Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A learned image codec that spreads channel energy evenly across packets gains 1.84 dB over the previous loss-resilient method when 20% of packets are lost.

desk verdict Solid systems paper with a real contribution, but the headline 20% loss numbers quietly assume the hyperprior packet never dies; the disclosed FEC doesn't fully fix that at 20% uniform loss. read the letter →

arxiv 2608.11096 v1 pith:74MFCAA6 submitted 2026-08-11 cs.CV

Yuhang Wei (1) , Chuqin Zhou (1) , Yibo Shi (2) , Jing Wang (2) , Guo Lu (1) ((1) Shanghai Jiao Tong University , (2) Huawei Technologies Ltd.) This is my paper · ORCID
classification cs.CV
keywords imagecompressionlearnedpacketlossresilienceneuralnetworkchannelgroupingautoregressiveentropymodelhyperprior
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

Learned image compression degrades sharply under packet loss because critical information concentrates in a few packets and entropy decoding depends on earlier bits. This paper proposes a scheme that disperses channel energy across packets before transmission and shortens the autoregressive dependency chain to two layers, so reconstruction quality degrades gracefully no matter which packets are lost. At 20% packet loss the method gains 1.84 dB over the previous state of the art while cutting PSNR variance by an order of magnitude. It is trained under uniform random loss only, yet it generalizes to bursty loss from a Gilbert–Elliott channel and outperforms methods trained specifically for that condition. The headline numbers assume the small hyperprior bitstream arrives losslessly, which the paper argues is cheap to guarantee with Reed–Solomon coding.

What carries the argument

The load-bearing machinery is the trio of Inter-Channel Redistribution (ICR), Interleaved Channel Grouping (ICG), and a two-layer dual-branch autoregressive model. ICR is an attention-plus-shuffle module that evens out channel energy before packetization, with an inverse module at the decoder; ICG is a strided channel-partition rule that makes every packet comparable in importance while respecting packet-size limits; and the two-layer dual-branch autoregression predicts the second latent slice from the first while keeping branches independent, so a lost packet degrades only its own branch. Together they remove both causes of fragility: no packet is uniquely critical, and decoding dependencies do not cascade.

What would settle it

Run the paper's 20% uniform-loss protocol on the Kodak dataset with the hyperprior packet included in the loss process instead of protected: if dropping that single packet collapses PSNR by more than the claimed 1.84 dB margin (or fails to decode entirely), then the lossless-hyperprior assumption, rather than the dispersal scheme, is carrying the result.

Watch

Extended reading notes

Core claim

The paper's central claim is that packet loss in learned image compression can be largely neutralized by making the information content of every packet roughly equal, rather than by adding redundancy or retransmission. Three components achieve this: Inter-Channel Redistribution uses attention and channel shuffling to spread the energy that would otherwise sit in a few high-importance channels; Interleaved Channel Grouping assigns channels to packets in a strided pattern so each packet carries a comparable share of the information; and a two-layer dual-branch autoregressive model keeps decoding dependencies short and confines the impact of a lost packet to a single branch. Trained with packet-level masking under uniform random loss, the model preserves high mean PSNR with very low variance across loss patterns, and it transfers to bursty loss without retraining. The scheme assumes the hyperprior bitstream — the side stream carrying scale estimates — is received intact; with roughly 7% bandwidth overhead from Reed–Solomon coding, the authors argue this is a practical assumption.

Load-bearing premise

The reported gains assume the hyperprior bitstream — the small side stream carrying scale estimates for the main latent — arrives losslessly; if that stream is lost, the distribution estimates are corrupted and the stated robustness numbers do not apply.

Editorial extensions

If this is right

  • At 20% packet loss, reconstruction quality improves by 1.84 dB over the previous loss-resilient codec at similar bitrate, with PSNR variance about one order of magnitude lower.
  • Uniform-random-loss training transfers to Gilbert–Elliott bursty loss, beating methods trained specifically on that bursty model.
  • The hyperprior stream can be protected by RS(12,8) FEC at roughly 7% bandwidth overhead, preserving the lossless-side-channel assumption at low cost.
  • Because every packet carries comparable importance, losing any single packet costs at most about 1.5 dB in the tested case, instead of breaking the whole decode.
  • The method works under both 900-byte and 4500-byte packet-size constraints, whereas progressive baselines degrade sharply under the smaller packet size.

