Pith. sign in

REVIEW 3 major objections 5 minor 62 references

Learning Optimal Lattice Vector Quantizers for End-to-end Neural Image Compression

T0 review · 3 major / 5 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 This paper claims that a learned, near-orthogonal lattice vector quantizer can replace the uniform scalar quantizer in end-to-end neural image compression, cutting bitrate by up to 22.6 percent at equal quality while staying nearly as…

desk verdict The reported bitrate savings are likely real, but the orthogonality constraint collapses the "lattice" into a learned orthogonal transform plus scalar quantization, so the paper oversells the LVQ framing. read the letter →

arxiv 2411.16119 v1 pith:T3XG2J3W submitted 2024-11-25 eess.IV cs.CV

Xi Zhang , Xiaolin Wu This is my paper · ORCID
classification eess.IVcs.CV
keywords latticevectorquantizationneuralimagecompressionlearnedBabairoundingend-to-endoptimizationrate-distortionentropycoding
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to overturn a default choice in neural image compression: instead of quantizing each latent coordinate independently with a uniform scalar quantizer, it learns a lattice vector quantizer whose geometry is fitted to the statistics of the latent features. The central claim is that a learned lattice quantizer, constrained to keep its basis nearly orthogonal, can replace the scalar quantizer inside existing end-to-end codecs and reduce bitrate at equal reconstruction quality. The reported savings range from 22.6 percent on a lightweight convolutional codec with no context model down to 0.95 percent on a strong transformer-CNN codec with a spatial autoregressive context model. If the claim holds, it matters because it gives a way to capture inter-feature dependencies and approach the performance of much more expensive vector quantizers while staying almost as cheap as scalar quantization.

What carries the argument

The central object is the learnable lattice generator matrix $B\in\mathbb{R}^{n\times n}$ whose columns are lattice basis vectors; the lattice is $\Lambda=\{Bm:m\in\mathbb{Z}^n\}$. Because the lattice is fully determined by $B$, learning a codebook means learning $B$. The argument rides on two pieces of machinery: Babai rounding, $q_l(v)=B\lfloor B^{-1}v\rceil$, turns nearest-lattice-point search into per-coordinate rounding, and the orthogonality penalty $L=\sum_{i\ne j}|b_i^\top b_j|$ keeps $B$ close enough to orthogonal that the rounding error stays small and the training-time quantizer matches inference. The rate model then factors the joint probability of a lattice point into a product of univariate Gaussian mixtures over the integer coefficients $m_i$, which is only justified by the same near-orthogonality.

What would settle it

Take a trained OLVQ model and, on a sample of latent vectors, compare two decoders: the normal Babai-rounding quantizer and an exact nearest-lattice-point search (or a lattice-reduction-based approximation) for the same learned basis. If the exact-search reconstructions do not improve rate-distortion, or if the average distance between Babai's output and the true nearest lattice point is large, the orthogonality-based training/inference consistency story is not doing the work attributed to it.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that the lattice basis matrix $B$ is a quantizer codebook that can be learned end-to-end rather than fixed in advance. The authors define the quantizer as $q_l(v) = B\lfloor B^{-1}v\rceil$, where $\lfloor\cdot\rceil$ rounds each coordinate, and they train $B$ jointly with the autoencoder under a rate-distortion loss plus an orthogonality penalty $L=\sum_{i\ne j}|b_i^\top b_j|$. The orthogonality penalty is what makes Babai rounding a faithful proxy for true nearest-lattice-point quantization, so the codebook learned during training is the one used at inference. A mixture-of-Gaussians entropy model, factored across the integer coefficient coordinates, supplies the rate estimate. The empirical claim is that this learned, nearly orthogonal lattice quantizer beats both a uniform scalar quantizer and classical fixed lattices such as the Gosset, Barnes-Wall, and Leech lattices on rate-distortion, and comes close to a general vector quantizer while running roughly an order of magnitude faster.

Load-bearing premise

The scheme presumes that the learned lattice basis stays close enough to orthogonal that Babai rounding picks nearly the nearest lattice point, so the quantizer used in training is the quantizer used at inference; the orthogonality penalty is soft, and the paper does not quantify the residual non-orthogonality.

