Pith. sign in

REVIEW 3 major objections 5 minor 59 references

WTDUN: Wavelet Tree-Structured Sampling and Deep Unfolding Network for Image Compressed Sensing

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 WTDUN claims that organizing both sampling and reconstruction around the wavelet coefficient tree lifts image compressed sensing to higher PSNR than prior deep-learning methods at every tested sampling ratio.

desk verdict A well-engineered wavelet-domain deep unfolding system for image CS whose headline gains rest on an unstated oracle advantage in the adaptive sampling rule; fix that and the derivation slip, and it is a solid paper. read the letter →

arxiv 2411.16336 v1 pith:JWVJ7PAN submitted 2024-11-25 eess.IV cs.CV

Kai Han , Jin Wang , Yunhui Shi , Hanqin Cai , Nam Ling , Baocai Yin This is my paper · ORCID
classification eess.IVcs.CV
keywords compressedsensingdeepunfoldingwavelettransformadaptivesamplingtree-structuredsparsityimagereconstructionmulti-scalesubbands
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that image compressed sensing improves when both sampling and reconstruction are organized by the multi-scale wavelet structure of an image instead of treating all pixels or all frequency bands alike. Its method, WTDUN, does two linked things: it assigns the limited measurement budget across wavelet subbands according to each subband's energy and sparsity, and it unfolds an iterative optimization into a deep network whose steps are guided by the parent-child tree structure of wavelet coefficients. On Set5, Set11, Set14, and Urban100, the 128×128 variant of WTDUN reports higher PSNR than every compared deep-learning method at all tested sampling ratios, with gains of 0.20–2.22 dB at 10% and 1.01–5.57 dB at 50% on Urban100. If correct, this says that frequency-aware adaptive sampling plus cross-scale reconstruction is the right design direction for practical compressed-sensing encoders.

What carries the argument

The carrying object is the wavelet coefficient tree. A multi-level Haar transform splits each image block into a low-frequency LL subband plus high-frequency subbands at several scales, and WTDUN treats those subbands as leaves of a quadtree. In sampling, the standard deviation and mean of each subband's absolute coefficients are weighted to set the number of measurements that subband receives (Eq. 6 and the allocation algorithm of Section 3.2.1), so low-frequency energy and high-frequency texture each get a proportionate share of the budget. In reconstruction, the tree-structured prior expresses the structural sparsity of the quadtree: coefficients in parent-child groups are encouraged to shrink together through a group-wise soft threshold (Eq. 12), and an unfolded gradient descent module alternates with denoising and deblocking. The tree is what lets the network transfer information between scales, which the paper credits for its improved texture and edge recovery.

What would settle it

Run WTDUN on Urban100 at 50% sampling with the per-subband measurement allocation replaced by a uniform allocation of the same total budget while keeping every other component and the training procedure fixed; the paper's own ablations predict a PSNR drop of about 1.5–2.5 dB, so a result showing parity or a uniform-allocation win would refute the adaptive-sampling claim. Independently, reproducing the reported absolute PSNR numbers on the four benchmarks with the paper's protocol would confirm or refute the headline performance claim.

Watch

Extended reading notes

Core claim

The central discovery is that wavelet-domain adaptive allocation of measurements is a large, mostly untapped source of reconstruction gain in learned compressed sensing. The paper's ablation attributes about 1.5–2.5 dB of PSNR on Urban100 and Set11 to the subband-importance allocation, about 0.08–0.2 dB to the wavelet tree-structured prior, and about 1–2.5 dB to full-image deblocking. The full architecture—initial pseudo-inverse reconstruction, nine unfolded stages alternating gradient descent with group-wise soft shrinkage on parent-child coefficient groups, full-image deblocking, and cross-domain attention—achieves the best numbers in the comparison tables on all four datasets at every tested sampling ratio, with the largest margins at the 50% ratio where the measurement budget is most flexible.

Load-bearing premise

The performance figures presuppose that the encoder may look at the original image's wavelet subband statistics and spend a different number of measurements on each subband of each image, with the decoder knowing exactly how the budget was spent; if compressed sensing must use one fixed measurement operator independent of content, or if the allocation costs side information that must fit inside the same bit budget, the advertised 1.5–2.5 dB from adaptive sampling will not transfer.

