q)$ by Monte Carlo for a fixed latent distribution corrupted by $U(-0.5, 0.5)\cdot q$ and compare it with the paper's proxy $(D_\theta(x,q) - x)/q$; a systematic discrepancy would indicate that the Euler update in Eq. (11) is not reversing the forward process, breaking the claimed link between the compression schedule and the score.

Watch

Extended reading notes

Core claim

The paper claims that compression, implemented as quantization scaling of a latent feature, is a legitimate forward diffusion path described by a stochastic differential equation, and that reversing that path with an estimated score function yields the reconstructed image directly. The forward process is approximated by $x_t = x_0 + U(-0.5, 0.5)\cdot q(t)$, the reverse ODE is $dx = -(dq/dt)\nabla \log p(x; q(t))\,dt$, and the trained denoiser $D_\theta$ supplies the score through $\nabla \log p(x; q(t)) = (D_\theta(x_t, q_t) - x_t)/q_t$. Discretizing with Euler's method gives the update $x_{i+1} = x_i + \frac{q_i - q_{i+1}}{q_i}(D_\theta(x_i, q_i) - x_i)$, and adding a noise term with amplitude $\beta\sqrt{q_t - q_{\min}}$ yields the stochastic sampler. The reverse network is trained with a plain L2 reconstruction loss against the original latent, without adversarial or perceptual losses, and the authors report that their method outperforms existing generative compression approaches on perceptual distortion, statistical fidelity, and no-reference quality metrics across standard benchmarks.

Load-bearing premise

The paper assumes that the gradient of the log-probability of a compressed feature is exactly $(D_\theta(x, q) - x)/q$, an identity proven for Gaussian noise but assumed without proof for the uniform quantization noise that actually corrupts the features; if that equality is wrong, the reverse ODE and stochastic sampler are not following the true score, and the method's theoretical grounding collapses.

Editorial extensions

If this is right

  • A single trained model covers all bit rates by varying the quantization scale $q$, removing the need to retrain separate models per rate.
  • Only two reverse steps are needed at decode time, making the diffusion-based decoder faster than prior diffusion-based compression systems.
  • No GAN or perceptual-loss finetuning is required; the score-based reverse alone is claimed to achieve competitive perceptual quality.
  • The framework applies to any learned compression pipeline that can be expressed as additive uniform quantization noise.
  • The rate-distortion curve becomes continuously adjustable through the quantization scale, enabling smooth rate control in deployment.

Reading between the lines

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

  • The score formula $(D_\theta(x,q) - x)/q$ is proven for Gaussian noise, not for uniform noise; a variant using the exact score of a uniform-noise distribution would test whether the sampler's theoretical basis matters for the reported results.
  • Because the entropy model already estimates the distribution of quantized features, one could combine that learned prior with the estimated score to compute an exact likelihood, enabling rate control under an information-theoretic objective.
  • The randomness-injection amplitude $\alpha_t = \beta\sqrt{q_t - q_{\min}}$ is set empirically; a principled derivation from the reverse SDE's stationary distribution would turn this heuristic into a testable schedule.
  • The method's two-step decoding suggests the effective forward corruption is mild relative to Gaussian diffusion, so the same reverse network might transfer to other bounded-noise corruption processes, such as dropout or JPEG-style quantization.
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 a diffusion-style generative image compression framework in which rate-variable quantization is reinterpreted as a forward diffusion process. A reverse network is trained with an L2 loss to invert quantization, and an ODE/SDE sampler is derived to reconstruct images from the compressed latent. Experiments on DIV2K, Kodak, and CLIC2020 report competitive perceptual metrics (LPIPS, FID, KID, MUSIQ, CLIPIQA) with only two reverse steps, and the method is claimed to outperform prior generative compression approaches without GAN or perceptual-loss finetuning.

