Pith. sign in

REVIEW 4 major objections 5 minor 57 references

Self-supervised Deep Hyperspectral Inpainting with the Plug and Play and Deep Image Prior Models

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A convergence guarantee for self-supervised hyperspectral inpainting with a deep prior.

desk verdict The empirical work is solid and reproducible, but the claimed convergence theorem collapses on the strong-convexity step, so the paper needs a corrected proof or honest reframing before it can be believed. read the letter →

arxiv 2501.08195 v1 pith:UJCQP6U3 submitted 2025-01-14 cs.CV cs.LG

Shuo Li , Mehrdad Yaghoobi This is my paper · ORCID
classification cs.CVcs.LG
keywords hyperspectralimageinpaintingdeeppriorplug-and-playpriorslow-rankandsparsemodelfixed-pointconvergenceLyapunovstabilityself-supervisedlearning1-Lipschitzconstraint
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 is trying to establish that hyperspectral image inpainting, filling in pixels or entire spectral bands that are missing from a noisy observation, can be carried out by an iterative algorithm that is both self-supervised and provably convergent. The proposed LRS-PnP-DIP(1-Lip) algorithm keeps the joint low-rank and sparsity model of an earlier design but replaces the singular-value-thresholding low-rank step with a deep image prior (DIP), a network fitted to the corrupted image itself, and replaces the sparse-coding denoiser with a non-expansive averaged non-local-means denoiser. The central result is Theorem 1, a Lyapunov-function proof that, under a $\theta$-averaged linear denoiser and an $L$-Lipschitz DIP with $L \leq 1$, the iterates converge to a fixed point. If the theorem is right, it would matter because DIP-based inpainting has previously been unstable without early stopping or hand-tuned learning rates, and the paper shows that its constrained variant removes most of that instability while staying competitive with or ahead of the compared learning-based baselines.

What carries the argument

The load-bearing object is the Lyapunov function $H_k$ together with the two non-expansiveness assumptions that make it decrease. The operator $T$ in the sparse-coding step is assumed $\theta$-averaged, meaning it is a convex combination of the identity and a non-expansive operator; the averaged NLM denoiser satisfies this by construction. The deep image prior $f_\theta$ is constrained to be $L$-Lipschitz with $L \leq 1$ by projecting each layer's spectral norm to one, and skip connections are removed so the composition of 1-Lipschitz layers stays 1-Lipschitz. The ADMM updates give a closed-form $x$ update because the masks and patch operators are diagonal, so the matrix inversion in Eq. (15) reduces to element-wise division; this makes the certified iteration practical. What this machinery does is convert the question "does the iteration stabilize?" into checking a single scalar energy that is non-increasing.

What would settle it

Run Algorithm 3 on a masked Chikusei or Indian Pines patch and record the sequence $H_k$; if $H_k$ ever increases, the claimed monotone decrease is false. Independently, compute the smallest eigenvalue of $2M^TM$ for the binary mask: with any missing pixel it is exactly zero, which contradicts the strong-convexity inequality the proof uses at Eq. (A25).

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that two modest design changes turn an empirically successful but unstable iterative inpainting algorithm into one with a convergence certificate. The changes are to use an averaged NLM denoiser, whose doubly stochastic weight matrix has spectral norm at most one, and to enforce a 1-Lipschitz constraint on every layer of the deep image prior by spectral-norm projection while removing skip connections. Theorem 1 then defines the Lyapunov function $H_k = 2\|x^k - x^*\|^2 + \frac{1}{\mu^2}\|\lambda_1^k - \lambda_1^*\|^2 + \frac{1}{\mu^2}\|\lambda_2^k - \lambda_2^*\|^2$ and shows that it is non-increasing, which yields $\|x^k - x^*\| \to 0$, $\|\alpha^k - \alpha^*\| \to 0$, and $\|u^k - u^*\| \to 0$ as $k \to \infty$. The paper further argues, with experiments, that the DIP captures the low-rank subspace of hyperspectral data more faithfully than singular value thresholding because it preserves small singular values, and that the resulting inpaintings compare favorably with several learning-based baselines on masked real datasets.

