Pith. sign in

REVIEW 3 major objections 5 minor 32 references

Distributed Image Semantic Communication via Nonlinear Transform Coding

T0 review · 3 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 The paper claims that explicitly modeling the correlation between two camera views—probabilistically through a joint entropy model and geometrically through a spatial transformer—lets distributed image transmission beat purely data-driven…

desk verdict Strong engineering contribution to distributed image coding undermined by an unachievable rate accounting for the hyperpriors. read the letter →

arxiv 2506.07391 v1 pith:AGDQWIHX submitted 2025-06-09 cs.IT eess.SPmath.IT

Yufei Bo , Meixia Tao , Kai Niu This is my paper
classification cs.ITeess.SPmath.IT
keywords distributedsourcecodingjointsource-channelnonlineartransformsemanticcommunicationentropymodelspatialtransformermulti-viewimagecompressionvariationalinference
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 claims that distributed image transmission can be improved by modeling the correlation between two cameras' views explicitly, instead of letting a neural network learn it implicitly. To that end it builds two coding schemes on nonlinear transform coding (NTC): D-NTSC for separate source and channel coding and D-NTSCC for joint source-channel coding. Both use a learned joint entropy model of the two latent representations to allocate rate adaptively and a spatial transformer to align the two views before joint decoding. On KITTI Stereo and Cityscapes, the schemes are reported to achieve state-of-the-art rate-distortion results, with D-NTSC reducing bit rate by roughly 30% versus a leading distributed coding baseline and D-NTSCC saving up to 50% bandwidth over a distributed deep JSCC baseline. The broader point is that coupling variational autoencoders with explicit geometric alignment can turn distributed source coding into a practical, learnable system.

What carries the argument

The argument turns on three components. (1) A joint entropy model: the hyperpriors $z_1,z_2$ are modeled jointly by a pairwise factorized bivariate Gaussian mixture, giving a density $p(\tilde z_1,\tilde z_2)$ that approximates the true joint distribution of the latents and guides adaptive rate allocation for both users. (2) A transformation module based on spatial transformer networks, which estimates a projective (homography) matrix $M_i \in \mathbb{R}^{3\times 3}$ and warps the other user's latent representation into the expected viewpoint before the joint synthesis transform, maximizing usable side information. (3) A variational-inference loss derivation in which the Markov chain $\tilde y_1 - \tilde z_1 - \tilde z_2 - \tilde y_2$ lets the KL objective decompose into distortion plus rate terms, with the joint hyperprior density replacing an independence assumption. The Swin Transformer backbone supplies the nonlinear analysis and synthesis transforms in both schemes.

What would settle it

Train the D-NTSC encoder on a stereo dataset, then compute the empirical conditional mutual information $I(\tilde y_1;\tilde y_2 \mid \tilde z_1, \tilde z_2)$ on a held-out set. If the estimate is clearly nonzero, the joint entropy model fails to capture all source correlation and the derived loss is not the true variational objective.

Watch

Extended reading notes

Core claim

For two correlated image sources separately encoded at distinct transmitters and jointly decoded at a common receiver, the paper establishes that explicitly modeling their joint distribution in latent space yields measurable gains over purely data-driven correlation learning. The central result is the joint entropy model $p(\tilde z_1,\tilde z_2)$ built as a pairwise factorized bivariate Gaussian mixture, which lets each transmitter estimate the other's rate and drives adaptive rate allocation, together with a projective transformation module that warps one latent representation toward the other's viewpoint before decoding. Variational inference produces the loss functions (12) and (13), whose rate terms factor through the Markov chain $\tilde y_1 - \tilde z_1 - \tilde z_2 - \tilde y_2$, with the joint hyperprior density replacing the independence assumption used by prior distributed JSCC work. Experiments report D-NTSC outperforming LDMIC by 0.3--0.6 dB PSNR at comparable bit rates and reducing bit rate by 30.7% on KITTI Stereo and 26.0% on Cityscapes, while D-NTSCC saves up to 50% bandwidth over D-DJSCC and about 30% over DHF-JSCC at equal quality. An ablation attributes part of the gain to the joint entropy model itself, which reduces bit rate by up to 8.43% relative to an independent entropy model.

