Pith. sign in

REVIEW 3 major objections 5 minor 51 references

Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that one decoder-side latent diffusion module can make a fixed pretrained neural codec slide continuously between high-PSNR and high-perceptual-quality reconstructions at the same bitrate, using a user parameter τ to…

desk verdict Useful decoder-side plug-in for distortion-perception control on pretrained codecs, but the diffusion mechanism as written has a real training/inference mismatch that needs fixing before the claims hold. read the letter →

arxiv 2412.11379 v1 pith:AJKODFYM submitted 2024-12-16 eess.IV cs.CV

Chuqin Zhou , Guo Lu , Jiangchuan Li , Xiangyu Chen , Zhengxue Cheng , Li Song , Wenjun Zhang This is my paper · ORCID
classification eess.IVcs.CV
keywords neuralimagecompressionlatentdiffusiondistortion-perceptiontradeoffplug-and-playdecodermoduleLPIPS-BDRateadaptivefusionDDIMrate-distortion-perception
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 tries to establish that a fixed, already-trained neural image codec can be upgraded on the receiver side alone: one plug-and-play module runs a latent diffusion process on the decoded features and lets a user choose, at inference time, where to sit on the distortion–perception curve without sending a single extra bit. The central claim is that the same compressed latent can be decoded either as a high-PSNR reconstruction or as a perceptually richer one, with a single parameter τ controlling the blend. If true, codec users no longer have to fix their fidelity-versus-realism trade-off at training time; they can dial it after the bitstream is already in hand. The headline numbers are a 158.75% LPIPS-BDRate improvement over the Cheng2020 base codec at an average PSNR cost of 1.08 dB, with similar flexibility shown on the HiFiC base codec.

What carries the argument

The load-bearing object is the adaptive latent fusion module $D$, a latent diffusion network that operates directly on the compressed feature $\hat{y}$ rather than on pixels. It is conditioned by concatenating the decoded feature with the diffusion input, and it is trained to predict the perception-oriented feature $\bar{y}$ (produced by an auxiliary encoder $g'_a$ optimized with LPIPS against the frozen decoder) rather than predicting noise. Inference uses DDIM-style sampling with the trade-off parameter $\tau$; the update rule $\tilde{y}_{t-1} = \sqrt{\alpha_{t-1}}\left[(1-\tau^2)D(\tilde{y}_t,\hat{y},t/T)+\tau^2\hat{y}\right] + (1-\tau^2)\sqrt{1-\alpha_{t-1}}\,\epsilon(\tilde{y}_t,\hat{y},t/T)$ blends the diffusion output with the original latent, with $\tau^2$ used so the control feels linear. The fixed decoder $g_s$ then maps the fused latent to an image, so the entire base codec stays frozen and the bitstream remains unchanged.

What would settle it

One concrete test is to attach the module to a pretrained codec with a narrow or highly quantized latent manifold, train on a fixed dataset, and check whether the PSNR-versus-$\tau$ curve is monotone and whether $\tau=0$ yields clean images; if the diffusion output drifts off the decoder's trained manifold, artifacts appear and the claimed 158.75% LPIPS-BDRate gain at about 1 dB PSNR loss will not reproduce.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the distribution gap between distortion-optimized and perception-optimized latent features can be bridged in latent space by a diffusion module trained once against a frozen decoder, so the decoder output becomes a controllable interpolation. The method trains an auxiliary encoder $g'_a$ against the fixed decoder with an LPIPS loss to produce a perception-oriented target $\bar{y}$; the adaptive latent fusion module $D$ then learns to denoise the base codec's decoded feature $\hat{y}$ toward $\bar{y}$. At inference, DDIM sampling with a user parameter $\tau$ interpolates between the diffusion prediction and the original $\hat{y}$: $\tau=1$ reproduces the base codec exactly, and $\tau=0$ gives the fully perception-oriented reconstruction. On Kodak, the paper reports a $158.75\%$ LPIPS-BDRate improvement with an average $1.08$ dB PSNR drop on Cheng2020, matches or exceeds HiFiC on DISTS at $\tau=0$ while keeping a $1.48$ dB PSNR advantage, and saves $22.59\%$ bitrate for equal PSNR when applied to HiFiC.