Editorial extensions

If this is right

  • Swapping a uniform scalar quantizer for a learned lattice quantizer yields bitrate savings at equal distortion in every architecture and context model tested, with the largest savings, up to 22.6 percent, in lightweight codecs with no context model.
  • The learned lattice quantizer outperforms classical fixed lattices such as the Gosset, Barnes-Wall, and Leech lattices by a wide margin, and the gap grows with lattice dimension.
  • A learned lattice quantizer approaches the rate-distortion performance of a general vector quantizer while keeping inference time far lower, for instance roughly 40 ms versus 454 ms at dimension 32.
  • High-dimensional learned lattices are the most efficient: increasing the quantization dimension from 8 to 24 roughly doubles the reported savings.
  • The method can serve as a substitute for a strong autoregressive entropy model in low-bitrate, low-complexity settings, since OLVQ with a factorized entropy model approaches the autoregressive model's rate-distortion performance at low bitrates.

Reading between the lines

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

  • Because the gains shrink as the context model gets stronger, a plausible reading is that the learned lattice and the context model are doing overlapping work: both exploit inter-feature dependencies, and a direct test would be to train a codec with both OLVQ and a strong autoregressive context while ablating the context's receptive field.
  • The near-orthogonality penalty leaves residual non-orthogonality unmeasured; if exact nearest-lattice-point search replaces Babai rounding at inference, the reported savings might improve further, which would also reveal how much of the gain is attributable to the orthogonality constraint.
  • The method's logic is not specific to image latents; the same learned-lattice module could be applied to quantizing network weights or other neural representations, a generalization the authors note only briefly.
  • The trend that higher-dimensional lattices give larger gains suggests testing dimensions beyond 24, where exact lattice search becomes harder but Babai rounding plus orthogonality still keeps complexity linear in dimension.
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 OLVQ, a learnable lattice vector quantizer for end-to-end neural image compression. The generator matrix B of the lattice is trained jointly with the autoencoder and the entropy model under a soft orthogonality penalty. Quantization is performed with Babai rounding, and the entropy model is factorized across the lattice coefficients, exploiting the approximate orthogonality of the basis. Experiments on Kodak and CLIC report BD-rate savings over a uniform scalar quantizer across three networks (Bmshj2018, SwinT-ChARM, LIC-TCM) and four context models (Factorized, Checkerboard, Channel-wise Autoregressive, Spatial-wise Autoregressive), with gains ranging from -22.60% to -0.95%. The paper also compares with classical non-learned lattices and general vector quantizers, and reports inference times.

Significance. If the reported effects are real and attributable to lattice vector quantization, the method would provide a simple drop-in quantization module with substantial gains for lightweight compression models. The paper addresses an important practical problem, and the empirical trend (larger gains when the baseline entropy model is weaker) is internally consistent with the authors' stated limitation. However, the theoretical foundation is not yet established: the orthogonality constraint may collapse the quantizer to a scalar quantizer in a learned linear basis, and the Babai-rounding justification in the appendix is mathematically incomplete. The manuscript also releases no code or error bars, which limits verification. These gaps prevent the paper from fully supporting its central claim.