Editorial extensions

If this is right

  • If the reported numbers hold, WTDUN achieves the highest PSNR on these block-based compressed-sensing benchmarks at 10–50% sampling ratios, with the largest margins at higher ratios.
  • The ablation isolates where the gain comes from: adaptive subband allocation accounts for roughly 1.5–2.5 dB, so future learned CS encoders should treat measurement allocation as a first-class design variable rather than a fixed random matrix.
  • The wavelet tree-structured prior, though smaller in effect at about 0.08–0.2 dB, is a checkable mechanism for recovering fine textures and can be added to other unfolding networks without changing their sampling setup.
  • Full-image deblocking contributes about 1–2.5 dB, meaning block-based methods that ignore inter-block context are leaving a large part of achievable quality on the table.
  • Two-level wavelet decomposition with 64×64 blocks beats one-level and three-level decomposition, so the hierarchy depth matters: too shallow loses sparsity, and too deep starves the small subbands of features the network can use.

Reading between the lines

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

  • A practical consequence the paper does not develop: the adaptive allocation depends on the original image's own wavelet statistics, so a real single-shot sensor would need to transmit the allocation table as side information or estimate it from a coarse preview; the reported gains should be re-measured once that overhead is charged to the measurement budget.
  • The same wavelet-tree prior could be transplanted into other deep unfolding networks; since the paper's ablation shows the prior's independent gain is modest, a fair comparison would hold the sampling protocol fixed and swap only the reconstruction prior.
  • The large margin at 50% over uniform-sampling baselines suggests that most remaining headroom in block compressed sensing lies in how measurements are spent, not just in the reconstruction network, which generalizes the result beyond this specific architecture.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 WTDUN, a deep unfolding network for image compressed sensing that operates on Haar wavelet subbands. It combines a wavelet-domain adaptive sampling (WAS) module that allocates measurement counts per subband according to subband energy and sparsity, a tree-structured reconstruction prior encoded through group soft-thresholding, and K unfolding stages with deblocking and cross-domain attention. The authors report PSNR/SSIM gains over several recent deep CS methods on Set5, Set11, Set14, and Urban100, and ablations attribute large gains to WAS and to the deblocking module.

Significance. If the reported results are taken at face value and the sampling protocol is a legitimate compressed-sensing protocol, WTDUN is a strong empirical contribution: the tables cover four benchmarks and five sampling ratios, the ablation study isolates the effect of each module, and the WAS ablation cleanly demonstrates a large effect on reconstruction quality. However, two load-bearing issues, namely the self-referential update in Eq. (5) and the oracle-like adaptive sampling protocol in Section 3.2.1, mean that the paper's central claims of state-of-the-art performance and of optimization grounding are not yet established.

major comments (3)
  1. [Section 3.1, Eq. (5)] The update x^k = A^T z^{k-1} + x^{k-1} - (A^T A - I)(x - x^{k-1}) defines the new iterate in terms of the unknown ground-truth signal x. This cannot serve as a derivation of a computable reconstruction iteration, and it does not provide optimization grounding for the network, because the implemented module replaces the offending term with the learned denoising block N^k. Please replace this derivation with a valid proximal or denoising update, or state explicitly that the network is only inspired by this identity rather than derived from it.
  2. [Section 3.2.1 and Section 4.3.1] The WAS allocation rule uses the mean and standard deviation of |theta_s|, where theta_s = Psi X is the wavelet transform of the original image X. In one-shot compressed sensing, the encoder only has y = A x and cannot compute these statistics before choosing the measurement counts M_s. If the intended protocol is adaptive, the per-image or per-block allocation must be transmitted to the decoder as side information, and neither Tables 6-7 nor the comparisons in Tables 1-4 account for this cost; none of the comparison methods is given original-image subband statistics. Since the WAS ablation alone contributes about 1.5-2.5 dB at 20% and 50% ratios, the claim of higher PSNR than fixed-operator or learned-operator baselines at all sampling ratios is not supported by the current experiments. Please specify the inference-time protocol, report the side-information overhead, or evaluate under a fixed measurement operator.
  3. [Section 3.3.2, Eq. (18)] The text says the denoising block N^k(theta^{k-1}) removes noise from theta^{k-1}, but Eq. (18) and the surrounding derivation require this term to play the role of the unknown error (x - x^{k-1}) from Eq. (5). These are different objects, and a CNN block operating on the current estimate does not obviously compute either one. Please clarify the role of N^k; if it is meant to approximate the error term, explain how it is trained to do so.
minor comments (5)
  1. [Eq. (9)] The index condition 1 < i < L is inconsistent with Eq. (6), where 1 <= i <= l; it should presumably be 1 <= i <= L or the notation should be aligned.
  2. [Section 3.2.1] The notation for the bias term, written as (n sum_{i=0} bias_i = 0), is garbled; please define bias_i and its index range explicitly.
  3. [Section 4.1] The text says the control parameter alpha is initialized as 1, but Section 3.2.1 introduced the allocation control parameter eta; please clarify whether alpha and eta are the same parameter or different ones.
  4. [Tables 1 and 2] Several entries for DPA-Net are missing (denoted by *), yet Section 4.2.1 claims higher PSNR at all sampling ratios; please state explicitly that incomplete comparisons are excluded from that claim.
  5. [Eq. (12)] The soft-threshold formula divides by the norm of G theta^{k-1}; please specify the behavior when this norm is zero, for example by defining the expression to be zero in that case.