Load-bearing premise

The method assumes that the original decoder remains able to turn the diffusion-transformed feature codes into good images, and that a separately trained perception-oriented encoder produces a target those transformed codes can genuinely reach; if either fails, the whole trade-off dial stops working.

Editorial extensions

If this is right

  • A single trained module can be attached to an existing distortion-oriented codec (Cheng2020) and, at $\tau=0$, improve LPIPS-BDRate by 158.75% while costing only 1.08 dB PSNR on average.
  • The same module attached to a perception-oriented codec (HiFiC) moves the operating point the other way, saving 22.59% bitrate for equal PSNR at $\tau=0$.
  • Because the base codec is frozen and the bitstream is unchanged, users can decode the same file several times with different $\tau$ values, effectively choosing sharpness versus realism after transmission.
  • Combined with a variable-bitrate scheme (MSHyper), the method produces a three-dimensional rate–distortion–perception operating range with a smooth, near-linear $\tau$ traversal, unlike MRIC's less controllable $\beta$ sweep.
  • The module works with as few as 10 DDIM steps, keeping inference latency at 331 ms per Kodak image on a 4090 GPU, well below pixel-domain diffusion codecs.

Reading between the lines

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

  • Our inference: because the auxiliary encoder is trained without any rate constraint, the perception-oriented target can be richer than the bitstream would normally pay for; if that holds, the method effectively shifts some decoder-side computation into perceptual enhancement that costs no extra bits.
  • Our inference: the same latent-space interpolation idea could extend to other conditional targets, such as semantic maps, text prompts, or restoration cues, letting a receiver choose content-specific enhancements from one transmitted bitstream.
  • Our inference: a stress point is whether the frozen decoder remains on its trained manifold for fused latents at intermediate $\tau$; the paper's stability runs show small variance, but applying the module to codecs with very narrow or strongly quantized latent manifolds could break the monotone $\tau$ control.
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

3 major / 5 minor

Summary. The paper proposes a plug-and-play decoder-side module for pretrained neural image codecs that uses a latent diffusion process to transform decoded latent features, enabling a user-controlled distortion-perception trade-off without changing the bitstream. An auxiliary encoder is trained to produce a perception-oriented target latent, and an adaptive latent fusion module is trained with an LPIPS-based loss to map from the base codec's decoded latent to that target. At inference, the module is applied inside a DDIM-style sampling recurrence with a trade-off parameter τ that interpolates between the transformed and original latents. The method is evaluated on Kodak and CLIC2020 with Cheng2020, HiFiC, and MSHyper as base codecs, reporting improvements in LPIPS-BDRate, FID, CLIP-IQA, NIQE, and a user study.

Significance. If the technical foundation is sound, this is a practically valuable contribution: a single trained module can give a pretrained codec a wide, adjustable distortion-perception range at inference time, with no modification to the codec or the bitstream. The paper is strong on the empirical side: it reports eight image-quality metrics on two benchmark datasets, includes a user study, measures inference latency, and uses publicly available pretrained models for reproducibility. The authors also provide an ablation of the diffusion module and a stability analysis. However, the central claim that the method is a latent diffusion process is undermined by a mismatch between the training objective and the inference procedure, and the headline PSNR number in the abstract is inconsistent with the reported result. The paper's core idea remains interesting, but the technical story needs to be repaired before the results can be fully trusted.