Significance. If the theoretical framework were sound, this would be a meaningful contribution: it combines rate-variable compression with a minimal-step generative decoder and demonstrates strong perceptual results. The empirical study is broad, includes several benchmarks and metrics, and the design of the reverse network as a denoiser conditioned on the quantization scale is interesting. However, the central theoretical claim—that the learned network estimates the score of the corrupted feature distribution—is not established and, in the form stated, is false for the uniform quantization noise used in the forward process. The paper may still describe a useful heuristic denoiser, but the 'estimating gradients' framing and the SDE derivation do not hold as written.

major comments (4)
  1. [Sec. 3.2, Eq. (9)-(11)] The score identity ∇_x log p(x; q(t)) = (D_θ(x,q)-x)/q(t) is not valid for additive uniform quantization noise. For a data distribution p_data = δ_0 and q = 1, the corrupted density p(x; q) is uniform on [−0.5, 0.5], so the score is zero on the interior of the support, while the proposed formula gives −x, which is nonzero. The L2-trained denoiser D_θ (Eq. 15) estimates the posterior mean E[x_0 | x_t], not the score; for uniform noise there is no Tweedie-style identity connecting the MMSE denoiser to the score. Consequently, the Euler update in Eq. (11), and the reverse ODE/SDE steps derived from it, do not follow from the stated forward process.
  2. [Sec. 3.2, Eq. (10)] The probability-flow ODE dx = −(dq/dt) ∇_x log p(x; q(t)) dt is asserted without derivation. In standard score-based diffusion, this ODE is derived from the Fokker–Planck equation of a Gaussian SDE. Here the forward process is defined by Eq. (6)/(9) as quantization or additive uniform noise, and no such derivation is provided; the claim that the ODE preserves the marginal distributions p(x; q(t)) is an unproven assumption rather than a consequence of the construction.
  3. [Sec. 4.2, Fig. 3 and Eq. (13)] The stochastic sampler parameters are tuned on the evaluation benchmark. The randomness form ω and the coefficient β are selected by evaluating on DIV2K (Fig. 3), and the same DIV2K set is used for the main comparison in Fig. 2. This test-set tuning undermines the validity of the reported DIV2K superiority and makes the 'SOTA' claim hard to interpret. Additionally, the final model description does not state whether Gaussian or uniform noise is used for the reported results, so the experiments are not fully reproducible.
  4. [Algorithm 1, line 10] There is a clear implementation error in the displayed pseudocode: line 10, ¯y_{i+1} ← α(ϵ_i − d_i), overwrites the Euler update from line 8 instead of adding the stochastic term to it. The update should be ¯y_{i+1} ← ¯y_{i+1} + α(ϵ_i − d_i). This obscures the actual sampling procedure and should be corrected, along with a description of how the continuous Eq. (12) is discretized into lines 7–10.
minor comments (5)
  1. [Sec. 3.2 and Appendix A] The relation between the quantization scale q(t) and the time variable t is not stated in the main text; the substitution q(t) := t appears only in Appendix A. This should be clarified in Sec. 3.2, since the notation q(t_i) − q(t_{i+1}) in Eq. (11) depends on that choice.
  2. [Sec. 4.1] The definition of q_min as 'lower than the minimum support' is vague; it should be specified numerically or in terms of the entropy model's supported quantization range.
  3. [Eq. (13)] The text contains a typo: 'minimus' should be 'minimum'.
  4. [Abstract and Sec. 3.2] The statement 'without requiring Gaussian noise initialization' is confusing because the stochastic sampler in Eq. (12) and Algorithm 1 injects noise during decoding. The intended meaning is that the initial state is the compressed latent, not pure Gaussian noise; this should be stated explicitly.
  5. [Sec. 1 and reproducibility] The phrase 'The code is available here' is followed by no URL or link; provide the repository address or a supplementary archive for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the reverse sampler is a learned-denoiser interpolation, and the score identity is an unproven approximation rather than a circular reduction.