Reading between the lines

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

  • If the hyperprior remains the only packet that must be protected, an obvious next step is to fold it into the dispersal scheme itself rather than guarding it with FEC, trading a little rate for end-to-end robustness.
  • The ICR-and-ICG design could transfer to learned video compression, where packet loss is equally critical and the dependency chain is even longer.
  • Because the method's stability comes from equal-importance packets, one testable prediction is that worst-case (e.g., 1st-percentile) PSNR improves even more than the mean, which matters for emergency links where a single failed image can be decisive.
  • The model trained only on uniform loss outperforming bursty-trained methods suggests that a broader principle — training on the weakest, most memoryless loss model — may suffice for channel-agnostic robustness; testing this on other bursty models with longer bursts would confirm it.
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 / 4 minor

Summary. The paper proposes a learned image compression system designed to survive packet loss. Three mechanisms are introduced: Inter-Channel Redistribution (ICR) to homogenize channel energy before packetization, Interleaved Channel Grouping (ICG) to disperse latent channels across packets under size constraints, and a two-layer dual-branch autoregressive entropy model that shortens decoding dependency chains. Training uses structured packet-level masking, including propagation of losses across the two autoregressive layers. The scheme is evaluated on Kodak and CLIC under uniform packet loss at 5%, 10%, and 20%, and under a Gilbert-Elliott bursty-loss model, reporting mean PSNR, variance over ten trials, and comparisons against JPEG2000, ProgDTD, LossResilientLIC, and ResiComp. The central claim is that the method achieves state-of-the-art loss-resilient performance, with a headline 1.84 dB average PSNR gain over LossResilientLIC at 20% packet loss and an order-of-magnitude reduction in PSNR variance.

Significance. If the results hold, the paper makes a useful contribution to learned image compression under unreliable channels. The ablations are careful and include a range of design alternatives (partition strategies, autoregressive depth, training stages, masking propagation), and the evaluation reports variance over repeated random loss trials rather than a single draw, which is a strength. The complexity analysis in Appendix F is also valuable, showing a large FLOP reduction relative to ResiComp. However, the significance is constrained by a load-bearing caveat: the headline packet-loss numbers condition on the hyperprior bitstream being received losslessly, and the proposed RS(12,8) protection does not make this assumption true at the advertised 20% uniform-loss operating point. The paper is transparent about this assumption, but as currently framed the abstract overstates the robustness claim, and the effective rate of the protected system is not included in the comparisons.