major comments (3)
  1. [Adaptive Latent Fusion, Eq. (2) and Eqs. (3)-(4)] The training objective trains the module D to predict the auxiliary perceptual latent y_bar from a noised condition y_hat_t = sqrt(alpha_t) y_hat + sqrt(1-alpha_t) epsilon, i.e., from a noisy version of the base codec's decoded feature. At inference, however, D is used inside the DDIM recurrence as the posterior-mean predictor of y_bar from y_tilde_t, which is a noisy version of the target y_bar itself and is initialized as y_tilde_T ~ N(0,I). These are different input distributions; for small t the training inputs concentrate around y_hat, while the inference inputs concentrate around y_bar. The statement that 'we directly learn transformed features y_tilde instead of the noise' does not reconcile this mismatch. As a result, the claimed latent-diffusion interpretation is not the reverse of the training forward process, and the stability of the sampling procedure is not guaranteed by Eq. (2). Please either reformulate the training to noise the actual target y_bar (as in standard DDPM/DDIM) or provide explicit evidence (e.g., an ablation training D on noised y_bar or a distribution-shift analysis) that the current training yields a valid posterior-mean predictor for the inference-time inputs.
  2. [Abstract and Main Results] The abstract's headline claim 'without sacrificing more than 1 dB in PSNR' is contradicted by the reported Cheng2020 result in the Main Results section, which states an average PSNR degradation of 1.08 dB. Since 1.08 dB exceeds 1 dB, the abstract's quantitative claim is inaccurate as written. Please correct the abstract to 'about 1.1 dB' or adjust the reported number so the headline assertion is consistent with the experimental results.
  3. [Proposed Method and Main Results] The training pipeline (auxiliary encoder g'_a and Eq. (2)) is formulated under the explicit assumption that the base codec is distortion-oriented. The paper nonetheless reports results for Ours(HiFiC), a perception-oriented base codec, and claims the method can transform features in both directions. The paper never specifies how g'_a is trained, what the target y_bar represents, or how Eq. (2) is adapted for a perception-oriented base codec. Without this information, the HiFiC results are not reproducible and the claimed generality across codec types is unsupported. Please add the training equations and implementation details for perception-oriented base codecs.
minor comments (5)
  1. [Main Results] The phrase 'On the low-distortion side ( τ = 0 )' appears to be a typo; τ = 0 corresponds to the perception-oriented end of the trade-off, not the low-distortion side.
  2. [Rate-Distortion-Perception Trade-off] The text 'Their method shows little or no change from β = 0.64 to β = 2.56' does not match the β values shown in Fig. 8 (0, 0.08, 2.56); please correct the reference.
  3. [Adaptive Latent Fusion, Training] The sentence 'We disregard noise level t and directly input the result of D at each step into the decoder g_s.' is ambiguous; please clarify that the decoder is used during training to compute the image-domain loss term in Eq. (2).
  4. [Figures] The submitted PDF contains garbled character sequences in and around the figure captions (e.g., immediately after Figure 6 and within the caption of Figure 7); these should be cleaned in the final version.
  5. [Evaluation] Because Eq. (2) directly optimizes LPIPS and the headline result is LPIPS-BDRate, the paper should explicitly acknowledge that the LPIPS improvement is partially an outcome of optimizing the evaluation metric; the non-LPIPS metrics (FID, CLIP-IQA, NIQE, user study) provide independent evidence and should be emphasized accordingly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the LPIPS training/evaluation overlap and DDIM training/inference mismatch are correctness caveats, not circular reductions.

full rationale

Walked the claimed derivation chain: (i) an auxiliary encoder g'_a is trained with perceptual (LPIPS) loss through the frozen decoder to produce a perception-oriented target y_bar; (ii) the adaptive latent fusion module D is trained with Eq. (2) to map the noised decoded condition to y_bar and to minimize LPIPS of the decoded output; (iii) at inference, Eq. (4) interpolates between D's output and the original latent y_hat under user parameter tau. No load-bearing step defines its output as its own input. The tau = 1 endpoint exactly recovers the base codec by Eq. (4), but this is a sanity endpoint, not the headline claim. The headline LPIPS-BDRate improvement is a held-out measurement on Kodak and CLIC2020 against external pretrained baselines; although Eq. (2) contains an LPIPS reconstruction term, the test images were not used for fitting, and the central claims are corroborated by metrics not present in the training loss (PSNR, SSIM, DISTS, FID, CLIP-IQA, NIQE, and a user study), so the result is externally falsifiable rather than forced. The paper's citations to latent diffusion and DDIM are standard external works (Rombach et al. 2022; Song et al. 2021), not a self-citation chain, and no uniqueness theorem or ansatz is imported from the authors' own prior work. The review concern that the DDIM sampling distribution is not the reverse of the training forward process is a correctness and generalization caveat, not circularity: it identifies a mismatch rather than an equation reducing to itself. No significant circularity found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper's central claim rests on standard ML training assumptions, domain assumptions about perceptual metrics and generalization, and two paper-specific assumptions about the compatibility of a fixed decoder with features produced by an auxiliary encoder and by latent diffusion. The controllable parameter τ, the auxiliary encoder, and the fusion module are new design components rather than new physical entities.