full rationale

The paper's derivation chain is not circular. The forward process is defined by rate-variable quantization approximated as additive uniform noise (Eq. 9), the reverse ODE is borrowed from score-based SDE theory (Eq. 10), and the score is replaced by the denoiser expression (x_hat_0 - x_t)/q(t). The denoiser D_theta is trained independently with an L2 loss (Eq. 15) on actual compressed latents produced by the entropy model, and the final sampler (Eq. 11) is a linear interpolation between the corrupted latent and D_theta's estimate. This is a direct use of a learned denoiser rather than a redefinition of the target in terms of the output; the sampler's behavior is not forced to equal a fitted parameter renamed as a prediction. The main weakness is that the score identity is mathematically unjustified for uniform quantization noise: the true score of a uniformly corrupted distribution is not generally (posterior_mean - x)/q, so the theoretical claim that the network estimates the score and thereby reverses the compression process is not established. That is a correctness and validity concern, not a circularity. Likewise, the stochasticity hyperparameter beta is selected using DIV2K metrics (Fig. 3) and then reported on DIV2K (Fig. 2), which is a benchmark-tuning weakness but does not make the reconstruction a consequence of the fit by construction. The only self-citation, [15], supplies the entropy-model backbone and is not load-bearing for the diffusion-sampler claim. No step reduces by construction to its own inputs, so no circularity is found.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The framework rests on the approximation of quantization as uniform noise and on two unproven identities: the probability-flow ODE for this noise process and the score formula in terms of the denoiser. One hyperparameter (β) is selected on the test set. No new physical entities are postulated.

free parameters (2)
  • β (stochasticity strength) = 0.075
    Controls the magnitude of random injection in the sampling schedule (Eq 13). Chosen by testing on DIV2K (Sec 4.2, Fig 3), the same benchmark used for the main state-of-the-art comparison.
  • Number of reverse steps N = 2
    Fixed to 2 for all bit rates; a hand-chosen design decision that trades quality for latency.
assumptions (5)
  • domain assumption The quantization operation is approximated as additive uniform noise: x_t = x0 + U(-0.5,0.5) q(t) (Eq 9).
    This is a standard approximation in learned image compression, but it ignores content-dependent quantization error and correlation between latent dimensions.
  • ad hoc to paper The probability-flow ODE dx = -(dq/dt) ∇ log p(x;q) dt (Eq 10) governs the evolution of samples under this forward process.
    This ODE is borrowed from Gaussian diffusion theory; for uniform-noise corruption it is asserted without proof and is not a consequence of an underlying SDE for the compression process.
  • ad hoc to paper The score equals (Dθ(x,q)-x)/q, where Dθ is an L2-trained denoiser.
    The Gaussian score identity is (D-x)/σ^2; the analogous /q form is not justified for uniform noise. The entire reverse sampling relies on this.
  • ad hoc to paper The reverse SDE in Eq 12 with random injection α(t) dω ± α(t) ∇ log p and schedule α_t = β sqrt(q_t - q_min) is a valid stochastic sampler.
    The form of the SDE and the sign of the last term are not derived; β is tuned empirically on DIV2K.
  • domain assumption A pretrained VAE (Rombach et al. [36]) and the rate-variable entropy model from Han et al. [15] provide a suitable latent space.
    The method operates in the latent space of a pretrained autoencoder, so its quality depends on those components.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution." pith.science (2026). https://pith.science/paper/7DVTIKLT

@misc{pith2026250520984,
  author       = {Pith},
  title        = {Pith review of: Generative Image Compression by Estimating Gradients of the Rate-variable Feature Distribution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7DVTIKLT}},
  note         = {Machine review of arXiv:2505.20984}
}
read the original abstract

While learned image compression (LIC) focuses on efficient data transmission, generative image compression (GIC) extends this framework by integrating generative modeling to produce photo-realistic reconstructed images. In this paper, we propose a novel diffusion-based generative modeling framework tailored for generative image compression. Unlike prior diffusion-based approaches that indirectly exploit diffusion modeling, we reinterpret the compression process itself as a forward diffusion path governed by stochastic differential equations (SDEs). A reverse neural network is trained to reconstruct images by reversing the compression process directly, without requiring Gaussian noise initialization. This approach achieves smooth rate adjustment and photo-realistic reconstructions with only a minimal number of sampling steps. Extensive experiments on benchmark datasets demonstrate that our method outperforms existing generative image compression approaches across a range of metrics, including perceptual distortion, statistical fidelity, and no-reference quality assessments.

Figures

Figures reproduced from arXiv: 2505.20984 by the authors.