major comments (3)
  1. [Section 3.1, Section 4.1, Appendix C] The headline claim in the abstract—“at 20% packet loss, it achieves an average PSNR gain of 1.84 dB over LossResilientLIC”—is evaluated under the assumption, stated in Section 3.1, that the hyperprior bitstream is lossless. Section 4.1 says the hyperprior is encapsulated as a standalone packet and must be reliably received. At 20% independent uniform loss, an unprotected single-packet hyperprior is lost with probability 20%, and the RS(12,8) scheme of Appendix C leaves roughly a 7% residual failure probability under independent erasures (about 93% success across 12 packets), not the “>99%” figure, which is only claimed for the GE model. The approximately 7% FEC overhead is also not included in the reported bitrates in Tables 1 and 2. Please re-report the end-to-end performance either by including the hyperprior packet in the loss process, or by explicitly re-labeling the results as conditioned on protected side information and showing the total rate with protection. A concrete test: report mean PSNR and variance over trials that include hyperprior loss, with and without RS protection, at 20% uniform loss.
  2. [Section 4.2, Table 1] The text states that at a 10% packet loss rate the method outperforms ResiComp by 0.48 dB, 0.33 dB, and 0.73 dB at low, medium, and high bitrates. Table 1 gives the corresponding numbers as 27.470 − 26.991 = 0.479 dB, 28.845 − 28.516 = 0.329 dB, and 30.271 − 30.232 = 0.039 dB. The high-bitrate gain is therefore 0.039 dB, not 0.73 dB, and at the 5% high-bitrate operating point Table 1 shows the method is actually 0.028 dB below ResiComp. This is not a presentation nuance; it directly affects the claim of consistent superiority across bitrate regimes. Either correct the text or the table, and re-word the summary of the comparison accordingly.
  3. [Section 4.2, Appendix E.2] The comparison against LossResilientLIC is not performed under a fully matched protocol. Section 4.1 specifies a 1500-byte packet size for the main evaluation, while Appendix E.2 states that the CLIC comparison follows LossResilientLIC’s setup with a 4500-byte packet size “to ensure fair comparisons.” Section 4.2 says the Kodak LossResilientLIC results are sourced from the original paper. Since packet size and the packet-loss simulation protocol determine how many channels are lost per packet and how those losses propagate, the headline 1.84 dB gain may conflate the proposed method’s robustness with a difference in evaluation protocol. Please either run LossResilientLIC under the same packetization and loss simulator used for the other baselines, or explicitly state the original paper’s settings and demonstrate that the comparison is unaffected by the protocol mismatch.
minor comments (4)
  1. [Section 4.4] The word “Dispite” should be “Despite” in the sentence “Dispite no GE-base simulation during training.”
  2. [Section 4.1 and Appendix E.2] The packet-size configuration is not consistently stated: the main text says 1500 bytes, while Appendix E.2 introduces 4500-byte and 900-byte settings. Please label each table with the packet size used so the reader can track which results correspond to which setting.
  3. [Captions of Figures 4, 11, and 12] Figure 4’s caption says “All methods lose the first two packets,” whereas Figures 11 and 12 add “while in our method, the packet of y3 is also lost due to the autoregressive dependency.” Harmonize the captions so the loss patterns are described consistently.
  4. [Abstract and Section 4.2] The abstract’s “average PSNR gain of 1.84 dB” should be qualified with the dataset and bitrate regime over which the average is taken; Section 4.2 does not clearly identify that this figure comes from averaging the low, medium, and high bitrate rows on Kodak.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reasoning: the claims are empirical evaluations on held-out benchmarks against external baselines; the hyperprior-losslessness assumption is a disclosed scope limitation, not a derivation from the claimed result.

full rationale

The paper's derivation chain is architecture design followed by empirical evaluation, not a fitted-parameter-then-prediction loop. The ICR/ICG modules and two-layer dual-branch autoregression are motivated by the energy-distribution analysis of Section 3.2 and validated through ablations in Section 4.6 and Appendix E.4 on held-out Kodak and CLIC images. The headline 1.84 dB gain over LossResilientLIC at 20% packet loss is a measured quantity from Tables 1 and 2 under a uniform-loss protocol averaged over ten trials, and the Gilbert-Elliott generalization claim of Section 4.4 is an out-of-distribution test, not a quantity defined by the training objective. No parameter is fitted to the test-set numbers and then reported as a prediction; the 'Baseline' and 'Random' variants are ablation checkpoints rather than fitted proxies for the headline metric. The hyperprior-losslessness assumption stated in Sections 3.1, 4.1, and the Conclusion is a real scope restriction, since all packet-loss numbers condition on the hyperprior packet being received intact; however, this is an explicitly disclosed limitation and a deployment caveat, not circular reasoning, and it does not make the measured PSNR gains equal to the model's inputs by construction. Self-citations in the reference list (e.g., the same group's earlier compression papers and the Flickr2W dataset paper) are not load-bearing for the loss-resilience claim, which is established against external published baselines such as LossResilientLIC, ResiComp, ProgDTD, and JPEG2000. Therefore the honest finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The ledger shows no free parameters fitted to the target loss-resilience results. The listed lambda values and p_max are standard training hyperparameters. The central assumptions are the lossless hyperprior, packet-level latent-space loss simulation, mask-conditioned restoration, and cross-dataset generalization. No new unobserved entities are introduced; ICR, ICG and the autoregressive structure are architectural components, not invented physical quantities.