major comments (3)
  1. [Sections 3.3 and 3.5, Eqs. (7) and (13)] The orthogonality constraint b_i^T b_j = 0 and the penalty L are load-bearing for Babai rounding, but when B is exactly orthogonal, B = QD with Q orthogonal and D diagonal, making the quantizer in Eq. (6) equivalent to scalar quantization in a rotated and scaled coordinate system, and the entropy model in Eq. (11) exactly factorized. The paper does not report the actual degree of non-orthogonality of the learned B at convergence, nor does it compare against a baseline that applies a learned linear transform (e.g., a 1x1 convolution) before a scalar quantizer. Without these measurements, the bitrate savings in Table 1 cannot be attributed to lattice vector quantization rather than to the added linear transformation capacity.
  2. [Appendix A.4, Eq. (6)] The proof of the Babai rounding bound is not a bound for the algorithm actually used. The appendix analyzes rounding in the Gram-Schmidt basis and claims the bound transfers to the reconstructed lattice point v = sum round(c_i) b_i, but the step "Since v and t' differ only by the orthogonalization process, the distance bound remains valid" is unsupported; in general t - v is not equal to sum (c_i - round(c_i)) b_i^*. As a result, the paper does not establish that the quantization error of Eq. (6) is small enough for the training-time additive noise model to match the inference quantizer, which is the premise of the entire training scheme.
  3. [Section 4.2, Table 1] The empirical comparison is only against a uniform scalar quantizer without a learnable linear transform. Given that the method adds a learnable matrix B, the reported savings (e.g., -22.60% for Bmshj2018+Factorized) could be due to the extra transform parameters rather than the lattice structure. The authors should include an ablated baseline where B is constrained to be a fixed orthogonal matrix (e.g., identity) or where a learned orthogonal transform precedes a scalar quantizer, with the same training budget and entropy model. This control is necessary to support the central claim that OLVQ improves over scalar quantization via lattice vector quantization.
minor comments (5)
  1. [Table 2] The entry "10.92%" under "Optimized 8-dimension lattice" should be "-10.92%" to be consistent with the other negative values in the table.
  2. [Sections 4.3 and 4.4] There are typos such as "comarisons agasin" and "perfromance"; a thorough proofreading of the experiment sections is recommended.
  3. [Section 3.4, Eqs. (10) and (11)] The notation for the mixture weights is inconsistent: Eq. (10) defines Phi ~ Categorical(K, phi), while Eq. (11) uses phi_k^i without defining how phi is obtained or how the per-dimension weights relate to the joint mixture. Please clarify the model.
  4. [Appendix A.4] The described Babai algorithm steps do not match Eq. (6), which uses B * round(B^{-1} v). Please align the pseudocode with the actual rounding procedure and state explicitly how the bound applies to that procedure, or remove the appendix proof.
  5. [Eq. (8)] The initialization interval is written as U(-1/(n sqrt(S) - 1), 1/(n sqrt(S) - 1)), which is ambiguous and likely missing parentheses. Specify the intended range and provide a derivation or citation for this choice.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the empirical claims are supported by held-out evaluation, though the orthogonality constraint makes the LVQ-vs-transform-coding attribution conceptually blurry.

full rationale

The paper's central empirical result, the BD-rate savings in Tables 1-3, is obtained by end-to-end training on ImageNet crops and evaluation on held-out Kodak/CLIC datasets; there is no fitting to the test set, and the comparison against pre-defined lattices (E8, A16, A24) and general vector quantizers provides external benchmarks. The codebook B, entropy model parameters, and transforms are jointly optimized under Eq. (13), so the measured gains are independent empirical facts rather than constructions from the paper's assumptions. The phrase 'rate-distortion optimal' means optimal with respect to the authors' own approximate objective, which is tautological but standard in learned compression and not a hidden circular derivation. The most serious concern is that the orthogonality constraint in Eq. (7), introduced to justify Babai rounding in Eq. (6) and the product-form entropy model in Eq. (11), makes the learned quantizer approximately equivalent to a learned orthogonal linear transform followed by scalar quantization; the reported savings over a scalar-quantizer baseline without such a transform may therefore be transform-coding gains rather than evidence of non-rectangular LVQ cells. This is an attribution/framing weakness, not a circular derivation, because the empirical BD-rate numbers stand independently. Separately, the appendix proof (A.4.3) is mathematically incomplete: the bound on the Gram-Schmidt projection t' does not by itself transfer to the reconstructed lattice point v, since v - t' = sum round(c_i)(b_i - b_i^*) is uncontrolled. That is a correctness risk outside the circularity score. No load-bearing self-citation or imported uniqueness theorem was found.