Circularity Check

2 steps flagged · score 6.0 of 10

The unfolding step in Eq. (5) is a tautology involving the unknown image, and the WAS gain (1.5–2.5 dB) is obtained by allocating measurements from ground-truth subband statistics, so the reported SOTA advantage is partially circular and oracle-driven.

  1. self definitional [Section 3.1, Eq. (5)]
    "The clean signal can be obtained by calculating: 𝒙𝑘 = 𝑨𝑇 𝒛𝑘−1+ 𝒙𝑘−1−( 𝑨𝑇 𝑨− 𝑰)( 𝒙− 𝒙𝑘−1)."

    The right-hand side contains the unknown ground-truth signal 𝒙, so Eq. (5) is an algebraic identity obtained by substituting 𝒆 = (𝑨𝑇𝑨−𝑰)(𝒙−𝒙𝑘−1) into 𝒙 = 𝑨𝑇𝒛𝑘−1+𝒙𝑘−1−𝒆. It is not a computable update for 𝒙𝑘; the quantity to be recovered appears on both sides, so the 'clean signal' is defined in terms of itself. The implemented network replaces this term with a learned denoiser in Eq. (18), so actual performance comes from a data-fitted network rather than from this self-referential equation.

  2. other [Section 3.2.1 (Eq. (6) and allocation algorithm); gain reported in Section 4.3.1]
    "𝜽𝑠 = Ψ𝑿, ... In the allocation algorithm, we initially calculate the standard deviation 𝜎 and mean 𝜇 of absolute values |𝜽| for each wavelet subband. ... By utilizing the percentages 𝑃𝐻 and 𝑃𝐿𝐿, we can derive the corresponding quantities of CS measurements."

    Eq. (6) defines 𝜽 from the original image 𝑿, and the allocation weight for each subband is computed from the mean and standard deviation of |𝜽| of that same original image. Thus the per-subband measurement counts are functions of the ground-truth signal that the CS system is supposed to recover. At test time, in a one-shot CS protocol the encoder only has 𝒚 = 𝑨𝒙 and cannot know |𝜽| without an oracle or an uncompensated side channel. The reported WAS ablation gain of about 1.5–2.5 dB therefore comes from using the answer to choose the measurement operator, not from a prediction based on compressive measurements; comparison methods are evaluated with fixed operators without this oracle information.

full rationale

The core deep-unfolding reconstruction modules are independently implemented with learned denoisers and are not reduced to a self-citation chain; there are no load-bearing self-citations or imported uniqueness theorems. However, the paper's derivation and evaluation contain two partially circular elements. First, Eq. (5) presents a 'clean signal' update whose right-hand side includes the unknown 𝒙, making it a tautology rather than a derivation; the network later replaces that term with a learned denoiser, so the optimization grounding is not the true source of performance. Second, the wavelet-domain adaptive sampling (WAS) component allocates measurements from the original image's subband statistics, meaning the decoder receives information derived from the ground-truth image. Section 4.3.1 attributes 1.5–2.5 dB to this component, and the paper's headline claim of higher PSNR at all sampling ratios rests substantially on this oracle-driven allocation. The reconstruction architecture itself has independent empirical content, but the central state-of-the-art claim is partially circular because a major reported gain is obtained by using the target signal to design the measurement operator.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method's central claim rests on a large, unspecified set of learned network weights plus several hand-set hyperparameters in the allocation algorithm (eta/alpha, Theta, bias). It also assumes natural images obey a wavelet quad-tree sparsity structure and that a CNN can learn the correction term that the derivation defines via ground truth. These are plausible engineering assumptions but are not independently evidenced.