free parameters (2)
  • Rate-distortion trade-off lambda = 0.0018, 0.0035, 0.0067
    Standard Lagrangian controls selecting low, medium, and high bitrate; not fitted to loss-resilience claims and not load-bearing for the central result.
  • Maximum training packet loss probability p_max = 0.3
    Chosen by hand for the three-stage training schedule; affects robustness at high loss rates but is not fitted to the reported test results.
assumptions (6)
  • domain assumption Hyperprior bitstream ẑ is transmitted losslessly (or protected by FEC) and is available at the decoder.
    Section 3.1 states "we assume ẑ of the hyperprior branch to be lossless during transmission"; all experiments exclude hyperprior loss. If this fails, decoding is infeasible, so it is load-bearing. The paper mitigates with RS(12,8) coding in Appendix C.
  • domain assumption Packet loss can be simulated as zeroing all channels of lost packets in the latent space, with mask propagation to dependent second-layer packets.
    Section 3.6 adopts packet-level loss simulation rather than channel-level loss. This assumes feature-domain zeroing matches binary-file packet dropout in testing.
  • domain assumption Mask Conditional Aggregation can restore missing channels from received channels.
    Section 3.6: "the mask is fed into a mask-conditioned aggregation module [37], which identifies positions of lost channels and performs feature restoration." Restoration quality bounds the achievable PSNR.
  • domain assumption The HPCM-based non-autoregressive baseline is a fair codec backbone for isolating loss-resilience gains.
    Appendix A: Baseline is a non-autoregressive variant with independent estimation of each latent element; the comparisons rely on this baseline being architecturally representative.
  • domain assumption Standard arithmetic coding is assumed error-free for received packets; only packet erasures are modeled.
    Section 3.6 and Appendix D: testing discards binary files to emulate packet loss; bit-level errors within received packets are not modeled.
  • domain assumption Training on Flickr2W 256x256 crops transfers to Kodak and CLIC evaluation.
    Section 4.1: training uses Flickr2W and evaluation uses Kodak/CLIC; cross-dataset generalization is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression." pith.science (2026). https://pith.science/paper/74MFCAA6

@misc{pith2026260811096,
  author       = {Pith},
  title        = {Pith review of: Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/74MFCAA6}},
  note         = {Machine review of arXiv:2608.11096}
}
read the original abstract

Learned image compression (LIC) has achieved impressive rate-distortion performance. However, existing methods remain highly vulnerable to packet loss, a common challenge in satellite and emergency communications. This vulnerability stems from non-uniform information distribution at the packetization stage and sequential decoding dependencies at the entropy coding stage. We propose an end-to-end loss-resilient image compression scheme that addresses both. Before packetization, we introduce an Inter-Channel Redistribution (ICR) mechanism to redistribute channel energy, preventing critical information concentrating in a small subset of channels. Then, an Interleaved Channel Grouping (ICG) strategy partitions latent channels in a strided manner to disperse information across packets, with each packet kept within constrained sizes. To limit cascading errors from lost packets, we adopt a two-layer dual-branch autoregressive structure to shorten the dependency chain. Extensive experiments demonstrate that our method consistently outperforms existing approaches in both reconstruction quality and stability. At 20% packet loss, it achieves an average PSNR gain of 1.84 dB over LossResilientLIC while reducing PSNR variance by an order of magnitude. Notably, trained under uniform random loss only, our model generalizes to bursty loss modeled by the Gilbert-Elliott channel, outperforming methods explicitly trained for such conditions.

Figures

Figures reproduced from arXiv: 2608.11096 by the authors.

Figure 1
Figure 1. PSNR gains over JPEG2000 at medium bitrate under [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of our proposed method. The Inter-Channel Redistribution (ICR) module shuffles and disperses channel [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Analysis of information loss across spatial and channel dimensions. “Baseline” denotes the case without element [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Visualization of reconstructed images on Kodak dataset. All methods lose the first two packets. Our proposed method [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Rate-distortion performance under GE model sim [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Ablation on module design, partition strategy, and autoregression mode at packet loss rates of 5%, 10%, and 20% (left [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: The left two blocks depict details of Channel Unshuffle/Shuffle and Channel Attention modules in ICR and Inv [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: GE-based simulation of packet loss transmission. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Rate-Distortion curves under no packet loss, and [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: The rate-distortion curves of our model on the [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Visualization of reconstructed images on Kodak dataset. All methods lose the first two packets, while in our method, [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Visualization of reconstructed images on CLIC dataset. All methods lose the first two packets, while in our method, [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 48 canonical work pages