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

The paper's central method introduces one learned matrix B and relies on several distributional approximations. There are no physically invented entities, but the orthogonality coupling between the codebook and the rate model is the main structural assumption. All of the method's parameters, including B and the entropy model parameters, are fitted to the training data, which is standard supervised learning.

free parameters (4)
  • Lattice generator matrix B = Trained end-to-end; dimension 32x32 for main results, 8x8, 16x16, 24x24 in comparisons
    This is the central learned quantity. The entire method is defined by B, and its values are fit to ImageNet training data. It directly determines the quantizer cell geometry.
  • Orthogonality penalty weight λ2 = Not reported
    Appendix A.3 says both too large and too small values hurt performance, but the chosen value is not given. It controls the soft orthogonality constraint that the method depends on.
  • Lattice dimension n = 32 for main results; 8, 16, 24 in comparisons
    The grouping of latent features into vectors of dimension n is a design choice. The paper does not provide a principled justification for 32 beyond experimental comparisons.
  • Initialization codebook size S in Eq. (8) = Not reported
    The basis initialization formula depends on an unspecified desired codebook size S, which affects the starting lattice density and therefore training.
assumptions (4)
  • domain assumption Babai rounding on an approximately orthogonal basis returns a lattice point close enough to the true nearest lattice point.
    Section 3.3 introduces Babai rounding in Eq. (6) and imposes orthogonality in Eq. (7) to reduce the rounding error, but no bound on the residual non-orthogonality or its effect on distortion and rate is provided.
  • domain assumption The integer coefficient vector m factorizes into independent univariate Gaussian mixture components under approximate orthogonality of B.
    Eq. (11) replaces the integral over the Voronoi polytope by a product of one-dimensional integrals. The factorization error is not quantified.
  • domain assumption Grouping latent features by channels captures the strongest inter-feature dependencies.
    Section 4.1 states that features across channels at the same spatial location are strongly correlated, motivating 32-dimensional channel-wise vectors. This is an empirical claim with no supporting analysis in the paper.
  • ad hoc to paper The soft orthogonality penalty term L in Eq. (13) keeps B sufficiently orthogonal throughout training.
    The method relies on a soft penalty with tuned weight λ2, but residual non-orthogonality is not measured and the training-inference gap is not quantified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Optimal Lattice Vector Quantizers for End-to-end Neural Image Compression." pith.science (2026). https://pith.science/paper/T3XG2J3W

@misc{pith2026241116119,
  author       = {Pith},
  title        = {Pith review of: Learning Optimal Lattice Vector Quantizers for End-to-end Neural Image Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/T3XG2J3W}},
  note         = {Machine review of arXiv:2411.16119}
}
read the original abstract

It is customary to deploy uniform scalar quantization in the end-to-end optimized Neural image compression methods, instead of more powerful vector quantization, due to the high complexity of the latter. Lattice vector quantization (LVQ), on the other hand, presents a compelling alternative, which can exploit inter-feature dependencies more effectively while keeping computational efficiency almost the same as scalar quantization. However, traditional LVQ structures are designed/optimized for uniform source distributions, hence nonadaptive and suboptimal for real source distributions of latent code space for Neural image compression tasks. In this paper, we propose a novel learning method to overcome this weakness by designing the rate-distortion optimal lattice vector quantization (OLVQ) codebooks with respect to the sample statistics of the latent features to be compressed. By being able to better fit the LVQ structures to any given latent sample distribution, the proposed OLVQ method improves the rate-distortion performances of the existing quantization schemes in neural image compression significantly, while retaining the amenability of uniform scalar quantization.

Figures

Figures reproduced from arXiv: 2411.16119 by the authors.

Figure 1
Figure 1. Two-dimensional visual examples for (a) scalar quantizer, (b) lattice vector quantizer and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Rate-distortion curves for the Bmshj2018 model on the Kodak and CLIC datasets, evaluated [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 57 canonical work pages

  1. [1]

    Optimization of lattices for quantization

    Erik Agrell and Thomas Eriksson. Optimization of lattices for quantization. IEEE Transactions on Information Theory, 44(5):1814–1828, 1998

    work page 1998

  2. [2]

    Closest point search in lattices

    Erik Agrell, Thomas Eriksson, Alexander Vardy, and Kenneth Zeger. Closest point search in lattices. IEEE transactions on information theory , 48(8):2201–2214, 2002

    work page 2002

  3. [3]