free parameters (6)
  • Allocation control parameter eta (and alpha in Section 4.1) = initialized to 1
    Section 3.2.1 uses eta to weight sigma and mu; Section 4.1 initializes alpha to 1. The final value or update rule is not reported.
  • Upper bound Theta for low-frequency measurements
    Introduced in Section 3.2.1 to prevent sample overflow; its value is never specified.
  • Bias terms b_i in measurement allocation
    Added in Section 3.2.1 with the constraint that they sum to zero; individual values are not given.
  • Regularization parameters beta and lambda in the tree-prior model
    Appear in Eqs. (8), (11), and (14); in the implementation they are absorbed into learned thresholds and group weights, so no explicit fitted values are reported.
  • Loss regularization weights gamma and mu = gamma=0.001, mu=0.01
    Stated in Section 3.4; chosen by hand without sensitivity analysis.
  • Unfolding stages K and wavelet decomposition level = K=9, level=2
    Selected from experiments in Sections 4.3.5 and 4.3.6 after observing PSNR saturation; this is a post-hoc design choice.
assumptions (4)
  • domain assumption Wavelet coefficients of natural images form a quad-tree structure where parent and child coefficients tend to vanish together.
    Invoked in Section 3.3 to justify the group-sparsity term ||G theta||_2 in Eq. (8); treated as a property of natural images with no demonstration on the test datasets.
  • ad hoc to paper Subband energy (mean of absolute values) and sparsity (standard deviation) are reliable signals for allocating CS measurements.
    Section 3.2.1 defines the allocation weight from sigma and mu; this heuristic is introduced specifically for this paper and is not derived from rate-distortion or CS theory.
  • ad hoc to paper The noise term (A^T A - I)(x - x^(k-1)) can be approximated by a learned denoiser from the previous iterate alone.
    Eq. (18) replaces the ground-truth-dependent correction in Eq. (5) with a learned denoiser; this learnability premise is load-bearing for the unfolding claim.
  • domain assumption Block-wise wavelet sampling with fixed block sizes is an appropriate model for image CS and does not lose critical global context.
    Section 3.2.2 partitions images into n x n blocks; the need for a deblocking module in Section 3.3.2 acknowledges the artifacts this assumption introduces.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WTDUN: Wavelet Tree-Structured Sampling and Deep Unfolding Network for Image Compressed Sensing." pith.science (2026). https://pith.science/paper/JWVJ7PAN

@misc{pith2026241116336,
  author       = {Pith},
  title        = {Pith review of: WTDUN: Wavelet Tree-Structured Sampling and Deep Unfolding Network for Image Compressed Sensing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JWVJ7PAN}},
  note         = {Machine review of arXiv:2411.16336}
}
read the original abstract

Deep unfolding networks have gained increasing attention in the field of compressed sensing (CS) owing to their theoretical interpretability and superior reconstruction performance. However, most existing deep unfolding methods often face the following issues: 1) they learn directly from single-channel images, leading to a simple feature representation that does not fully capture complex features; and 2) they treat various image components uniformly, ignoring the characteristics of different components. To address these issues, we propose a novel wavelet-domain deep unfolding framework named WTDUN, which operates directly on the multi-scale wavelet subbands. Our method utilizes the intrinsic sparsity and multi-scale structure of wavelet coefficients to achieve a tree-structured sampling and reconstruction, effectively capturing and highlighting the most important features within images. Specifically, the design of tree-structured reconstruction aims to capture the inter-dependencies among the multi-scale subbands, enabling the identification of both fine and coarse features, which can lead to a marked improvement in reconstruction quality. Furthermore, a wavelet domain adaptive sampling method is proposed to greatly improve the sampling capability, which is realized by assigning measurements to each wavelet subband based on its importance. Unlike pure deep learning methods that treat all components uniformly, our method introduces a targeted focus on important subbands, considering their energy and sparsity. This targeted strategy lets us capture key information more efficiently while discarding less important information, resulting in a more effective and detailed reconstruction. Extensive experimental results on various datasets validate the superior performance of our proposed method.