  1. [1]

    Muhammad Salman Ali, Yeongwoong Kim, Maryam Qamar, Sung-Chang Lim, Donghyun Kim, Chaoning Zhang, Sung-Ho Bae, and Hui Yong Kim. 2023. To- wards efficient image compression without autoregressive models.Adv. Neural Inf. Process. Syst.36 (2023), 7392–7404

    work page 2023

  2. [2]

    Johannes Ballé, Valero Laparra, and Eero P Simoncelli. 2017. End-to-end Opti- mized Image Compression. InInt. Conf. Learn. Represent

    work page 2017

  3. [3]

    Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick John- ston. 2018. Variational image compression with a scale hyperprior. InInt. Conf. Learn. Represent

    work page 2018

  4. [4]

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

    work page 2018

  5. [5]

    Sullivan, and Jens-Rainer Ohm

    Benjamin Bross, Ye-Kui Wang, Yan Ye, Shan Liu, Jianle Chen, Gary J. Sullivan, and Jens-Rainer Ohm. 2021. Overview of the Versatile Video Coding (VVC) Standard and its Applications.IEEE Trans. Circuits Syst. Video Technol.31, 10 (2021), 3736–3764

    work page 2021

  6. [6]

    Yunuo Chen, Bing He, Zezheng Lyu, Hongwei Hu, Qunshan Gu, Yuan Tian, and Guo Lu. 2026. Adaptive Learned Image Compression with Graph Neural Networks. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.12150–12161

    work page 2026

  7. [7]

    Yunuo Chen, Zezheng Lyu, Bing He, Ning Cao, Gang Chen, Guo Lu, and Wenjun Zhang. 2025. Knowledge Distillation for Learned Image Compression. InProc. IEEE/CVF Int. Conf. Comput. Vis.4996–5006

    work page 2025

  8. [8]

    Yunuo Chen, Zezheng Lyu, Bing He, Hongwei Hu, Qi Wang, Yuan Tian, Li Song, Wenjun Zhang, and Guo Lu. 2026. Content-aware mamba for learned image compression. InInt. Conf. Learn. Represent., Vol. 2026. 78452–78477

    work page 2026

Show all 51 references
  1. [9]

    Yihua Cheng, Ziyi Zhang, Hanchen Li, Anton Arapin, Yue Zhang, Qizheng Zhang, Yuhan Liu, Kuntai Du, Xu Zhang, Francis Y Yan, et al. 2024. GRACE: Loss-Resilient Real-Time Video Through Neural Codecs. InProc. USENIX Symp. Netw. Syst. Des. Implement.509–531

    2024

  2. [10]

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto. 2020. Learned Image Compression With Discretized Gaussian Mixture Likelihoods and Atten- tion Modules. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit

    2020

  3. [11]

    Wen-Jeng Chu and Jin-Jang Leou. 1998. Detection and concealment of transmis- sion errors in H. 261 images.IEEE Trans. Circuits Syst. Video Technol.8, 1 (1998), 74–84

    1998

  4. [12]

    Ze Cui, Jing Wang, Shangyin Gao, Tiansheng Guo, Yihui Feng, and Bo Bai. 2021. Asymmetric Gained Deep Image Compression With Continuous Rate Adaptation. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.10532–10541

    2021

  5. [13]

    Zhihao Duan, Ming Lu, Jack Ma, Yuning Huang, Zhan Ma, and Fengqing Zhu

  6. [14]

    Zhihao Duan, Ming Lu, Zhan Ma, and Fengqing Zhu. 2023. Lossy image com- pression with quantized hierarchical vaes. InProc. IEEE/CVF Winter Conf. Appl. Comput. Vis.198–207

    2023

  7. [15]

    Donghui Feng, Zhengxue Cheng, Shen Wang, Ronghua Wu, Hongwei Hu, Guo Lu, and Li Song. 2025. Linear attention modeling for learned image compression. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.7623–7632

    2025

  8. [16]