    Soft-to-hard vector quantization for end-to-end learning compressible representations

    Eirikur Agustsson, Fabian Mentzer, Michael Tschannen, Lukas Cavigelli, Radu Timofte, Luca Benini, and Luc Van Gool. Soft-to-hard vector quantization for end-to-end learning compressible representations. In Advances in Neural Information Processing Systems 30 , pages 1141–1151, 2017

    work page 2017

  4. [4]

    Generative adversarial networks for extreme learned image compression

    Eirikur Agustsson, Michael Tschannen, Fabian Mentzer, Radu Timofte, and Luc Van Gool. Generative adversarial networks for extreme learned image compression. In Proceedings of the IEEE International Conference on Computer Vision , pages 221–231, 2019

    work page 2019

  5. [5]

    On lovász’lattice reduction and the nearest lattice point problem

    László Babai. On lovász’lattice reduction and the nearest lattice point problem. Combinatorica, 6:1–13, 1986

    work page 1986

  6. [6]

    Simoncelli

    Johannes Ballé, Valero Laparra, and Eero P. Simoncelli. End-to-end optimized image compres- sion. In 5th International Conference on Learning Representations, ICLR , 2017

    work page 2017

  7. [7]

    Variational image compression with a scale hyperprior

    Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compression with a scale hyperprior. In 6th International Conference on Learning Representations, ICLR. OpenReview.net, 2018

    work page 2018

  8. [8]

    Workshop and challenge on learned image compression

    Johannes Ballé, George Toderici, Luca Versari, and etal. Workshop and challenge on learned image compression. http://compression.cc/, 2018

    work page 2018

Show all 62 references
  1. [9]

    Compressai: a py- torch library and evaluation platform for end-to-end compression research

    Jean Bégaint, Fabien Racapé, Simon Feltman, and Akshay Pushparaja. Compressai: a py- torch library and evaluation platform for end-to-end compression research. arXiv preprint arXiv:2011.03029, 2020

    2011 arXiv

  2. [10]

    Calculation of average psnr differences between rd-curves

    Gisle Bjontegaard. Calculation of average psnr differences between rd-curves. ITU SG16 Doc. VCEG-M33, 2001

    2001

  3. [11]

    Towards image compression with perfect realism at ultra-low bitrates

    Marlène Careil, Matthew J Muckley, Jakob Verbeek, and Stéphane Lathuilière. Towards image compression with perfect realism at ultra-low bitrates. In The Twelfth International Conference on Learning Representations, 2023

    2023

  4. [12]

    Learned image compression with discretized gaussian mixture likelihoods and attention modules

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto. Learned image compression with discretized gaussian mixture likelihoods and attention modules. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR , pages 7936–7945, 2020

    2020

  5. [13]

    Variable rate deep image compression with a conditional autoencoder

    Yoojin Choi, Mostafa El-Khamy, and Jungwon Lee. Variable rate deep image compression with a conditional autoencoder. In Proceedings of the IEEE International Conference on Computer Vision, pages 3146–3154, 2019

    2019

  6. [14]

    Kodak lossless true color image suite

    Eastman Kodak Company. Kodak lossless true color image suite. https://r0k.us/ graphics/kodak/, 2013

    2013

  7. [15]

    Imagenet: A large- scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large- scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  8. [16]

    Nvtc: Nonlinear vector transform cod- ing

    Runsen Feng, Zongyu Guo, Weiping Li, and Zhibo Chen. Nvtc: Nonlinear vector transform cod- ing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 6101–6110, 2023

    2023

  9. [17]

    Neural image compression via attentional multi-scale back projection and frequency decompo- sition

    Ge Gao, Pei You, Rong Pan, Shunyuan Han, Yuanyuan Zhang, Yuchao Dai, and Hojae Lee. Neural image compression via attentional multi-scale back projection and frequency decompo- sition. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 14677–1468...