Figures

Figures reproduced from arXiv: 2411.16336 by the authors.

Figure 1
Figure 1. Illustration of WTDUN, which consists of a sampling module, an initial module, and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The detailed design of one single phase in WTDUN. The GDM(gradient descent module) represents [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Detailed design of each module. (1) indicates the sampling process. (2) is the initial reconstruction [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Visual quality comparisons between our proposed method and recently state-of-the-art CS methods [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Visual quality comparisons between our WTDUN and recently state-of-the-art CS methods on Set11 [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Visual quality comparisons between our WTDUN and recently state-of-the-art CS methods on Set14 [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Visual quality comparisons between the proposed WTDUN and recently state-of-the-art CS methods [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: The PSNR/SSIM curve with the increase of iterative stage [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 59 canonical work pages

  1. [1]

    Amir Beck and Marc Teboulle. 2009. A fast Iterative Shrinkage-Thresholding Algorithm with application to wavelet- based image deblurring. In IEEE International Conference on Acoustics, Speech and Signal Processing(ICASSP) . 693–696

    work page 2009

  2. [2]

    Candes, J

    E.J. Candes, J. Romberg, and T. Tao. 2006. Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information. IEEE Transactions on Information Theory 52, 2 (2006), 489–509

    work page 2006

  3. [3]

    Thuong Nguyen Canh and Byeungwoo Jeon. 2018. Multi-Scale Deep Compressive Sensing Network. In 2018 IEEE Visual Communications and Image Processing (VCIP) . 1–4

    work page 2018

  4. [4]

    Cheng Chen and Junzhou Huang. 2014. The benefit of tree sparsity in accelerated MRI. Medical image analysis 18, 6 (2014), 834–842

    work page 2014

  5. [5]

    Jiwei Chen, Yubao Sun, Qingshan Liu, and Rui Huang. 2020. Learning Memory Augmented Cascading Network for Compressed Sensing of Images. In European Conference on Computer Vision

    work page 2020

  6. [6]

    Donoho, and Michael A

    Scott Saobing Chen, David L. Donoho, and Michael A. Saunders. 1998. Atomic Decomposition by Basis Pursuit. SIAM J. Sci. Comput. 20 (1998), 33–61. WTDUN: Wavelet Tree-Structured Sampling and Deep Unfolding Network for Image Compressed Sensing 19

    work page 1998

  7. [7]

    Wenjun Chen, Chunling Yang, and Xin Yang. 2022. FSOINET: Feature-Space Optimization-Inspired Network For Image Compressive Sensing. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . 2460–2464

    work page 2022

  8. [8]

    Zan Chen, Xingsong Hou, Xueming Qian, and Chen Gong. 2018. Efficient and Robust Image Coding and Transmission Based on Scrambled Block Compressive Sensing. IEEE Transactions on Multimedia 20, 7 (2018), 1610–1621

    work page 2018

Show all 59 references
  1. [9]

    Wenxue Cui, Shaohui Liu, and Debin Zhao. 2022. Fast Hierarchical Deep Unfolding Network for Image Compressed Sensing. In Proceedings of the 30th ACM International Conference on Multimedia . 2739–2748

    2022

  2. [10]

    D.L. Donoho. 2006. Compressed sensing. IEEE Transactions on Information Theory 52, 4 (2006), 1289–1306

    2006

  3. [11]

    Donoho, Arian Maleki, and Andrea Montanari

    David L. Donoho, Arian Maleki, and Andrea Montanari. 2010. Message passing algorithms for compressed sensing: I. motivation and construction. In 2010 IEEE Information Theory Workshop on Information Theory (ITW 2010, Cairo) . 1–5

    2010

  4. [12]

    Jiang Du, Xuemei Xie, Chenye Wang, Guangming Shi, Xun Xu, and Yuxiang Wang. 2019. Fully convolutional measurement network for compressive sensing image reconstruction. Neurocomputing 328 (2019), 105–112

    2019

  5. [13]