free parameters (4)
  • λ (loss balance in Eq. 2) = not reported
    Balances the latent regression loss against the LPIPS image reconstruction loss during training; the value changes the trained module and is not stated.
  • inference diffusion timesteps = 10
    Chosen as the default based on the speed-quality tradeoff in Fig. 9; not derived.
  • τ² interpolation weighting = τ in [0,1]
    The squared weighting is a hand-selected curve shape to make the tradeoff appear linear; no derivation is given.
  • noise schedule β_t and T = not reported
    The paper states a monotonically increasing β_t but does not give the schedule values used in training or inference.
assumptions (5)
  • standard math DDPM/DDIM theory and latent diffusion sampling equations are valid and transfer to compressed latent spaces.
    The paper relies on Ho et al., Song et al., and Rombach et al. for the diffusion framework.
  • domain assumption LPIPS (and the other chosen metrics) approximates human perceptual quality well enough to serve as both training loss and evaluation.
    The method optimizes LPIPS and then evaluates LPIPS as evidence of perceptual improvement.
  • ad hoc to paper A fixed pretrained decoder can decode features produced by a separately trained auxiliary encoder and by the diffusion module.
    The paper does not prove this; it depends on the decoder's latent manifold being broad enough.
  • domain assumption Training on Flickr2W random 256x256 crops transfers to full-resolution Kodak and CLIC2020 images.
    The module is trained at one resolution and evaluated on full-resolution datasets.
  • ad hoc to paper The auxiliary encoder's perceptual latent y_bar is a valid regression target for the diffusion module starting from y_hat.
    No argument is given that y_bar is reachable from y_hat through the diffusion process; this is an empirical assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression." pith.science (2026). https://pith.science/paper/AJKODFYM

@misc{pith2026241211379,
  author       = {Pith},
  title        = {Pith review of: Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AJKODFYM}},
  note         = {Machine review of arXiv:2412.11379}
}
read the original abstract

Neural image compression often faces a challenging trade-off among rate, distortion and perception. While most existing methods typically focus on either achieving high pixel-level fidelity or optimizing for perceptual metrics, we propose a novel approach that simultaneously addresses both aspects for a fixed neural image codec. Specifically, we introduce a plug-and-play module at the decoder side that leverages a latent diffusion process to transform the decoded features, enhancing either low distortion or high perceptual quality without altering the original image compression codec. Our approach facilitates fusion of original and transformed features without additional training, enabling users to flexibly adjust the balance between distortion and perception during inference. Extensive experimental results demonstrate that our method significantly enhances the pretrained codecs with a wide, adjustable distortion-perception range while maintaining their original compression capabilities. For instance, we can achieve more than 150% improvement in LPIPS-BDRate without sacrificing more than 1 dB in PSNR.

Figures

Figures reproduced from arXiv: 2412.11379 by the authors.

Figure 1
Figure 1. Overview of our proposed method. D represents a plug-and-play adaptive latent fusion module at decoder side for a base neural codec. We can achieve different distortion (PSNR) and perception (LPIPS) trade-offs, controlled by τ . For simplicity, quantization and entropy coding are omitted. bitrate. For example, HiFiC (Mentzer et al. 2020) proposes to use Generative Adversarial Networks (GANs) (Goodfel￾low et al. 2014… view at source ↗
Figure 1
Figure 1. Specifically, we introduce a plug-and-play adaptive [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the proposed method. For simplicity, we assume the base NIC is distortion-oriented. (a) represents the [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figures from the paper (19 more)
Figure 3
Figure 3. Figure 3: Overview of the latent diffusion process. For sim [PITH_FULL_IMAGE:figures/full_fig_p004_3.png]
Figure 5
Figure 5. Figure 5: Trade-offs between bitrate and different metrics for various base codecs tested on Kodak dataset. Arrows in the plot [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Kodak reconstructions of our method for different rate-distortion-perception. Shown scores are for full image. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Trade-offs between bitrate and different metrics for [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 9
Figure 9. Figure 9: Reconstruction quality in the terms of PSNR and [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]
Figure 10
Figure 10. Figure 10: Comparison of PSNR values with varying hyper [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]
Figure 1
Figure 1. Figure 1: Trade-offs between bitrate and different metrics for various models tested on Kodak dataset. Arrows in the plot titles [PITH_FULL_IMAGE:figures/full_fig_p011_1.png]
Figure 2
Figure 2. Figure 2: Distortion (PSNR) vs. perception(LPIPS) on Kodak for different rate-distortion-perception tradeoffs. [PITH_FULL_IMAGE:figures/full_fig_p012_2.png]
Figure 3
Figure 3. Figure 3: Visualization of original image x, compressed feature yˆ and residual with transformed feature y˜. For better observa￾tion, feature and residual are normalized separately [PITH_FULL_IMAGE:figures/full_fig_p012_3.png]
Figure 4
Figure 4. Figure 4: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p016_11.png]
Figure 12
Figure 12. Figure 12: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p017_12.png]
Figure 13
Figure 13. Figure 13: Reconstructions of Ours(Cheng2020) with different [PITH_FULL_IMAGE:figures/full_fig_p017_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 39 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Agustsson, E.; Minnen, D.; Toderici, G.; and Mentzer, F. 2023. Multi-Realism Image Compression with a Conditional Generator. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 22324--22333