Load-bearing premise

The load-bearing premise is that the data-fidelity term $\|y - Mx\|^2$ is strongly convex, so that the inequality at Eq. (A25) holds, and that the DIP acts as a fixed 1-Lipschitz operator while the iteration runs; for a binary mask the Hessian $2M^TM$ has zero directions on missing pixels and Algorithm 3 keeps updating the network weights, so both premises are strained.

Editorial extensions

If this is right

  • If the theorem is correct, hyperspectral inpainting can be deployed with a certified fixed-point guarantee rather than relying on early stopping to avoid DIP divergence.
  • The 1-Lipschitz DIP and averaged NLM denoiser provide concrete, implementable sufficient conditions; the experiments indicate they cost only a small MPSNR reduction relative to the unconstrained version.
  • Replacing singular value thresholding with a DIP means small but important singular values are preserved, so the low-rank prior becomes data-adaptive and can represent structures a fixed subspace model would discard.
  • Because the whole pipeline is self-supervised and the dictionary is learned from the corrupted image itself, the method is usable when no clean reference or external training set is available, which is the typical onboard remote-sensing setting.

Reading between the lines

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

  • Our inference: the same Lyapunov-style certificate would extend to other inverse problems only when the data-fidelity term is strongly convex; for pure masking $2M^TM$ is singular, so a proximal or regularized fidelity term would be needed for the proof to survive.
  • Our inference: the layerwise product bound on the Lipschitz constant is not tight, so a joint or layer-grouped spectral-norm analysis could recover some expressivity lost to the current constraint.
  • Our inference: the design recipe of a doubly stochastic denoiser plus a spectral-normalized network without skip connections offers a testable template for stabilizing other iterative deep-prior schemes in video, multi-modal, or medical imaging.
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

4 major / 5 minor

Summary. The paper proposes LRS-PnP-DIP(1-Lip), a self-supervised hyperspectral inpainting algorithm that combines a low-rank and sparse model with a plug-and-play (PnP) denoiser and a deep image prior (DIP) whose layers are constrained to be 1-Lipschitz. The main theoretical claim is Theorem 1, which states that, under assumptions of a linear θ-averaged denoiser and an L-Lipschitz DIP with L ≤ 1, the trajectories of the algorithm converge to a fixed point in the Lyapunov sense. The paper also presents experimental comparisons on the Chikusei and Indian Pines datasets, reporting competitive or superior MPSNR/MSSIM/MSAM values against several learning-based baselines.

Significance. The convergence guarantee, if valid, would be a useful contribution to the PnP and DIP literature, as it would provide the first stability analysis for a DIP-based low-rank replacement in an iterative solver. However, the proof of Theorem 1 contains a load-bearing technical error: it relies on strong convexity of f(x)=||y-Mx||^2, which does not hold for the inpainting problem with a binary mask. Consequently, the central theoretical contribution is not supported. The experimental results are reasonably extensive and the code is made available, but the empirical findings do not compensate for the invalidity of the main theorem.