Load-bearing premise

The two images' latent representations are assumed to be independent once their hyperpriors are known, so any correlation left between the latents after conditioning on the hyperpriors is invisible to the rate model and would make the reported bit rates underestimates.

Editorial extensions

If this is right

  • D-NTSC and D-NTSCC give system designers two operating points: choose D-NTSCC for low rates and low signal-to-noise ratios, and D-NTSC for higher rates where separate source and channel coding performs better.
  • The joint entropy modeling can be extended to more than two users by generalizing the bivariate Gaussian mixture to multivariate Gaussian variables, as the paper notes.
  • Because each transmitter can estimate the other's coding rate from the shared joint probability model, adaptive rate allocation and resource coordination can happen without a central encoder.
  • The probabilistic correlation model and the geometric alignment module are separable components that could be combined with other multi-view coding or transmission pipelines.
  • The reported gains suggest that explicit correlation structure—both probabilistic and geometric—is a stronger inductive bias than implicit attention-based correlation learning for distributed image coding.

Reading between the lines

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

  • The Markov-chain assumption could be tested directly: estimate the empirical conditional mutual information $I(\tilde y_1;\tilde y_2 \mid \tilde z_1, \tilde z_2)$ on the learned latents; if it is clearly nonzero, the joint entropy model undercounts the true rate and the derived loss is only an approximation to the variational objective.
  • The framework likely transfers to other multi-view tasks such as multi-camera surveillance or light-field compression, where pose alignment and joint entropy are natural priors, but the paper evaluates only standard stereo image pairs, so the transfer remains untested.
  • The bandwidth-savings numbers compare against baselines with different architectures; a controlled comparison using the same Swin Transformer backbone with and without the joint entropy model would isolate the contribution of explicit probabilistic correlation modeling.
  • The rate-estimation procedure assumes the channel conditions used in training; performance under SNR mismatch or fading channels is not examined, and that is where the JSCC scheme's adaptive rate mechanism would face a realistic stress test.
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 D-NTSC and D-NTSCC, two learning-based distributed image transmission schemes for correlated sources, corresponding to separate source-channel coding (SSCC) and joint source-channel coding (JSCC). Both schemes use nonlinear transform coding with Swin Transformer backbones, a joint entropy model for the hyperpriors z1 and z2, and a spatial transformation module that warps one latent representation to align with the other before joint synthesis decoding. Loss functions are derived from variational inference, with rate terms including the claimed joint entropy H(z̄1,z̄2) and per-user distortion terms. Experiments on KITTI Stereo and Cityscapes report state-of-the-art rate-distortion and channel-rate performance relative to learning-based distributed coding baselines such as LDMIC, D-DJSCC, and DHF-JSCC, plus ablations for the joint entropy model and the transformation module.

Significance. If the reported rate-distortion gains are real, the paper makes a useful contribution by moving distributed image semantic communication from purely data-driven correlation learning toward an explicit probabilistic model of source correlation, and by introducing a geometric alignment module that is broadly applicable. The experimental section is broad, includes two datasets and multiple metrics, and the ablation studies isolate the joint entropy model and the transformation module. The promise to release code is a welcome step for reproducibility. However, the core theoretical and practical claims rest on two load-bearing assumptions that are not adequately supported: the achievability of the joint hyperprior entropy by two separate arithmetic encoders, and a Markov-chain conditional-independence structure for the latent representations. These issues directly affect the numerical bit-rate and bandwidth-savings claims, so the significance of the empirical gains cannot be assessed until they are resolved.