    Zi-En Fan, Feng Lian, and Jia-Ni Quan. 2022. Global Sensing and Measurements Reuse for Image Compressed Sensing. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 8944–8953

    2022

  6. [14]

    Jing Fang, Yinbo Yu, Zhongyuan Wang, Xin Ding, and Ruimin Hu. 2023. An Image Arbitrary-Scale Super-Resolution Network Using Frequency-Domain Information. ACM Trans. Multimedia Comput. Commun. Appl. 20, 3 (2023)

    2023

  7. [15]

    Fowler, Sungkwang Mun, and Eric W

    James E. Fowler, Sungkwang Mun, and Eric W. Tramel. 2011. Multiscale block compressed sensing with smoothed projected Landweber reconstruction. In 2011 19th European Signal Processing Conference . 564–568

    2011

  8. [16]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative Adversarial Nets. In Proceedings of the 27th International Conference on Neural Information Processing Systems. 2672–2680

    2014

  9. [17]

    Tao Huang, Weisheng Dong, Xin Yuan, Jinjian Wu, and Guangming Shi. 2021. Deep Gaussian Scale Mixture Prior for Spectral Compressive Imaging. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 16211–16220

    2021

  10. [18]

    Tao Huang, Xin Yuan, Weisheng Dong, Jinjian Wu, and Guangming Shi. 2023. Deep Gaussian Scale Mixture Prior for Image Reconstruction. IEEE Transactions on Pattern Analysis and Machine Intelligence 45, 9 (2023), 10778–10794

    2023

  11. [19]

    Shihao Ji, Ya Xue, and Lawrence Carin. 2008. Bayesian Compressive Sensing. IEEE Transactions on Signal Processing 56, 6 (2008), 2346–2356

    2008

  12. [20]

    Zutao Jiang, Changlin Li, Xiaojun Chang, Ling Chen, Jihua Zhu, and Yi Yang. 2023. Dynamic Slimmable Denoising Network. IEEE Transactions on Image Processing 32 (2023), 1583–1598

    2023

  13. [21]

    Chengbo Li, Wotao Yin, Hong Jiang, and Yin Zhang. 2013. An efficient augmented Lagrangian method with applications to total variation minimization. Computational Optimization and Applications 56 (2013), 507 – 530

    2013

  14. [22]

    Weiqi Li, Bin Chen, Shuai Liu, Shijie Zhao, Bowen Du, Yongbing Zhang, and Jian Zhang. 2024. D3C2-Net: Dual-Domain Deep Convolutional Coding Network for Compressive Sensing. IEEE Transactions on Circuits and Systems for Video Technology 34, 10 (2024), 9341–9355

    2024

  15. [23]

    Suhas Lohit, Kuldeep Kulkarni, Ronan Kerviche, Pavan Turaga, and Amit Ashok. 2018. Convolutional Neural Networks for Noniterative Reconstruction of Compressively Sensed Images. IEEE Transactions on Computational Imaging 4, 3 (2018), 326–340

    2018

  16. [24]

    Jiawei Ma, Xiao-Yang Liu, Zheng Shou, and Xin Yuan. 2019. Deep Tensor ADMM-Net for Snapshot Compressive Imaging. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV) . 10222–10231

    2019

  17. [25]

    Mallat and Z

    S. Mallat and Z. Zhang. 1992. Adaptive time-frequency decomposition with matching pursuits. In Proceedings of the IEEE-SP International Symposium on Time-Frequency and Time-Scale Analysis . 7–10

    1992

  18. [26]

    Metzler, Ali Mousavi, and Richard G

    Christopher A. Metzler, Ali Mousavi, and Richard G. Baraniuk. 2017. Learned D-AMP: Principled Neural Network Based Compressive Image Recovery. InProceedings of the 31st International Conference on Neural Information Processing Systems(NIPS). 1770–1781

    2017

  19. [27]

    Patel, and Richard G

    Ali Mousavi, Ankit B. Patel, and Richard G. Baraniuk. 2015. A deep learning approach to structured signal recovery. In 2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) . 1336–1343

    2015

  20. [28]

    Sungkwang Mun and James E. Fowler. 2009. Block compressed sensing of images using directional transforms. In 2009 16th IEEE International Conference on Image Processing (ICIP) . 3021–3024

    2009

  21. [29]