    2021

  10. [18]

    Lattice quantization

    Jerry D Gibson and Khalid Sayood. Lattice quantization. In Advances in electronics and electron physics, volume 72, pages 259–330. Elsevier, 1988

    1988

  11. [19]

    Elic: Efficient learned image compression with unevenly grouped space-channel contextual adaptive coding

    Dailan He, Ziming Yang, Weikun Peng, Rui Ma, Hongwei Qin, and Yan Wang. Elic: Efficient learned image compression with unevenly grouped space-channel contextual adaptive coding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5718–5...

    2022

  12. [20]

    Checkerboard context model for efficient learned image compression

    Dailan He, Yaoyan Zheng, Baocheng Sun, Yan Wang, and Hongwei Qin. Checkerboard context model for efficient learned image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14771–14780, 2021

    2021

  13. [21]

    Mlic: Multi-reference entropy model for learned image compression

    Wei Jiang, Jiayu Yang, Yongqi Zhai, Peirong Ning, Feng Gao, and Ronggang Wang. Mlic: Multi-reference entropy model for learned image compression. In Proceedings of the 31st ACM International Conference on Multimedia, pages 7618–7627, 2023

    2023

  14. [22]

    Improved lossy image compression with priming and spatially adaptive bit rates for recurrent networks

    Nick Johnston, Damien Vincent, David Minnen, Michele Covell, Saurabh Singh, Troy Chinen, Sung Jin Hwang, Joel Shor, and George Toderici. Improved lossy image compression with priming and spatially adaptive bit rates for recurrent networks. In Proceedings of the IEEE Conference...

    2018

  15. [23]

    Ll-vq-vae: Learnable lattice vector-quantization for efficient representations

    Ahmed Khalil, Robert Piechocki, and Raul Santos-Rodriguez. Ll-vq-vae: Learnable lattice vector-quantization for efficient representations. arXiv preprint arXiv:2310.09382, 2023

    2023 arXiv

  16. [24]

    Joint global and local hierarchical priors for learned image compression

    Jun-Hyuk Kim, Byeongho Heo, and Jong-Seok Lee. Joint global and local hierarchical priors for learned image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5992–6001, 2022

    2022

  17. [25]

    Lvq-vae: End-to-end hyperprior-based variational image compression with lattice vector quantization

    Shinobu Kudo, Yukihiro Bandoh, Seishi Takamura, and Masaki Kitahara. Lvq-vae: End-to-end hyperprior-based variational image compression with lattice vector quantization. OpenReview, 2023

    2023

  18. [26]

    Context-adaptive entropy model for end-to-end optimized image compression

    Jooyoung Lee, Seunghyun Cho, and Seung-Kwon Beack. Context-adaptive entropy model for end-to-end optimized image compression. In 7th International Conference on Learning Representations, ICLR, 2019

    2019

  19. [27]

    Approaching rate-distortion limits in neural compression with lattice transform coding

    Eric Lei, Hamed Hassani, and Shirin Saeedi Bidokhti. Approaching rate-distortion limits in neural compression with lattice transform coding. arXiv preprint arXiv:2403.07320, 2024

    2024 arXiv

  20. [28]

    Content-oriented learned image compression

    Meng Li, Shangyin Gao, Yihui Feng, Yibo Shi, and Jing Wang. Content-oriented learned image compression. arXiv preprint arXiv:2207.14168, 2022

    2022 arXiv

  21. [29]

    Learning convolutional networks for content-weighted image compression

    Mu Li, Wangmeng Zuo, Shuhang Gu, Debin Zhao, and David Zhang. Learning convolutional networks for content-weighted image compression. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 3214–3223, 2018

    2018

  22. [30]

    A spatial rnn codec for end-to-end image compression

    Chaoyi Lin, Jiabao Yao, Fangdong Chen, and Li Wang. A spatial rnn codec for end-to-end image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13269–13277, 2020

    2020

  23. [31]

    Learned image compression with mixed transformer- cnn architectures

    Jinming Liu, Heming Sun, and Jiro Katto. Learned image compression with mixed transformer- cnn architectures. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14388–14397, 2023

    2023

  24. [32]