major comments (3)
  1. [Section IV-C, Eqs. (7)-(9), and Figs. 2-3] The reported hyperprior rate of (1/2)H(z̄1,z̄2) per user is not achievable by the architecture described. Each transmitter runs its own arithmetic encoder on its own z̄_i, and the paper does not specify or implement any distributed source coding mechanism, such as Slepian-Wolf binning, syndrome-based coding, or distributed arithmetic coding. The joint model in Eq. (16) is used only for rate estimation and for providing the MMSE estimate z2* of the other user's hyperprior for rate allocation (Section IV-C, Remarks), which does not shorten the entropy-coded bitstream. The actual rate produced by two independent arithmetic encoders is approximately H(z̄1)+H(z̄2), not H(z̄1,z̄2). Consequently, Eqs. (8) and (9) understate the true bit rate by H(z̄_i)-1/2H(z̄1,z̄2) per user. This affects all reported rate-distortion curves (Figs. 9-13) and specifically the claimed 8.43% bit-rate saving of the joint entropy model in Fig. 10 and the 30.7% bit-rate reduction over LDMIC in Section VI-B. Please either implement a genuine distributed entropy-coding scheme for the hyperpriors or report all rates using the marginal entropy models that match the actual arithmetic coders.
  2. [Appendix A, Eqs. (20)-(22), and Proposition 1] The factorization p(y1,y2|z1,z2)=p(y1|z1)p(y2|z2) is asserted to follow from the Markov chain y1 - z1 - z2 - y2, but this chain is neither proven nor empirically validated. The analysis transforms ga1 and ga2 are learned independently and the hyperpriors zi are low-dimensional summaries, so there is no evident reason that all correlation between y1 and y2 is mediated by z1 and z2. If residual correlation remains, Eq. (10) is not the correct variational objective and the model may underestimate the joint rate. I recommend an empirical test on held-out data: compare the model's rate -log p(y1|z1)-log p(y2|z2)-log p(z1,z2) with the actual bitstream length produced by the separate arithmetic encoders, or fit a richer conditional model and report the rate difference. The analogous Markov chain ŝ1 - z1 - z2 - ŝ2 in Appendix B raises the same concern for D-NTSCC.
  3. [Section VI-C and Eq. (18)] The bandwidth allocation in Eq. (18) is a heuristic nearest-neighbor discretization of -η log p(y^j_1|z̄1) onto the finite set V, and the paper does not analyze or ablate this approximation. Since the rate estimates used in this allocation are themselves affected by the unachievable hyperprior cost in Eq. (9), it is unclear how much of the reported bandwidth savings in Figs. 11-12 comes from the coding scheme and how much from the optimistic rate accounting. Please provide results with the corrected rate definition and, if possible, an ablation of the discretization granularity.
minor comments (5)
  1. [Section II heading] The section heading contains a typo: 'BACKGROND' should be 'BACKGROUND'.
  2. [Eq. (16)] The notation N(m_k, Σ_k) * U([-1/2,1/2]^2) should specify that this is a two-dimensional convolution of the Gaussian density with the bivariate uniform density; the current notation is easy to misread as a product of densities.
  3. [Fig. 2 and Fig. 3] The AE/AD blocks are drawn separately for each user's hyperprior, which visually reinforces the concern that the two hyperpriors are entropy-coded independently; if joint decoding of z1 and z2 is intended, the figures should show the side-information path.
  4. [Section VI-F, Table I] The complexity comparison would benefit from stating the input image resolution and channel-use count used for the FLOPs measurements, since these quantities strongly affect the comparison.
  5. [Section IV-C, Remarks] The phrase 'both transmitters share this joint probability function, they can predict each other's transmission rate' is about rate estimation, not about joint entropy coding; the distinction should be stated explicitly to avoid implying that the joint model is used by the arithmetic encoders.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the variational losses and rate estimates are standard NTC self-consistent quantities, and the central comparisons are against external baselines.

full rationale

The derivation chain is self-contained and does not reduce to its own inputs. The loss functions (12) and (13) are obtained by minimizing the expected KL divergence between a parameterized variational density and the assumed generative posterior, which is the standard variational-inference route used in nonlinear transform coding. The rate terms in those losses are the negative log-likelihoods of the learned entropy models; reporting those same quantities as bit rates is the usual NTC operating definition, not a circular prediction, because the distortion values in the RD curves are independently measured PSNR/MS-SSIM and the baselines (LDMIC, D-DJSCC, DHF-JSCC, BPG, NTSCC) are external systems. The self-citations, notably the preliminary version [1] and the NTSCC architecture [7] on which the JSCC codec builds, are not load-bearing evidence for the paper's central claims; [7] is used as a baseline and starting point, not as an unverified uniqueness or correctness argument. Two genuine concerns exist but are not circularity: the asserted Markov chain y1 - z1 - z2 - y2 in Propositions 1 and 2 and Appendix A is an unproven modeling assumption, and the use of H(z1,z2) as the hyperprior rate in Eqs. (7)-(9) may not be achievable by two physically separate arithmetic encoders without an explicit Slepian-Wolf or distributed arithmetic-coding mechanism. These are correctness and achievability risks, not cases where an equation is defined in terms of its claimed output or a fitted parameter is renamed as a prediction. No circular step was found.