    work page 2023

  4. [4]

    J.; and Johnston, N

    Ball \' e , J.; Minnen, D.; Singh, S.; Hwang, S. J.; and Johnston, N. 2018. Variational image compression with a scale hyperprior. In 6th International Conference on Learning Representations(ICLR)

    work page 2018

  5. [5]

    B \'e gaint, J.; Racap \'e , F.; Feltman, S.; and Pushparaja, A. 2020. CompressAI: a PyTorch library and evaluation platform for end-to-end compression research. arXiv preprint arXiv:2011.03029

    arXiv 2020

  6. [6]

    Bellard, F. 2018. Bpg image format, https://bellard.org/bpg/

    work page 2018

  7. [7]

    Blau, Y.; Mechrez, R.; Timofte, R.; Michaeli, T.; and Zelnik - Manor, L. 2018. The 2018 PIRM Challenge on Perceptual Image Super-Resolution. In Computer Vision - ECCV 2018 Workshops , volume 11133 of Lecture Notes in Computer Science, 334--355. Springer

    work page 2018

  8. [8]

    Blau, Y.; and Michaeli, T. 2019. Rethinking Lossy Compression: The Rate-Distortion-Perception Tradeoff. In Proceedings of the 36th International Conference on Machine Learning(ICML), 675--685. PMLR

    work page 2019

Show all 51 references
  1. [9]

    J.; and Ohm, J.-R

    Bross, B.; Wang, Y.-K.; Ye, Y.; Liu, S.; Chen, J.; Sullivan, G. J.; and Ohm, J.-R. 2021. Overview of the Versatile Video Coding (VVC) Standard and its Applications. IEEE Transactions on Circuits and Systems for Video Technology, 31(10): 3736--3764

    2021

  2. [10]

    J.; Verbeek, J.; and Lathuili \`e re, S

    Careil, M.; Muckley, M. J.; Verbeek, J.; and Lathuili \`e re, S. 2023. Towards image compression with perfect realism at ultra-low bitrates. In The Twelfth International Conference on Learning Representations(ICLR)

    2023

  3. [11]

    Cheng, Z.; Sun, H.; Takeuchi, M.; and Katto, J. 2020. Learned Image Compression With Discretized Gaussian Mixture Likelihoods and Attention Modules. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR)

    2020

  4. [12]

    Cui, Z.; Wang, J.; Gao, S.; Guo, T.; Feng, Y.; and Bai, B. 2021. Asymmetric Gained Deep Image Compression With Continuous Rate Adaptation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 10532--10541

    2021

  5. [13]

    F.; Petersen, J.; Wiggers, A.; Xu, T.; and Sautiere, G

    Ghouse, N. F.; Petersen, J.; Wiggers, A.; Xu, T.; and Sautiere, G. 2023. A residual diffusion model for high perceptual quality codec augmentation. arXiv preprint arXiv:2301.05489

    2023 arXiv

  6. [14]

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

    2014

  7. [15]

    He, D.; Zheng, Y.; Sun, B.; Wang, Y.; and Qin, H. 2021. Checkerboard Context Model for Efficient Learned Image Compression. In IEEE Conference on Computer Vision and Pattern Recognition(CVPR) , 14771--14780. Computer Vision Foundation / IEEE

    2021

  8. [16]

    Heusel, M.; Ramsauer, H.; Unterthiner, T.; Nessler, B.; and Hochreiter, S. 2017. GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium. In Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Syste...