major comments (4)
  1. [Appendix A.2, Eq. (A25)] The proof applies Lemma 2 to f(x)=||y-Mx||^2 and uses the resulting inequality <∇f(x^{k+1})-∇f(x*), x^{k+1}-x*> ≥ ρ||x^{k+1}-x*||^2. For the inpainting problem, M is a diagonal binary mask with zero entries on missing pixels, so the Hessian 2M^T M is singular. Taking v supported only on missing pixels gives Mv=0 and hence <∇f(x+v)-∇f(x), v>=2||Mv||^2=0, which contradicts strong convexity with any modulus ρ>0. Therefore inequality (A25), and everything derived from it including (A26) and the Lyapunov decrease in (A32)-(A34), is unsupported. This is not a minor gap; it invalidates the proof of Theorem 1.
  2. [Assumption 2 and Algorithm 3] Assumption 2 treats the DIP f_θ as a fixed L-Lipschitz operator with L ≤ 1, while Algorithm 3 updates the parameters θ with the Adam optimizer at every iteration. The proof does not establish that the trained, time-varying network satisfies the Lipschitz bound uniformly across iterations, nor that the convergence argument extends to the implemented algorithm. This is a second independent mismatch between the theoretical object and the algorithm whose convergence is claimed.
  3. [Appendix A.2, Eqs. (A17)-(A20) and Eq. (17)] The proof assumes a constant penalty parameter µ in the updates (A17)-(A20) and in the definition of H_k, using 1/µ^2 factors and factoring µ out of inner products. Algorithm 3, however, updates µ1 and µ2 multiplicatively as µ^{k+1}_i = ρ_i µ^k_i in Eq. (17). The theorem statement mentions 'with penalty µ' but does not specify how this relates to the growing penalties of the algorithm. As written, the proof does not cover the algorithm with adaptive penalty parameters.
  4. [Appendix A.2, Eqs. (A32)-(A34)] Even under the disputed strong-convexity assumption, the derivation of H_k - H_{k+1} ≥ 0 is incomplete. The authors claim that the last two inner-product terms in (A32) are non-negative by plugging k = k-1 into the first line of (A29), but that relation involves x^{k+1}_e, not x^k_e, so the inference is invalid. The conclusion that the right-hand side of (A32) is non-negative is therefore not justified, and the Lyapunov decrease does not follow from the stated inequalities.
minor comments (5)
  1. [Appendix A.1.2, Eq. (A5)] There is a typographical error: 'f2(x1 - f2(x2))' should read 'f2(x1) - f2(x2)'.
  2. [Table 1, Mask Type 3, MSSIM for DeepHyIn] The entry '08950' appears to be missing a decimal point; it should likely be '0.8950'.
  3. [Abstract and Section 1] There is an extra space before the comma in 'under mild assumptions ,' and similar minor punctuation issues elsewhere.
  4. [Appendix A.1.2 and Appendix A.3.5] The paper states in Appendix A.1.2 that skip connections are removed from the 1-Lipschitz DIP to keep the Lipschitz constant bounded by 1, but Appendix A.3.5 reports that Skip-Net 2D achieves the best performance and is selected as the backbone for LRS-PnP-DIP(1-Lip) and all competing methods. It should be clarified which architecture is actually used in the experiments.
  5. [Figure 3 caption] The caption says 'Training is conducted with a single-band HS image,' which is confusing because the text describes varying the number of input channels. Please reword to describe the experiment accurately.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Theorem 1 is a conditional convergence proof built on external lemmas, and the design choices that satisfy its assumptions are engineering constraints, not circular reductions.

full rationale

The paper's derivation chain is not circular. The central claim, Theorem 1, is a conditional convergence theorem: if the denoiser is linear and θ-averaged and the DIP is L-Lipschitz with L ≤ 1, then the Lyapunov function H_k is non-increasing and the iterates converge. The assumptions are not defined in terms of the conclusion; Assumptions 1 and 2 state properties of the denoiser and DIP, and the proof derives the decrease using external lemmas (Lemma 2 from Ekeland–Temam, Lemmas 1 and 3 from Nair et al.), rather than assuming the decrease. The design of LRS-PnP-DIP(1-Lip) with an averaged NLM denoiser and a spectrally normalized 1-Lipschitz DIP ensures the hypotheses hold by construction, but that is an architectural choice, not a logical equivalence between input and output. Experimental results are compared against independent baselines, and no fitted parameter is relabeled as a prediction. Self-citations to the authors' preliminary report [20] describe the predecessor algorithm and the reported instability, but they are not used as the justification for the new convergence proof; the instability is also attributed to external DIP-related works [13,21–23]. The reader's concern about strong convexity of f(x)=‖y–Mx‖^2 is a correctness gap, because M is a binary mask with singular M^T M, so Lemma 2 does not apply; similarly, the constant-µ proof does not match the geometric penalty growth in Algorithm 3. These are validity mismatches, not circular reductions. No step of the derivation reduces to its own input by definition.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The convergence theorem depends on assumptions that either fail, such as strong convexity of the masked fidelity term, or are not verified in the implemented algorithm, such as a fixed DIP operator, a linear averaged NLM denoiser, and a constant penalty parameter. The remaining hyperparameters are tuned on the test datasets, so the empirical claim of state-of-the-art is not independent of manual selection.