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

The central measurement claims rest on a few modeling choices: the conditional-independence structure of the latent and hyperprior variables, the homography alignment model, and the heuristic mapping from estimated bits to channel uses. The free parameters are operating points and design choices whose values are not fully reported, but the method's qualitative gains are not solely due to them.

free parameters (5)
  • β1, β2 (D-NTSC rate-distortion weights) = not reported; swept
    Eq. (12) uses β1 and β2 to trade distortion against rate. The paper says the compression ratio is controlled by changing these values but does not list the numbers used for the curves.
  • α1, α2 (D-NTSCC rate-distortion weights) = not reported; swept
    Eq. (13) uses α1 and α2 similarly; values are not specified.
  • η (bit-to-channel-use proportionality) = 1
    Fixed at 1 in Eq. (13); converts the estimated bit-rate of y_i into transmission cost. There is no derivation, and no sensitivity study is reported.
  • K (GMM components in joint hyperprior model) = 1
    The joint density in Eq. (15) is a GMM; K is set to 1 (a single bivariate Gaussian per element). No ablation over K is provided.
  • V (discrete bandwidth-cost set) = {8a | a=1,...,20}
    Used in Eq. (18) to quantize the channel-use allocation; the spacing and range are hand-chosen and not justified.
assumptions (5)
  • standard math Variational autoencoder interpretation of NTC (Ballé et al.)
    The loss derivations in Section IV rest on the VAE view of nonlinear transform coding from [29], which is accepted background in learned image compression.
  • domain assumption Conditional independence of latents given hyperpriors: p(y1,y2|z1,z2)=p(y1|z1)p(y2|z2)
    Introduced in Proposition 1 and Appendix A (Eq. 20-21) as the Markov chain y1 - z1 - z2 - y2. It is a modeling assumption that the correlated latents are independent once hyperpriors are known, and it is not validated against data.
  • domain assumption Error-free hyperprior transmission at cost H(z1,z2)/2 per user
    Used in Eqs. (8) and (9) to compute total rate and channel use; assumes capacity-achieving codes for the hyperprior bitstream, an idealization common in NTC literature.
  • domain assumption Global projective homography relates the two latent representations
    The transformation module (Section V-B, Eq. 17) assumes a single 3x3 homography warps one latent map onto the other, valid for planar scenes under a pinhole camera model; not validated for arbitrary multi-view geometry.
  • standard math Gaussian likelihood for distortion terms in the loss
    The distortion terms in (12) and (13) are the negative log of a Gaussian likelihood, equivalent to MSE; this is the standard assumption when optimizing PSNR.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributed Image Semantic Communication via Nonlinear Transform Coding." pith.science (2026). https://pith.science/paper/AGDQWIHX

@misc{pith2026250607391,
  author       = {Pith},
  title        = {Pith review of: Distributed Image Semantic Communication via Nonlinear Transform Coding},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AGDQWIHX}},
  note         = {Machine review of arXiv:2506.07391}
}
read the original abstract

This paper investigates distributed source-channel coding for correlated image semantic transmission over wireless channels. In this setup, correlated images at different transmitters are separately encoded and transmitted through dedicated channels for joint recovery at the receiver. We propose a general approach for distributed image semantic communication that applies to both separate source and channel coding (SSCC) and joint source-channel coding (JSCC). Unlike existing learning-based approaches that implicitly learn source correlation in a purely data-driven manner, our method leverages nonlinear transform coding (NTC) to explicitly model source correlation from both probabilistic and geometric perspectives. A joint entropy model approximates the joint distribution of latent representations to guide adaptive rate allocation, while a transformation module aligns latent features for maximal correlation learning at the decoder. We implement this framework as D-NTSC for SSCC and D-NTSCC for JSCC, both built on Swin Transformers for effective feature extraction and correlation exploitation. Variational inference is employed to derive principled loss functions that jointly optimize encoding, decoding, and joint entropy modeling. Extensive experiments on real-world multi-view datasets demonstrate that D-NTSC and D-NTSCC outperform existing distributed SSCC and distributed JSCC baselines, respectively, achieving state-of-the-art performance in both pixel-level and perceptual quality metrics.