Figure 1
Figure 1. Left: The forward and reverse process of legacy diffusion for comparison. The legacy diffusion consists of transforming data to a simple noise distribution and a reverse ODE to restore the original data. Right: The overview pipeline of our method. The forward process is defined as the entropy model compressing the data. With the bit rates decrease, the compressed images retain less details (please zoom in for better… view at source ↗
Figure 2
Figure 2. Comparisons of methods across various distortion and statistical fidelity metrics for the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Evaluation of randomness injection schedules when the scale parameter [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Visualization of the reconstructed images (top to bottom: [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the reconstructed images ( [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Visualization of the reconstructed images ( [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Visualization of the reconstructed images ( [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Comparisons of methods across various metrics on the CLIC2020 dataset. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Comparisons of methods across various metrics on the Kodak dataset. [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 35 canonical work pages

  1. [1]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Agustsson, E., Minnen, D., Toderici, G., Mentzer, F.: Multi-realism image compression with a conditional generator. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 22324–22333 (2023) 4, 8

    work page 2023

  2. [2]

    In: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (July 2017) 7, 8

    Agustsson, E., Timofte, R.: Ntire 2017 challenge on single image super-resolution: Dataset and study. In: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (July 2017) 7, 8

    work page 2017

  3. [3]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Agustsson, E., Tschannen, M., Mentzer, F., Timofte, R., Gool, L.V .: Generative adversarial net- works for extreme learned image compression. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 221–231 (2019) 1, 4

    work page 2019

  4. [4]

    arXiv preprint arXiv:1611.01704 (2016) 1, 2, 3, 6

    Ballé, J., Laparra, V ., Simoncelli, E.P.: End-to-end optimized image compression. arXiv preprint arXiv:1611.01704 (2016) 1, 2, 3, 6

    arXiv 2016

  5. [5]

    arXiv preprint arXiv:1802.01436 (2018) 1, 2, 3, 6

    Ballé, J., Minnen, D., Singh, S., Hwang, S.J., Johnston, N.: Variational image compression with a scale hyperprior. arXiv preprint arXiv:1802.01436 (2018) 1, 2, 3, 6

    arXiv 2018

  6. [6]

    Bellard, F.: Bpg image format (2015), https://bellard.org/bpg 1, 8

    work page 2015

  7. [7]

    arXiv preprint arXiv:1801.01401 (2018) 8

    Bi´nkowski, M., Sutherland, D.J., Arbel, M., Gretton, A.: Demystifying mmd gans. arXiv preprint arXiv:1801.01401 (2018) 8

    arXiv 2018

  8. [8]

    In: The Twelfth International Conference on Learning Representations (2023) 1, 4

    Careil, M., Muckley, M.J., Verbeek, J., Lathuilière, S.: Towards image compression with perfect realism at ultra-low bitrates. In: The Twelfth International Conference on Learning Representations (2023) 1, 4

    work page 2023

Show all 48 references
  1. [9]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Cheng, Z., Sun, H., Takeuchi, M., Katto, J.: Learned image compression with discretized gaus- sian mixture likelihoods and attention modules. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 7939–7948 (2020) 1

    2020

  2. [10]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Choi, Y ., El-Khamy, M., Lee, J.: Variable rate deep image compression with a conditional autoencoder. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 3146–3154 (2019) 3

    2019

  3. [11]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Cui, Z., Wang, J., Gao, S., Guo, T., Feng, Y ., Bai, B.: Asymmetric gained deep image compression with continuous rate adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 10532–10541 (2021) 3

    2021

  4. [12]

    In: 2009 IEEE conference on computer vision and pattern recognition

    Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large-scale hierarchical image database. In: 2009 IEEE conference on computer vision and pattern recognition. pp. 248–255. Ieee (2009) 7

    2009

  5. [13]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Esser, P., Rombach, R., Ommer, B.: Taming transformers for high-resolution image synthesis. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 12873–12883 (2021) 1

    2021

  6. [14]

    Advances in neural information processing systems 27 (2014) 1

    Goodfellow, I.J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y .: Generative adversarial nets. Advances in neural information processing systems 27 (2014) 1

    2014

  7. [15]

    arXiv preprint arXiv:2410.04847 (2024) 1, 6, 7

    Han, M., Jiang, S., Li, S., Deng, X., Xu, M., Zhu, C., Gu, S.: Causal context adjustment loss for learned image compression. arXiv preprint arXiv:2410.04847 (2024) 1, 6, 7

    2024 arXiv

  8. [16]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    He, D., Yang, Z., Peng, W., Ma, R., Qin, H., Wang, Y .: Elic: Efficient learned image compression with unevenly grouped space-channel contextual adaptive coding. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 5718–5727 (2022) 1