    2017

  9. [17]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising Diffusion Probabilistic Models. In Advances in Neural Information Processing Systems(NeurIPS) 33

    2020

  10. [18]

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

    2023 arXiv

  11. [19]

    Hu, Z.; Lu, G.; and Xu, D. 2021. FVC: A New Framework Towards Deep Video Compression in Feature Space. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 1502--1511

    2021

  12. [20]

    Iwai, S.; Miyazaki, T.; and Omachi, S. 2024. Controlling rate, distortion, and realism: Towards a single comprehensive neural image compression model. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2900--2909

    2024

  13. [21]

    Li, J.; Li, B.; and Lu, Y. 2023. Neural Video Compression with Diverse Contexts. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 22616--22626

    2023

  14. [22]

    Lin, X.; He, J.; Chen, Z.; Lyu, Z.; Fei, B.; Dai, B.; Ouyang, W.; Qiao, Y.; and Dong, C. 2023. DiffBIR: Towards Blind Image Restoration with Generative Diffusion Prior. arXiv preprint arXiv:2308.15070

    2023 arXiv

  15. [23]

    Liu, J.; Lu, G.; Hu, Z.; and Xu, D. 2020. A Unified End-to-End Framework for Efficient Deep Image Compression. arXiv preprint arXiv:2002.03370

    2020 arXiv

  16. [24]

    Liu, L.; Zhao, M.; Yuan, S.; Lyu, W.; Zhou, W.; Li, H.; Wang, Y.; and Tian, Q. 2023. Exploring Effective Mask Sampling Modeling for Neural Image Compression. arXiv preprint arXiv: 2306.05704

    2023 arXiv

  17. [25]

    Lu, G.; Ge, X.; Zhong, T.; Hu, Q.; and Geng, J. 2024 a . Preprocessing Enhanced Image Compression for Machine Vision. IEEE Transactions on Circuits and Systems for Video Technology

    2024

  18. [26]

    Lu, G.; Ouyang, W.; Xu, D.; Zhang, X.; Cai, C.; and Gao, Z. 2019. DVC: An End-To-End Deep Video Compression Framework. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 11006--11015

    2019

  19. [27]

    Lu, G.; Zhang, X.; Ouyang, W.; Chen, L.; Gao, Z.; and Xu, D. 2021. An End-to-End Learning Framework for Video Compression. IEEE Trans. Pattern Anal. Mach. Intell. , 43(10): 3292--3308

    2021

  20. [28]

    Lu, H.; Yang, G.; Fei, N.; Huo, Y.; Lu, Z.; Luo, P.; and Ding, M. 2024 b . VDT: General-purpose Video Diffusion Transformers via Mask Modeling. In The Twelfth International Conference on Learning Representations(ICLR)

    2024

  21. [29]

    Luo, X.; Xie, Y.; Qu, Y.; and Fu, Y. 2024. SkipDiff: Adaptive Skip Diffusion Model for High-Fidelity Perceptual Image Super-resolution. In Thirty-Eighth AAAI Conference on Artificial Intelligence(AAAI) , 4017--4025

    2024

  22. [30]

    Mentzer, F.; Toderici, G.; Tschannen, M.; and Agustsson, E. 2020. High-Fidelity Generative Image Compression. In Advances in Neural Information Processing Systems(NeurIPS) 33

    2020

  23. [31]

    Minnen, D.; Ball \' e , J.; and Toderici, G. 2018. Joint Autoregressive and Hierarchical Priors for Learned Image Compression. In Advances in Neural Information Processing Systems(NeurIPS) 31, 10794--10803

    2018

  24. [32]

    K.; and Bovik, A

    Mittal, A.; Moorthy, A. K.; and Bovik, A. C. 2011. Blind/Referenceless Image Spatial Quality Evaluator. In Conference Record of the Forty Fifth Asilomar Conference on Signals, Systems and Computers(ACSCC), 723--727. IEEE

    2011

  25. [33]

    Completely Blind

    Mittal, A.; Soundararajan, R.; and Bovik, A. C. 2013. Making a "Completely Blind" Image Quality Analyzer. IEEE Signal Process. Lett. , 20(3): 209--212

    2013

  26. [34]

    J.; El - Nouby, A.; Ullrich, K.; J \' e gou, H.; and Verbeek, J

    Muckley, M. J.; El - Nouby, A.; Ullrich, K.; J \' e gou, H.; and Verbeek, J. 2023. Improving Statistical Fidelity for Neural Image Compression with Implicit Local Likelihood Models. In International Conference on Machine Learning(ICML), volume 202, 25426--25443. PMLR