    Sujit Kumar Sahoo and Anamitra Makur. 2015. Signal Recovery from Random Measurements via Extended Orthogonal Matching Pursuit. IEEE Transactions on Signal Processing 63, 10 (2015), 2572–2581

    2015

  22. [30]

    Minghe Shen, Hongping Gan, Chao Ning, Yi Hua, and Tao Zhang. 2022. TransCS: A Transformer-Based Hybrid Architecture for Image Compressed Sensing. IEEE Transactions on Image Processing 31 (2022), 6991–7005

    2022

  23. [31]

    Wuzhen Shi, Feng Jiang, Shaohui Liu, and Debin Zhao. 2019. Scalable Convolutional Neural Network for Image Compressed Sensing. In 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 12282–12291

    2019

  24. [32]

    Wuzhen Shi, Feng Jiang, Shaohui Liu, and Debin Zhao. 2020. Image Compressed Sensing Using Convolutional Neural Network. IEEE Transactions on Image Processing 29 (2020), 375–388. 20 Kai Han, Jin Wang, Yunhui Shi, Hanqin Cai, Nam Ling, and Baocai Yin

    2020

  25. [33]

    Wuzhen Shi, Shaohui Liu, Feng Jiang, and Debin Zhao. 2021. Video Compressed Sensing Using a Convolutional Neural Network. IEEE Transactions on Circuits and Systems for Video Technology 31, 2 (2021), 425–438

    2021

  26. [34]

    Jiechong Song, Bin Chen, and Jian Zhang. 2021. Memory-Augmented Deep Unfolding Network for Compressive Sensing. In Proceedings of the 29th ACM International Conference on Multimedia . 4249–4258

    2021

  27. [35]

    Jiechong Song, Bin Chen, and Jian Zhang. 2023. Dynamic Path-Controllable Deep Unfolding Network for Compressive Sensing. IEEE Transactions on Image Processing 32 (2023), 2202–2214

    2023

  28. [36]

    Jiechong Song and Jian Zhang. 2023. SODAS-Net: Side-Information-Aided Deep Adaptive Shrinkage Network for Compressive Sensing. IEEE Transactions on Instrumentation and Measurement 72 (2023), 1–12

    2023

  29. [37]

    Yubao Sun, Jiwei Chen, Qingshan Liu, Bo Liu, and Guodong Guo. 2020. Dual-Path Attention Network for Compressed Sensing Image Reconstruction. IEEE Transactions on Image Processing 29 (2020), 9482–9495

    2020

  30. [38]

    Yubao Sun, Ying Yang, Qingshan Liu, Jiwei Chen, Xiao-Tong Yuan, and Guodong Guo. 2020. Learning Non-Locally Regularized Compressed Sensing Network With Half-Quadratic Splitting. IEEE Transactions on Multimedia 22, 12 (2020), 3236–3248

    2020

  31. [39]

    Anhong Wang, Lei Liu, Bing Zeng, and Huihui Bai. 2011. Progressive image coding based on an adaptive block compressed sensing. IEICE Electron. Express 8 (2011), 575–581

    2011

  32. [40]

    Xuemei Xie, Chenye Wang, Jiang Du, and Guangming Shi. 2018. Full Image Recover for Block-Based Compressive Sensing. In 2018 IEEE International Conference on Multimedia and Expo (ICME) . 1–6

    2018

  33. [41]

    Xuemei Xie, Yuxiang Wang, Guangming Shi, Chenye Wang, Jiang Du, and Xiao Han. 2017. Adaptive Measurement Network for CS Image Reconstruction. In Computer Vision. 407–417

    2017

  34. [42]

    Jingyu Yang, Xiangjun Yin, Mengxi Zhang, Huihui Yue, Xingyu Cui, and Huanjing Yue. 2022. Learning Image Formation and Regularization in Unrolling AMP for Lensless Image Reconstruction. IEEE Transactions on Computational Imaging 8 (2022), 479–489

    2022

  35. [43]

    Peihao Yang, Linghe Kong, Meikang Qiu, Xue Liu, and Guihai Chen. 2021. Compressed Imaging Reconstruction with Sparse Random Projection. ACM Trans. Multimedia Comput. Commun. Appl. 17, 1 (2021)

    2021

  36. [44]