    2022

  9. [17]

    Advances in neural information processing systems 30 (2017) 8 10

    Heusel, M., Ramsauer, H., Unterthiner, T., Nessler, B., Hochreiter, S.: Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems 30 (2017) 8 10

    2017

  10. [18]

    Advances in neural information processing systems 33, 6840–6851 (2020) 1

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. Advances in neural information processing systems 33, 6840–6851 (2020) 1

    2020

  11. [19]

    arXiv preprint arXiv:2305.18231 (2023) 1, 4

    Hoogeboom, E., Agustsson, E., Mentzer, F., Versari, L., Toderici, G., Theis, L.: High-fidelity image compression with score-based generative models. arXiv preprint arXiv:2305.18231 (2023) 1, 4

    2023 arXiv

  12. [20]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Jia, Z., Li, J., Li, B., Li, H., Lu, Y .: Generative latent coding for ultra-low bitrate image compression. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 26088–26098 (2024) 1, 4

    2024

  13. [21]

    Advances in neural information processing systems35, 26565–26577 (2022) 4, 5, 6, 7

    Karras, T., Aittala, M., Aila, T., Laine, S.: Elucidating the design space of diffusion-based generative models. Advances in neural information processing systems35, 26565–26577 (2022) 4, 5, 6, 7

    2022

  14. [22]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Ke, J., Wang, Q., Wang, Y ., Milanfar, P., Yang, F.: Musiq: Multi-scale image quality transformer. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 5148–5157 (2021) 8

    2021

  15. [23]

    Kodak, E.: Kodak lossless true color image suite (photocd pcd0992) (1993), http://r0k.us/ graphics/kodak 7

    1993

  16. [24]

    International journal of computer vision 128(7), 1956–1981 (2020) 7

    Kuznetsova, A., Rom, H., Alldrin, N., Uijlings, J., Krasin, I., Pont-Tuset, J., Kamali, S., Popov, S., Malloci, M., Kolesnikov, A., et al.: The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale. International jou...

    2020

  17. [25]

    arXiv preprint arXiv:2307.01944 (2023) 1, 4

    Lei, E., Uslu, Y .B., Hassani, H., Bidokhti, S.S.: Text+ sketch: Image compression at ultra low rates. arXiv preprint arXiv:2307.01944 (2023) 1, 4

    2023 arXiv

  18. [26]

    arXiv preprint arXiv:2406.00758 (2024) 1, 4, 8, 13

    Li, A., Li, F., Liu, Y ., Cong, R., Zhao, Y ., Bai, H.: Once-for-all: Controllable generative image compression with dynamic granularity adaption. arXiv preprint arXiv:2406.00758 (2024) 1, 4, 8, 13

    2024 arXiv

  19. [27]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Liu, J., Sun, H., Katto, J.: Learned image compression with mixed transformer-cnn architectures. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 14388–14397 (2023) 1

    2023

  20. [28]

    arXiv preprint arXiv:1711.05101 (2017) 7

    Loshchilov, I., Hutter, F.: Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101 (2017) 7

    2017 arXiv

  21. [29]

    arXiv preprint arXiv:2504.00496 (2025) 1

    Lu, J., Zhang, L., Zhou, X., Li, M., Li, W., Gu, S.: Learned image compression with dictionary- based entropy model. arXiv preprint arXiv:2504.00496 (2025) 1

    2025 arXiv

  22. [30]

    In: 2024 Data Compression Conference (DCC)

    Mao, Q., Yang, T., Zhang, Y ., Wang, Z., Wang, M., Wang, S., Jin, L., Ma, S.: Extreme image compression using fine-tuned vqgans. In: 2024 Data Compression Conference (DCC). pp. 203–212. IEEE (2024) 1

    2024

  23. [31]

    Advances in neural information processing systems 33, 11913–11924 (2020) 1, 4, 8, 9, 13

    Mentzer, F., Toderici, G.D., Tschannen, M., Agustsson, E.: High-fidelity generative image compression. Advances in neural information processing systems 33, 11913–11924 (2020) 1, 4, 8, 9, 13

    2020

  24. [32]

    Advances in neural information processing systems 31 (2018) 1

    Minnen, D., Ballé, J., Toderici, G.D.: Joint autoregressive and hierarchical priors for learned image compression. Advances in neural information processing systems 31 (2018) 1