Figures

Figures reproduced from arXiv: 2506.07391 by the authors.

Figure 1
Figure 1. Left: Operational diagram of NTC. The label U|Q denotes the use of uniform noise during training and actual quantization during testing. Right: Graphical representation of the analysis transform as an inference model, and the synthesis transform as a generative model in NTC. Hollow nodes indicate random variables, solid nodes represent parameters, and arrows denote conditional dependencies. of the latent representat… view at source ↗
Figure 2
Figure 2. Overall architecture of the proposed D-NTSC scheme. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Overall architecture of the proposed D-NTSCC scheme. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Graphical representation of distributed analysis transforms as an inference model, and the joint synthesis transform as a generative model for D-NTSC. mean and variance ϵ 2 i . And we assume equal noise variance for both users, i.e., ϵ 2 1 = ϵ 2 2 = ϵ 2 . Thus, the two…
Figure 5
Figure 5. Figure 5: Graphical representation for D-NTSCC. variational density q(ˆs1, ˆs2, z˜1, z˜2|x1, x2) and the intractable joint posterior p(ˆs1, ˆs2, z˜1, z˜2|x1, x2). Proposition 2: The expectation of the KL divergence between q(ˆs1, ˆs2, z˜1, z˜2|x1, x2) and p(ˆs1, ˆs2, z˜1, z˜2|x1…
Figure 7
Figure 7. Figure 7: Left: Transformation module. Right: Detailed architec [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 6
Figure 6. Figure 6: NN architectures of nonlinear transform modules. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: NN architectures of JSCC codecs. The perspective transformation can relate any two images of the same planar surface under the assumption of a pinhole camera model [32], which makes it particularly useful for aligning images captured from different viewpoints [PITH_FU…
Figure 9
Figure 9. Figure 9: PSNR performances at varying bit rates. pairs, and 790 testing stereo image pairs. A stereo image pair refers to a pair of images each taken at the same time from a different camera. The other is Cityscapes, which contains 2975 training stereo image pairs, 500 validati…
Figure 10
Figure 10. Figure 10: Ablation study on the joint entropy model. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Performances at varying transmission rates when [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]
Figure 12
Figure 12. Figure 12: Performances at varying SNRs when r = 0.03. to the two existing distributed coding schemes, D-DJSCC and DHF-JSCC, D-NTSCC achieves up to 50% and 30% bandwidth savings, respectively. This performance gain is attributed to the joint entropy modeling and the “upgrade” of…
Figure 14
Figure 14. Figure 14: Ablation study on the transformation module. [PITH_FULL_IMAGE:figures/full_fig_p011_14.png]
Figure 15
Figure 15. Figure 15: Visualization of the effect of the transformation module. [PITH_FULL_IMAGE:figures/full_fig_p012_15.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 25 canonical work pages

  1. [1]

    Distributed Nonlinear Transform Source-Channel Coding for Wireless Correlated Image Transmission

    Y . Bo and M. Tao, “Distributed nonlinear transform source-channel coding for wireless correlated image transmission,”arXiv preprint arXiv:2503.21249, 2025

    work page Pith review arXiv 2025

  2. [2]

    AI empowered wireless communications: From bits to semantics,

    Z. Qin, L. Liang, Z. Wang, S. Jin, X. Tao, W. Tong, and G. Y . Li, “AI empowered wireless communications: From bits to semantics,”Proc. IEEE, vol. 112, no. 7, pp. 621–652, 2024

    2024

  3. [3]

    Beyond transmitting bits: Context, semantics, and task-oriented communications,

    D. G ¨und¨uz, Z. Qin, I. E. Aguerri, H. S. Dhillon, Z. Yang, A. Yener, K. K. Wong, and C.-B. Chae, “Beyond transmitting bits: Context, semantics, and task-oriented communications,”IEEE J. Sel. Areas Commun., vol. 41, no. 1, pp. 5–41, 2022

    work page 2022

  4. [4]