    Runsen Feng, Zongyu Guo, Weiping Li, and Zhibo Chen. 2023. Nvtc: Nonlinear vector transform coding. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. 6101–6110

    2023

  9. [17]

    Rich Franzen. 1993. Kodak Lossless True Color Image Suite (PhotoCD PCD0992). http://r0k.us/graphics/kodak/

    1993

  10. [18]

    Tobias Gruber, Sebastian Cammerer, Jakob Hoydis, and Stephan Ten Brink. 2017. On deep learning-based channel decoding. InProc. Annu. Conf. Inf. Sci. Syst.IEEE, 1–6

    2017

  11. [19]

    Gerhard Haßlinger and Oliver Hohlfeld. 2008. The Gilbert-Elliott model for packet loss in real time services on the Internet. InProc. GI/ITG Conf. Meas. Model. Eval. Comput. Commun. Syst.VDE, 1–15

    2008

  12. [20]

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

  13. [21]

    Dailan He, Yaoyan Zheng, Baocheng Sun, Yan Wang, and Hongwei Qin. 2021. Checkerboard Context Model for Efficient Learned Image Compression. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.14771–14780

    2021

  14. [22]

    Ali Hojjat, Janek Haberer, and Olaf Landsiedel. 2023. ProgDTD: Progressive learned image compression with double-tail-drop training. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.1130–1139

    2023

  15. [23]

    Ismaeil Ismaeil, Shahram Shirani, Faouzi Kossentini, and Rabab Ward. 2000. An efficient, similarity-based error concealment method for block-based coded images. InProc. IEEE Int. Conf. Image Process., Vol. 3. IEEE, 388–391

    2000

  16. [24]

    Wei Jiang and Ronggang Wang. 2023. MLIC++: Linear Complexity Multi- Reference Entropy Modeling for Learned Image Compression. InProc. Int. Conf. Mach. Learn. Workshops

    2023

  17. [25]

    Wei Jiang, Jiayu Yang, Yongqi Zhai, Peirong Ning, Feng Gao, and Ronggang Wang

  18. [26]

    Jae-Han Lee, Seungmin Jeon, Kwang Pyo Choi, Youngo Park, and Chang-Su Kim

  19. [27]

    Han Li, Shaohui Li, Wenrui Dai, Maida Cao, Nuowen Kan, Chenglin Li, Junni Zou, and Hongkai Xiong. 2025. On Disentangled Training for Nonlinear Transform in Learned Image Compression. InInt. Conf. Learn. Represent

    2025

  20. [28]

    Mlic: Multi-reference entropy model for learned image compression. In Proc. ACM Int. Conf. Multimedia. 7618–7627

  21. [29]

    Yuqi Li, Haotian Zhang, Li Li, and Dong Liu. 2025. Learned image compression with hierarchical progressive context modeling. InProc. IEEE/CVF Int. Conf. Comput. Vis.18834–18843

    2025

  22. [30]

    DPICT: Deep progressive image compression using trit-planes. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.16113–16122

  23. [31]

    David Minnen, Johannes Ballé, and George Toderici. 2018. Joint Autoregressive and Hierarchical Priors for Learned Image Compression. InAdv. Neural Inf. Process. Syst.10794–10803

    2018

  24. [32]

    Yanghao Li, Tongda Xu, Yan Wang, Jingjing Liu, and Ya-Qin Zhang. 2023. Idem- potent learned image compression with right-inverse.Adv. Neural Inf. Process. Syst.36 (2023), 12878–12896

    2023

  25. [33]

    Guanbo Pan, Guo Lu, Zhihao Hu, and Dong Xu. 2022. Content adaptive latents and decoder for neural image compression. InProc. Eur. Conf. Comput. Vis.556– 573

    2022

  26. [34]

    Jiaheng Liu, Guo Lu, Zhihao Hu, and Dong Xu. 2020. A Unified End-to-End Frame- work for Efficient Deep Image Compression.arXiv preprint arXiv:2002.03370 (2020)

    2020 arXiv

  27. [35]

    Irving S Reed and Gustave Solomon. 1960. Polynomial codes over certain finite fields.Journal of the society for industrial and applied mathematics8, 2 (1960), 300–304