    2018

  25. [33]

    In: 2020 IEEE International Conference on Image Processing (ICIP)

    Minnen, D., Singh, S.: Channel-wise autoregressive entropy models for learned image com- pression. In: 2020 IEEE International Conference on Image Processing (ICIP). pp. 3339–3343. IEEE (2020) 1

    2020

  26. [34]

    In: International Conference on Machine Learning

    Muckley, M.J., El-Nouby, A., Ullrich, K., Jégou, H., Verbeek, J.: Improving statistical fidelity for neural image compression with implicit local likelihood models. In: International Conference on Machine Learning. pp. 25426–25443. PMLR (2023) 1, 4, 8, 9, 13

    2023

  27. [35]

    arXiv preprint arXiv:2202.05492 (2022) 1 11

    Qian, Y ., Lin, M., Sun, X., Tan, Z., Jin, R.: Entroformer: A transformer-based entropy model for learned image compression. arXiv preprint arXiv:2202.05492 (2022) 1 11

    2022 arXiv

  28. [36]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022) 6

    2022

  29. [37]

    The Bell system technical journal 27(3), 379–423 (1948) 3

    Shannon, C.E.: A mathematical theory of communication. The Bell system technical journal 27(3), 379–423 (1948) 3

    1948

  30. [38]

    arXiv preprint arXiv:2010.02502 (2020) 5, 6

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020) 5, 6

    2020 arXiv

  31. [39]

    Advances in neural information processing systems 32 (2019) 1, 2, 3, 4, 5

    Song, Y ., Ermon, S.: Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems 32 (2019) 1, 2, 3, 4, 5

    2019

  32. [40]

    arXiv preprint arXiv:2011.13456 (2020) 2, 3, 4, 5, 6

    Song, Y ., Sohl-Dickstein, J., Kingma, D.P., Kumar, A., Ermon, S., Poole, B.: Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456 (2020) 2, 3, 4, 5, 6

    2020 arXiv

  33. [41]

    Team, J.V .E.: Versatile video coding reference software version 23.9(vtm-23.9) (2025),https: //vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/-/tags/VTM-23.9 1, 8, 13

    2025

  34. [42]

    arXiv preprint arXiv:1511.06085 (2015) 3

    Toderici, G., O’Malley, S.M., Hwang, S.J., Vincent, D., Minnen, D., Baluja, S., Covell, M., Sukthankar, R.: Variable rate image compression with recurrent neural networks. arXiv preprint arXiv:1511.06085 (2015) 3

    2015 arXiv

  35. [43]

    In: CVPR (2020) 7

    Toderici, G., Shi, W., Timofte, R., Theis, L., Ballé, J., Agustsson, E., Johnston, N., Mentzer, F.: Workshop and challenge on learned image compression (clic2020). In: CVPR (2020) 7

    2020

  36. [44]

    Advances in neural information processing systems 31 (2018) 1

    Tschannen, M., Agustsson, E., Lucic, M.: Deep generative models for distribution-preserving lossy compression. Advances in neural information processing systems 31 (2018) 1

    2018

  37. [45]

    arXiv preprint arXiv:2302.05071 (2023) 3

    Wang, G.H., Li, J., Li, B., Lu, Y .: Evc: Towards real-time neural image compression with mask decay. arXiv preprint arXiv:2302.05071 (2023) 3

    2023 arXiv

  38. [46]

    In: Proceedings of the AAAI conference on artificial intelligence

    Wang, J., Chan, K.C., Loy, C.C.: Exploring clip for assessing the look and feel of images. In: Proceedings of the AAAI conference on artificial intelligence. vol. 37, pp. 2555–2563 (2023) 8

    2023

  39. [47]

    Advances in Neural Information Processing Systems 36, 64971–64995 (2023) 1, 4, 8, 9, 13

    Yang, R., Mandt, S.: Lossy image compression with conditional diffusion models. Advances in Neural Information Processing Systems 36, 64971–64995 (2023) 1, 4, 8, 9, 13

    2023

  40. [48]

    method [bpp]

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 586–595 (2018) 8 12 A Derivation of Discrete ODE Solver A ...

    2018

Pith tools

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