    Conditional probability models for deep image compression

    Fabian Mentzer, Eirikur Agustsson, Michael Tschannen, Radu Timofte, and Luc Van Gool. Conditional probability models for deep image compression. In 2018 IEEE Conference on Computer Vision and Pattern Recognition, CVPR, pages 4394–4402, 2018

    2018

  25. [33]

    High-fidelity generative image compression

    Fabian Mentzer, George D Toderici, Michael Tschannen, and Eirikur Agustsson. High-fidelity generative image compression. Advances in Neural Information Processing Systems , 33:11913– 11924, 2020

    2020

  26. [34]

    The hardness of the closest vector problem with preprocessing

    Daniele Micciancio. The hardness of the closest vector problem with preprocessing. IEEE Transactions on Information Theory, 47(3):1212–1215, 2001

    2001

  27. [35]

    Joint autoregressive and hierarchical priors for learned image compression

    David Minnen, Johannes Ballé, and George Toderici. Joint autoregressive and hierarchical priors for learned image compression. In Advances in Neural Information Processing Systems 31, pages 10794–10803, 2018

    2018

  28. [36]

    Channel-wise autoregressive entropy models for learned image compression

    David Minnen and Saurabh Singh. Channel-wise autoregressive entropy models for learned image compression. In 2020 IEEE International Conference on Image Processing (ICIP) , pages 3339–3343. IEEE, 2020. 12

    2020

  29. [37]

    Content adaptive latents and decoder for neural image compression

    Guanbo Pan, Guo Lu, Zhihao Hu, and Dong Xu. Content adaptive latents and decoder for neural image compression. In European Conference on Computer Vision, pages 556–573. Springer, 2022

    2022

  30. [38]

    Real-time adaptive image compression

    Oren Rippel and Lubomir Bourdev. Real-time adaptive image compression. arXiv preprint arXiv:1705.05823, 2017

    2017 arXiv

  31. [39]

    Arithmetic coding

    Jorma Rissanen and Glen G Langdon. Arithmetic coding. IBM Journal of research and development, 23(2):149–162, 1979

    1979

  32. [40]

    Multiple description lattice vector quantization

    Sergio D Servetto, Vinay A Vaishampayan, and Neil JA Sloane. Multiple description lattice vector quantization. In Proceedings DCC’99 Data Compression Conference (Cat. No. PR00096), pages 13–22. IEEE, 1999

    1999

  33. [41]

    Expanded adaptive scaling normalization for end to end image compression

    Chajin Shin, Hyeongmin Lee, Hanbin Son, Sangjin Lee, Dogyoon Lee, and Sangyoun Lee. Expanded adaptive scaling normalization for end to end image compression. In European Conference on Computer Vision, pages 390–405. Springer, 2022

    2022

  34. [42]

    Evc: Towards real-time neural image compres- sion with mask decay

    Guo-Hua Wang, Jiahao Li, Bin Li, and Yan Lu. Evc: Towards real-time neural image compres- sion with mask decay. arXiv preprint arXiv:2302.05071, 2023

    2023 arXiv

  35. [43]

    Arithmetic coding for data compression

    Ian H Witten, Radford M Neal, and John G Cleary. Arithmetic coding for data compression. Communications of the ACM, 30(6):520–540, 1987

    1987

  36. [44]

    Slimmable compressive au- toencoders for practical neural image compression

    Fei Yang, Luis Herranz, Yongmei Cheng, and Mikhail G Mozerov. Slimmable compressive au- toencoders for practical neural image compression. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4998–5007, 2021

    2021

  37. [45]

    Lossy image compression with conditional diffusion models

    Ruihan Yang and Stephan Mandt. Lossy image compression with conditional diffusion models. Advances in Neural Information Processing Systems , 36, 2024

    2024

  38. [46]

    Attention-guided image compression by deep reconstruction of compressive sensed saliency skeleton

    Xi Zhang and Xiaolin Wu. Attention-guided image compression by deep reconstruction of compressive sensed saliency skeleton. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13354–13364, 2021

    2021

  39. [47]