free parameters (5)
  • wlr and ws (low-rank and sparsity weights) = 1 (set via sensitivity analysis)
    Table A2 shows performance varies with the ratio; values are chosen by hand on the test datasets.
  • gamma (data fidelity weight) = 0.5
    Set manually and described as dependent on noise level; no principled fitting rule is given.
  • mu1 and mu2 (ADMM penalty parameters) = initial 1/2, multiplied by rho1/rho2 per iteration
    Hand-chosen; the convergence proof uses a fixed mu and does not cover the growth schedule.
  • rho1 and rho2 (penalty multipliers) = not reported numerically
    Appear in Eq. (17) and Algorithm 3 but are not included in the theorem or appendix proof.
  • DIP learning rate and early stopping window/patience = 0.1; window 20, patience 100
    Manual choices that control the effective DIP operator and reconstructed quality.
assumptions (6)
  • domain assumption The observation Y = M{X} + N with known binary mask and Gaussian noise holds.
    Standard model in Eq. (1); no verification on the real test images beyond the experimental setup.
  • ad hoc to paper The data fidelity term ||y - Mx||^2 is strongly convex.
    Used implicitly in Eq. (A25). For masked inpainting, M has zero diagonal entries, so the Hessian is singular and strong convexity fails.
  • domain assumption A stationary point (x*, alpha*, u*, lambda1*, lambda2*) exists.
    The Lyapunov proof requires such a point; no existence argument is supplied.
  • ad hoc to paper The DIP is a fixed 1-Lipschitz operator during the convergence analysis.
    Algorithm 3 updates theta by Adam each iteration; the proof treats u^{k+1} = f_theta(...) as if theta is constant.
  • domain assumption The averaged NLM denoiser is linear and theta-averaged.
    Assumption 1. Practical NLM weights depend on the image, so linearity is not established.
  • domain assumption The online-learned dictionary Phi captures the spectral subspace well.
    Learned from noisy, incomplete pixels; the sparse coding convergence depends on this via Lemma 3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self-supervised Deep Hyperspectral Inpainting with the Plug and Play and Deep Image Prior Models." pith.science (2026). https://pith.science/paper/UJCQP6U3

@misc{pith2026250108195,
  author       = {Pith},
  title        = {Pith review of: Self-supervised Deep Hyperspectral Inpainting with the Plug and Play and Deep Image Prior Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UJCQP6U3}},
  note         = {Machine review of arXiv:2501.08195}
}
read the original abstract

Hyperspectral images are typically composed of hundreds of narrow and contiguous spectral bands, each containing information regarding the material composition of the imaged scene. However, these images can be affected by various sources of noise, distortions, or data loss, which can significantly degrade their quality and usefulness. This paper introduces a convergent guaranteed algorithm, LRS-PnP-DIP(1-Lip), which successfully addresses the instability issue of DHP that has been reported before. The proposed algorithm extends the successful joint low-rank and sparse model to further exploit the underlying data structures beyond the conventional and sometimes restrictive unions of subspace models. A stability analysis guarantees the convergence of the proposed algorithm under mild assumptions , which is crucial for its application in real-world scenarios. Extensive experiments demonstrate that the proposed solution consistently delivers visually and quantitatively superior inpainting results, establishing state-of-the-art performance.

Figures

Figures reproduced from arXiv: 2501.08195 by the authors.