    Yuxiang Yang, Qi Cao, Jing Zhang, and Dacheng Tao. 2022. CODON: On Orchestrating Cross-Domain Attentions for Depth Super-Resolution. Int. J. Comput. Vision 130, 2 (2022), 267–284

    2022

  37. [45]

    Yan Yang, Jian Sun, Huibin Li, and Zongben Xu. 2020. ADMM-CSNet: A Deep Learning Approach for Image Compressive Sensing. IEEE Transactions on Pattern Analysis and Machine Intelligence 42, 3 (2020), 521–538

    2020

  38. [46]

    Yixiao Yang, Ran Tao, Kaixuan Wei, and Ying Fu. 2022. Dynamic proximal unrolling network for compressive imaging. Neurocomputing 510 (2022), 203–217

    2022

  39. [47]

    Dongjie Ye, Zhangkai Ni, Hanli Wang, Jian Zhang, Shiqi Wang, and Sam Kwong. 2023. CSformer: Bridging Convolution and Transformer for Compressive Sensing. IEEE Transactions on Image Processing 32 (2023), 2827–2842

    2023

  40. [48]

    Di You, Jingfen Xie, and Jian Zhang. 2021. ISTA-NET++: Flexible Deep Unfolding Network for Compressive Sensing. In 2021 IEEE International Conference on Multimedia and Expo (ICME) . 1–6

    2021

  41. [49]

    Di You, Jian Zhang, Jingfen Xie, Bin Chen, and Siwei Ma. 2021. COAST: COntrollable Arbitrary-Sampling NeTwork for Compressive Sensing. IEEE Transactions on Image Processing 30 (2021), 6066–6080

    2021

  42. [50]

    Roman Zeyde, Michael Elad, and Matan Protter. 2012. On Single Image Scale-Up Using Sparse Representations. In Curves and Surfaces. 711–730

    2012

  43. [51]

    Jian Zhang and Bernard Ghanem. 2018. ISTA-Net: Interpretable Optimization-Inspired Deep Network for Image Compressive Sensing. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 1828–1837

    2018

  44. [52]

    Jian Zhang, Zhenyu Zhang, Jingfen Xie, and Yongbing Zhang. 2022. High-Throughput Deep Unfolding Network for Compressive Sensing MRI. IEEE Journal of Selected Topics in Signal Processing 16, 4 (2022), 750–761

    2022

  45. [53]

    Jian Zhang, Chen Zhao, and Wen Gao. 2020. Optimization-Inspired Compact Deep Compressive Sensing. IEEE Journal of Selected Topics in Signal Processing 14, 4 (2020), 765–774

    2020

  46. [54]

    Jian Zhang, Debin Zhao, and Wen Gao. 2014. Group-Based Sparse Representation for Image Restoration. IEEE Transactions on Image Processing 23, 8 (2014), 3336–3351

    2014

  47. [55]

    Jian Zhang, Debin Zhao, Ruiqin Xiong, Siwei Ma, and Wen Gao. 2014. Image Restoration Using Joint Statistical Modeling in a Space-Transform Domain. IEEE Transactions on Circuits and Systems for Video Technology 24, 6 (2014), 915–928

    2014

  48. [56]

    Zhonghao Zhang, Yipeng Liu, Jiani Liu, Fei Wen, and Ce Zhu. 2021. AMP-Net: Denoising-Based Deep Unfolding for Compressive Image Sensing. IEEE Transactions on Image Processing 30 (2021), 1487–1500

    2021

  49. [57]

    Chen Zhao, Siwei Ma, Jian Zhang, Ruiqin Xiong, and Wen Gao. 2017. Video Compressive Sensing Reconstruction via Reweighted Residual Sparsity. IEEE Transactions on Circuits and Systems for Video Technology 27, 6 (2017), 1182–1195

    2017

  50. [58]

    Siwang Zhou, Yan He, Yonghe Liu, Chengqing Li, and Jianming Zhang. 2021. Multi-Channel Deep Networks for Block-Based Image Compressive Sensing. IEEE Transactions on Multimedia 23 (2021), 2627–2640

    2021

  51. [59]

    Shuyuan Zhu, Bing Zeng, and Moncef Gabbouj. 2014. Adaptive reweighted compressed sensing for image compression. In 2014 IEEE International Symposium on Circuits and Systems (ISCAS) . 1–4

    2014

Pith tools

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