    Deep separate source- channel coding for semantic-aware image transmission,

    J. Huang, D. Li, C. Huang, X. Qin, and W. Zhang, “Deep separate source- channel coding for semantic-aware image transmission,” inIEEE Int. Conf. Commun. (ICC), Italy, May 2023, pp. 5626–5631

    work page 2023

  5. [5]

    Deep joint source- channel coding for wireless image transmission,

    E. Bourtsoulatze, D. Burth Kurka, and D. G ¨und¨uz, “Deep joint source- channel coding for wireless image transmission,”IEEE Trans. Cogn. Commun. Netw., vol. 5, no. 3, pp. 567–579, 2019

    2019

  6. [6]

    Deep learning enabled semantic communication systems,

    H. Xie, Z. Qin, G. Y . Li, and B.-H. Juang, “Deep learning enabled semantic communication systems,”IEEE Trans. Signal Process., vol. 69, pp. 2663–2675, 2021

    2021

  7. [7]

    Nonlinear transform source-channel coding for semantic communications,

    J. Dai, S. Wang, K. Tan, Z. Si, X. Qin, K. Niu, and P. Zhang, “Nonlinear transform source-channel coding for semantic communications,”IEEE J. Sel. Areas Commun., vol. 40, no. 8, pp. 2300–2316, 2022

    2022

  8. [8]

    Joint coding-modulation for dig- ital semantic communications via variational autoencoder,

    Y . Bo, Y . Duan, S. Shao, and M. Tao, “Joint coding-modulation for dig- ital semantic communications via variational autoencoder,”IEEE Trans. Commun., vol. 72, no. 9, pp. 5626–5640, 2024

    work page 2024

Show all 32 references
  1. [9]

    Semantic communication systems for speech transmission,

    Z. Weng and Z. Qin, “Semantic communication systems for speech transmission,”IEEE J. Sel. Areas Commun., vol. 39, no. 8, pp. 2434– 2444, 2021

    2021

  2. [10]

    DeepWiVe: Deep-learning-aided wireless video transmission,

    T.-Y . Tung and D. G ¨und¨uz, “DeepWiVe: Deep-learning-aided wireless video transmission,”IEEE J. Sel. Areas Commun., vol. 40, no. 9, pp. 2570–2583, 2022

    2022

  3. [11]

    Learning task-oriented communication for edge inference: An information bottleneck approach,

    J. Shao, Y . Mao, and J. Zhang, “Learning task-oriented communication for edge inference: An information bottleneck approach,”IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 197–211, 2022

    2022

  4. [12]

    Toward semantic com- munications: Deep learning-based image semantic coding,

    D. Huang, F. Gao, X. Tao, Q. Du, and J. Lu, “Toward semantic com- munications: Deep learning-based image semantic coding,”IEEE J. Sel. Areas Commun., vol. 41, no. 1, pp. 55–71, 2023

    2023

  5. [13]

    Distributed deep joint source-channel coding over a multiple access channel,

    S. F. Yilmaz, C. Karamanlı, and D. G ¨und¨uz, “Distributed deep joint source-channel coding over a multiple access channel,” inIEEE Int. Conf. on Commun. (ICC), Italy, May 2023, pp. 1400–1405

    2023

  6. [14]

    Deep learning-based superposition coded modulation for hierarchical semantic communications over broadcast channels,

    Y . Bo, S. Shao, and M. Tao, “Deep learning-based superposition coded modulation for hierarchical semantic communications over broadcast channels,”IEEE Trans. Commun., vol. 73, no. 2, pp. 1186–1200, 2025

    2025

  7. [15]

    Task-oriented multi-user semantic com- munications for VQA,

    H. Xie, Z. Qin, and G. Y . Li, “Task-oriented multi-user semantic com- munications for VQA,”IEEE Wireless Commun. Lett., vol. 11, no. 3, pp. 553–557, 2022

    2022

  8. [16]

    Noiseless coding of correlated information sources,

    D. Slepian and J. Wolf, “Noiseless coding of correlated information sources,”IEEE Trans. Inf. Theory, vol. 19, no. 4, pp. 471–480, 1973

    1973

  9. [17]