    2023

  27. [35]

    Pan, Z.; Zhou, X.; and Tian, H. 2022. Extreme generative image compression by learning text embedding from diffusion models. arXiv preprint arXiv:2211.07793

    2022 arXiv

  28. [36]

    Podell, D.; English, Z.; Lacey, K.; Blattmann, A.; Dockhorn, T.; M \" u ller, J.; Penna, J.; and Rombach, R. 2024. SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis. In The Twelfth International Conference on Learning Representations(ICLR)

    2024

  29. [37]

    Prashnani, E.; Cai, H.; Mostofi, Y.; and Sen, P. 2018. PieAPP: Perceptual Image-Error Assessment Through Pairwise Preference. In 2018 IEEE Conference on Computer Vision and Pattern Recognition(CVPR) , 1808--1817

    2018

  30. [38]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-Resolution Image Synthesis with Latent Diffusion Models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 10674--10685

    2022

  31. [39]

    Song, J.; Meng, C.; and Ermon, S. 2021. Denoising Diffusion Implicit Models. In 9th International Conference on Learning Representations(ICLR)

    2021

  32. [40]

    J.; Ohm, J.; Han, W.; and Wiegand, T

    Sullivan, G. J.; Ohm, J.; Han, W.; and Wiegand, T. 2012. Overview of the High Efficiency Video Coding (HEVC) Standard. IEEE Transactions on Circuits and Systems for Video Technology, 22(12): 1649--1668

    2012

  33. [41]

    Wallace, G. K. 1991. The JPEG Still Picture Compression Standard. Communication ACM, 34(4): 30--44

    1991

  34. [42]

    Wang, J.; Chan, K. C. K.; and Loy, C. C. 2023. Exploring CLIP for Assessing the Look and Feel of Images. In Thirty-Seventh AAAI Conference on Artificial Intelligence(AAAI) , 2555--2563

    2023

  35. [43]

    C.; and Loy, C

    Wang, J.; Yue, Z.; Zhou, S.; Chan, K. C.; and Loy, C. C. 2023. Exploiting diffusion prior for real-world image super-resolution. arXiv preprint arXiv:2305.07015

    2023 arXiv

  36. [44]

    Yan, Z.; Wen, F.; and Liu, P. 2022. Optimally Controllable Perceptual Lossy Compression. In International Conference on Machine Learning(ICML), volume 162 of Proceedings of Machine Learning Research, 24911--24928. PMLR

    2022

  37. [45]

    Yang, R.; and Mandt, S. 2023. Lossy Image Compression with Conditional Diffusion Models. In Advances in Neural Information Processing Systems(NeurIPS) 36

    2023

  38. [46]

    Yu, F.; Gu, J.; Li, Z.; Hu, J.; Kong, X.; Wang, X.; He, J.; Qiao, Y.; and Dong, C. 2024. Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild. arXiv preprint arXiv:2401.13627

    2024 arXiv

  39. [47]

    Zhang, G.; Qian, J.; Chen, J.; and Khisti, A. 2021. Universal Rate-Distortion-Perception Representations for Lossy Compression. In Advances in Neural Information Processing Systems(NeurIPS) 34, 11517--11529

    2021

  40. [48]

    A.; Shechtman, E.; and Wang, O

    Zhang, R.; Isola, P.; Efros, A. A.; Shechtman, E.; and Wang, O. 2018. The Unreasonable Effectiveness of Deep Features as a Perceptual Metric. In IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR) , 586--595

    2018

  41. [49]

    Zhao, S.; Song, J.; and Ermon, S. 2017. Towards Deeper Understanding of Variational Autoencoding Models. arXiv preprint arXiv:1702.08658

    2017 arXiv

  42. [50]

    Zheng, H.; and Gao, W. 2024. End-to-End RGB-D Image Compression via Exploiting Channel-Modality Redundancy. In Thirty-Eighth AAAI Conference on Artificial Intelligence(AAAI) , 7562--7570

    2024

  43. [51]

    Zhu, Y.; Yang, Y.; and Cohen, T. 2022. Transformer-based Transform Coding. In The Tenth International Conference on Learning Representations(ICLR)

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.