    Multi-modality deep restoration of extremely compressed face videos

    Xi Zhang and Xiaolin Wu. Multi-modality deep restoration of extremely compressed face videos. IEEE Transactions on Pattern Analysis and Machine Intelligence , 45(2):2024–2037, 2022

    2024

  40. [48]

    Lvqac: Lattice vector quantization coupled with spatially adap- tive companding for efficient learned image compression

    Xi Zhang and Xiaolin Wu. Lvqac: Lattice vector quantization coupled with spatially adap- tive companding for efficient learned image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10239–10248, 2023

    2023

  41. [49]

    Unified multivariate gaussian mixture for efficient neural image compression

    Xiaosu Zhu, Jingkuan Song, Lianli Gao, Feng Zheng, and Heng Tao Shen. Unified multivariate gaussian mixture for efficient neural image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17612–17621, 2022

    2022

  42. [50]

    Transformer-based transform coding

    Yinhao Zhu, Yang Yang, and Taco Cohen. Transformer-based transform coding. InInternational Conference on Learning Representations, 2021. 13 A Appendix / supplemental material A.1 Comparison with other VQ based image compression methods We provide the comparison results with th...

    2021

  43. [51]

    Compute the Gram-Schmidt orthogonalization of the basis B, resulting in an orthogonal basis B∗ = [b∗ 1, b∗ 2, . . . ,b∗ n]

  44. [52]

    Express the target vector t in terms of the orthogonal basis B∗: t = nX i=1 cib∗ i where ci are the coordinates of t in the Gram-Schmidt basis B∗

  45. [53]

    ,round(cn))

    Round each coordinate ci to the nearest integer: c′ = (round(c1), round(c2), . . . ,round(cn))

  46. [54]

    Construct the approximate lattice point: v = nX i=1 round(ci)bi A.4.2 Example Suppose we have a 2-dimensional lattice with basis vectors b1 and b2, and a target vector t

  47. [55]

    Compute the Gram-Schmidt basis: b∗ 1 = b1 b∗ 2 = b2 − ⟨b2, b1⟩ ∥b1∥2 b1

  48. [56]

    Project t onto b∗ 1 and b∗ 2: c1 = ⟨t, b∗ 1⟩ ∥b∗ 1∥2 c2 = ⟨t, b∗ 2⟩ ∥b∗ 2∥2

  49. [57]

    Round c1 and c2: round(c1), round(c2)

  50. [58]

    Construct the approximate lattice point: v = round(c1)b1 + round(c2)b2 A.4.3 Proof of the Bound We want to show that the distance between the target vectort and the found lattice point v is bounded by half the sum of the lengths of the basis vectors

  51. [59]

    Decompose the Target Vector: t = nX i=1 cib∗ i Let t′ be the projection of t onto the lattice: t′ = nX i=1 round(ci)b∗ i 15 The difference between t and t′ is: t − t′ = nX i=1 (ci − round(ci))b∗ i Each term ci − round(ci) is at most 1 2 in magnitude because round(ci) is the ne...

  52. [60]

    Bound the Difference: ∥t − t′∥ ≤ nX i=1 |ci − round(ci)|∥b∗ i ∥ Since |ci − round(ci)| ≤1 2, we have: ∥t − t′∥ ≤ nX i=1 1 2 ∥b∗ i ∥ = 1 2 nX i=1 ∥b∗ i ∥

  53. [61]

    Since v and t′ differ only by the orthogonalization process, the distance bound remains valid

    Reconstruct the Lattice Point : The lattice point v constructed by Babai’s rounding algorithm is: v = nX i=1 round(ci)bi Note that t′ is the projection in the Gram-Schmidt basis, while v is the actual lattice point in the original basis. Since v and t′ differ only by the ortho...

  54. [62]

    A.5 Broader Impact The proposed Optimal Lattice Vector Quantization (LVQ) method for image compression offers several significant and wide-reaching benefits

    Final Bound: Since the Gram-Schmidt process does not increase the lengths of the basis vectors, we can state that: ∥t − v∥ ≤1 2 nX i=1 ∥bi∥ Thus, Babai’s rounding algorithm guarantees that the distance between the target vector t and the found lattice point v is within half th...

Pith tools

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