    1960

  28. [36]

    David Minnen and Saurabh Singh. 2020. Channel-wise autoregressive entropy models for learned image compression. InProc. IEEE Int. Conf. Image Process. IEEE, 3339–3343

    2020

  29. [37]

    Hongwei Sha, Muchen Dong, Quanyou Luo, Ming Lu, Hao Chen, and Zhan Ma

  30. [38]

    Shiyu Qin, Jinpeng Wang, Yimin Zhou, Bin Chen, Tianci Luo, Baoyi An, Tao Dai, Shutao Xia, and Yaowei Wang. 2024. MambaVC: Learned visual compression with selective state spaces.arXiv preprint arXiv:2405.15413(2024)

    2024 arXiv

  31. [39]

    Lucas Theis, Wenzhe Shi, Andrew Cunningham, and Ferenc Huszár. 2017. Lossy image compression with compressive autoencoders. InInt. Conf. Learn. Represent

    2017

  32. [40]

    Michael Rudow, Francis Y Yan, Abhishek Kumar, Ganesh Ananthanarayanan, Martin Ellis, and KV Rashmi. 2023. Tambur: Efficient loss recovery for videocon- ferencing via streaming codes. InProc. USENIX Symp. Netw. Syst. Des. Implement. 953–971

    2023

  33. [41]

    Gregory K. Wallace. 1991. The JPEG Still Picture Compression Standard.Com- munication ACM34, 4 (1991), 30–44

    1991

  34. [42]

    Sixian Wang, Jincheng Dai, Xiaoqi Qin, Ke Yang, Kai Niu, and Ping Zhang. 2025. ResiComp: Loss-Resilient Image Compression via Dual-Functional Masked Visual Token Modeling.IEEE Trans. Circuits Syst. Video Technol.(2025)

    2025

  35. [43]

    Athanassios Skodras, Charilaos Christopoulos, and Touradj Ebrahimi. 2001. The JPEG 2000 still image compression standard.IEEE Signal Process. Mag.18, 5 (2001), 36–58

    2001

  36. [44]

    Fanhu Zeng, Hao Tang, Yihua Shao, Siyu Chen, Ling Shao, and Yan Wang. 2025. MambaIC: State space models for high-performance learned image compression. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.18041–18050

    2025

  37. [45]

    George Toderici, Lucas Theis, Nick Johnston, Eirikur Agustsson, Fabian Mentzer, Johannes Ballé, Wenzhe Shi, and Radu Timofte. 2020. Clic 2020: Challenge on learned image compression. 29 (2020), 2021

    2020

  38. [46]

    MingSheng Zhou and MingMing Kong. 2025. GLIC: General Format Learned Image Compression. InProc. AAAI Conf. Artif. Intell., Vol. 39. 10815–10824. MM ’26, November 10–14, 2026, Rio de Janeiro, Brazil Yuhang Wei, Chuqin Zhou, Yibo Shi, Jing Wang, & Guo Lu A Details of Architectur...

    2025

  39. [48]

    Siqi Wu, Yinda Chen, Dong Liu, and Zhihai He. 2025. Conditional Latent Coding with Learnable Synthesized Reference for Deep Image Compression. InProc. AAAI Conf. Artif. Intell., Vol. 39. 12863–12871

    2025

  40. [50]

    Chuqin Zhou, Guo Lu, Jiangchuan Li, Xiangyu Chen, Zhengxue Cheng, Li Song, and Wenjun Zhang. 2025. Controllable distortion-perception tradeoff through latent diffusion for neural image compression. InProc. AAAI Conf. Artif. Intell., Vol. 39. 10725–10733

    2025

  41. [2022]

    ELIC: Efficient Learned Image Compression with Unevenly Grouped Space-Channel Contextual Adaptive Coding. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.5708–5717

  42. [2023]

    Pattern Anal

    QARV: Quantization-aware resnet vae for lossy image compression.IEEE Trans. Pattern Anal. Mach. Intell.46, 1 (2023), 436–450

    2023

  43. [2025]

    Towards loss-resilient image coding for unstable satellite networks. InProc. AAAI Conf. Artif. Intell., Vol. 39. 12506–12514

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.