Figure 1
Figure 1. Flow chart of the proposed LRS-PnP-DIP(1-Lip) algorithm. At each iteration, α k+1 u k+1 and x k+1 are sequentially updated. The 1-Lipschitz DIP is implemented by imposing Lipschitz constraints on all layers. We use red color to highlight the differences between this work and the LRS-PnP-DIP algorithm [20]. The detailed design of the 1-Lipschitz DIP is placed in Appendix A.1. Algorithm 3 LRS-PnP-DIP(1-Lip) algorithm.… view at source ↗
Figure 2
Figure 2. Comparison of MPSNR value of LRS-PnP among different τ (τ = ws/wlr) under different masks. 4.3. Low Rankness Due to DIP In the original LRS-PnP-DIP Algorithm [20], we propose solving the low-rank mini￾mization problem using DIP instead of the traditional SVT projection. A natural question arises: does the DIP (e.g., the architecture proposed in [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Learning capability of DIP vs. number of input and output channels. Training is conducted with a single-band HS image, meaning that the input HSIs are processed; there is no correlation in the spectral domain. There is a significant performance gain when there are more input bands, indicating that the DIP with 2D convolution has the ability to exploit the correlation between channels [PITH_FULL_IMAGE:figures/full_f… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The amplitude of the singular value of the reconstructed image upon converge. The important singular values are captured and preserved via the 2D-convolution DIP, which is even more accurate than the traditional SVT projection. In [PITH_FULL_IMAGE:figures/full_fig_p01…
Figure 5
Figure 5. Figure 5: Empirical converge of LRS-PnP-DIP(1-Lip) with modified NLM denoiser and non￾expansive/1-Lipschitz DIP. Top left, top right, and bottom left: successive difference of x, λ1 , and λ2 in the log scale, respectively. Bottom right: the inpainting MPSNR vs. the number of ite…
Figure 6
Figure 6. Figure 6: Different algorithms and their recovered spectrum of the center pixel with the assumption that the whole spectrum bands are missing. MPSNR increases from the thin line to the thick line. In [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Comparison between the proposed algorithm and other learning-based inpainting algo￾rithms on the Chikusei dataset. From top to bottom: (1) clean image, (2) input image, (3) GLON, (4) DHP, (5) R-DLRHyIn, (6) DeepRED, (7) DeepHyIn, (8) PnP-DIP, (9) DDS2M, and (10) LRS-Pn…
Figure 8
Figure 8. Figure 8: Comparison between the proposed algorithm and other learning-based inpainting al￾gorithms on the Indian Pines dataset. From top to bottom: (1) clean image, (2) input image, (3) GLON, (4) DHP, (5) R-DLRHyIn, (6) DeepRED, (7) DeepHyIn, (8) PnP-DIP, (9) DDS2M, and (10) LR…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 53 canonical work pages

  1. [1]

    Hyperspectral push-broom microscope development and characterization

    Ortega, S.; Guerra, R.; Diaz, M.; Fabelo, H.; López, S.; Callico, G.M.; Sarmiento, R. Hyperspectral push-broom microscope development and characterization. IEEE Access 2019, 7, 122473–122491

    work page 2019

  2. [2]

    EMIT L1B At-Sensor Calibrated Radiance and Geolocation Data 60 m V001. Available online: https://search.earthdata.nasa.gov/ search/granules?p=C2408009906-LPCLOUD&pg[0][v]=f&pg[0][gsk]=-start_date&q=%22EMIT%22&tl=1711560236!3!! (accessed on 27 August 2024)

    work page 2024

  3. [3]

    Fast hyperspectral image denoising and inpainting based on low-rank and sparse representations

    Zhuang, L.; Bioucas-Dias, J.M. Fast hyperspectral image denoising and inpainting based on low-rank and sparse representations. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2018, 11, 730–742

    work page 2018

  4. [4]

    Tensor completion via complementary global, local, and nonlocal priors

    Zhao, X.L.; Yang, J.H.; Ma, T.H.; Jiang, T.X.; Ng, M.K.; Huang, T.Z. Tensor completion via complementary global, local, and nonlocal priors. IEEE Trans. Image Process. 2021, 31, 984–999

    work page 2021

  5. [5]

    Tensor completion via collaborative sparse and low-rank transforms

    Li, B.Z.; Zhao, X.L.; Wang, J.L.; Chen, Y.; Jiang, T.X.; Liu, J. Tensor completion via collaborative sparse and low-rank transforms. IEEE Trans. Comput. Imaging 2021, 7, 1289–1303

    work page 2021

  6. [6]

    Self-supervised nonlinear transform-based tensor nuclear norm for multi-dimensional image recovery

    Luo, Y.S.; Zhao, X.L.; Jiang, T.X.; Chang, Y.; Ng, M.K.; Li, C. Self-supervised nonlinear transform-based tensor nuclear norm for multi-dimensional image recovery. IEEE Trans. Image Process. 2022, 31, 3793–3808

    work page 2022

  7. [7]

    MPCT: Multiscale point cloud transformer with a residual network

    Wu, Y.; Liu, J.; Gong, M.; Liu, Z.; Miao, Q.; Ma, W. MPCT: Multiscale point cloud transformer with a residual network. IEEE Trans. Multimed. 2023, 26, 3505–3516

    work page 2023

  8. [8]