    Multiterminal source coding,

    T. Berger, “Multiterminal source coding,”The inf. theory approach to commun., 1978

    1978

  10. [18]

    Tung,Multiterminal Source Coding.Cornell University, 1978

    S.-Y . Tung,Multiterminal Source Coding.Cornell University, 1978

    1978

  11. [19]

    Distributed source coding using syndromes (DISCUS): Design and construction,

    S. S. Pradhan and K. Ramchandran, “Distributed source coding using syndromes (DISCUS): Design and construction,”IEEE Trans. Inf. Theory, vol. 49, no. 3, pp. 626–643, 2003

    2003

  12. [20]

    On multiterminal source code design,

    Y . Yang, V . Stankovic, Z. Xiong, and W. Zhao, “On multiterminal source code design,”IEEE Trans. Inf. Theory, vol. 54, no. 5, pp. 2278–2302, 2008

    2008

  13. [21]

    DRASIC: Distributed recurrent au- toencoder for scalable image compression,

    E. Diao, J. Ding, and V . Tarokh, “DRASIC: Distributed recurrent au- toencoder for scalable image compression,” inData Compression Conf. (DCC), USA, March 2020, pp. 3–12

    2020

  14. [22]

    LDMIC: Learning-based distributed multi-view image coding,

    X. Zhang, J. Shao, and J. Zhang, “LDMIC: Learning-based distributed multi-view image coding,”arXiv preprint arXiv:2301.09799, 2023

    2023 arXiv

  15. [23]

    Distributed image transmission using deep joint source-channel coding,

    S. Wang, K. Yang, J. Dai, and K. Niu, “Distributed image transmission using deep joint source-channel coding,” inIEEE Int. Conf. Acoust. Speech Signal Process. (ICASSP), Singapore, May 2022, pp. 5208–5212

    2022

  16. [24]

    Content adaptive distributed joint source-channel coding for image transmission with hyperprior,

    Y . Li, X. Chen, X. Deng, and J. Gui, “Content adaptive distributed joint source-channel coding for image transmission with hyperprior,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 1, pp. 105–117, 2025

    2025

  17. [25]

    Robust deep joint source- channel coding enabled distributed image transmission with imperfect channel state information,

    B. Dong, B. Cao, G. Gui, and Q. Zhang, “Robust deep joint source- channel coding enabled distributed image transmission with imperfect channel state information,”arXiv preprint arXiv:2411.12228, 2024

    2024 arXiv

  18. [26]

    Nonlinear transform coding,

    J. Ball ´e, P. A. Chou, D. Minnen, S. Singh, N. Johnston, E. Agustsson, S. J. Hwang, and G. Toderici, “Nonlinear transform coding,”IEEE J. Sel. Top. Signal Process., vol. 15, no. 2, pp. 339–353, 2020

    2020

  19. [27]

    ITU-R Rec. T.81 & ISO/IEC 10918-1: Digital compression and coding of continuous-tone still images,

    “ITU-R Rec. T.81 & ISO/IEC 10918-1: Digital compression and coding of continuous-tone still images,” 1992

    1992

  20. [28]

    F. Bellard. (2014) Bpg image format. [Online]. Available: https://bellard.org/bpg/

    2014

  21. [29]

    Vari- ational image compression with a scale hyperprior,

    J. Ball ´e, D. Minnen, S. Singh, S. J. Hwang, and N. Johnston, “Vari- ational image compression with a scale hyperprior,”arXiv preprint arXiv:1802.01436, 2018

    2018 arXiv

  22. [30]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV), Montreal, Canada, Oct. 2021, pp. 10 012–10 022

    2021

  23. [31]

    Spatial transformer networks,

    M. Jaderberg, K. Simonyan, A. Zissermanet al., “Spatial transformer networks,” inAdv. Neural Inf. Process. Syst. (NeurIPS), Montreal, Canada, Dec. 2015

    2015

  24. [32]

    Perspective transformation layer,

    N. Khatri, A. Dasgupta, Y . Shen, X. Zhong, and F. Y . Shih, “Perspective transformation layer,” inInt. Conf. Comput. Sci. Comput. Intell. (CSCI), Las Vegas, USA, Dec. 2022, pp. 1395–1401

    2022

Pith tools

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