    EGST: Enhanced geometric structure transformer for point cloud registration

    Yuan, Y.; Wu, Y.; Fan, X.; Gong, M.; Ma, W.; Miao, Q. EGST: Enhanced geometric structure transformer for point cloud registration. IEEE Trans. Vis. Comput. Graph. 2023, 30, 6222–6234

    work page 2023

Show all 57 references
  1. [9]

    HSI-IPNet: Hyperspectral imagery inpainting by deep learning with adaptive spectral extraction

    Wong, R.; Zhang, Z.; Wang, Y.; Chen, F.; Zeng, D. HSI-IPNet: Hyperspectral imagery inpainting by deep learning with adaptive spectral extraction. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2020, 13, 4369–4380

    2020

  2. [10]

    Deep image prior

    Ulyanov, D.; Vedaldi, A.; Lempitsky, V . Deep image prior. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18–22 June 2018; pp. 9446–9454

    2018

  3. [11]

    Deep hyperspectral prior: Single-image denoising, inpainting, super-resolution

    Sidorov, O.; Yngve Hardeberg, J. Deep hyperspectral prior: Single-image denoising, inpainting, super-resolution. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, Seoul, Republic of Korea, 27 October–2 November 2019

    2019

  4. [12]

    Self-supervised neural networks for spectral snapshot compressive imaging

    Meng, Z.; Yu, Z.; Xu, K.; Yuan, X. Self-supervised neural networks for spectral snapshot compressive imaging. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 10–17 October 2021; pp. 2622–2631. Remote Sens. 2025, 1, 0 30 of 31

    2021

  5. [13]

    Deep Plug-and-Play Prior for Hyperspectral Image Restoration

    Lai, Z.; Wei, K.; Fu, Y. Deep Plug-and-Play Prior for Hyperspectral Image Restoration. Neurocomputing 2022, 481, 281–293

    2022

  6. [14]

    Adaptive Deep PnP Algorithm for Video Snapshot Compressive Imaging

    Wu, Z.; Yang, C.; Su, X.; Yuan, X. Adaptive Deep PnP Algorithm for Video Snapshot Compressive Imaging. arXiv 2022, arXiv:2201.05483

    2022 arXiv

  7. [15]

    Image reconstruction for mri using deep cnn priors trained without groundtruth

    Gan, W.; Eldeniz, C.; Liu, J.; Chen, S.; An, H.; Kamilov, U.S. Image reconstruction for mri using deep cnn priors trained without groundtruth. In Proceedings of the 2020 54th Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, 1–4 November 2020; pp. 475–479

    2020

  8. [16]

    Robust Hyperspectral Inpainting via Low-Rank Regularized Untrained Convolutional Neural Network

    Niresi, K.F.; Chi, C.Y. Robust Hyperspectral Inpainting via Low-Rank Regularized Untrained Convolutional Neural Network. IEEE Geosci. Remote Sens. Lett. 2023, 20, 1–5

    2023

  9. [17]

    Decoupled-and-coupled networks: Self-supervised hyperspectral image super-resolution with subpixel fusion

    Hong, D.; Yao, J.; Li, C.; Meng, D.; Yokoya, N.; Chanussot, J. Decoupled-and-coupled networks: Self-supervised hyperspectral image super-resolution with subpixel fusion. IEEE Trans. Geosci. Remote Sens. 2023, 61, 5527812

    2023

  10. [18]

    Cross-attention in coupled unmixing nets for unsupervised hyperspectral super-resolution

    Yao, J.; Hong, D.; Chanussot, J.; Meng, D.; Zhu, X.; Xu, Z. Cross-attention in coupled unmixing nets for unsupervised hyperspectral super-resolution. In Proceedings of the Computer Vision—ECCV 2020: 16th European Conference, Glasgow, UK, 23–28 August 2020; Part XXIX 16; pp. 208–224

    2020

  11. [19]

    Dds2m: Self-supervised denoising diffusion spatio-spectral model for hyperspectral image restoration

    Miao, Y.; Zhang, L.; Zhang, L.; Tao, D. Dds2m: Self-supervised denoising diffusion spatio-spectral model for hyperspectral image restoration. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Paris, France, 2–6 October 2023; pp. 12086–12096

    2023

  12. [20]

    Self-Supervised Hyperspectral Inpainting with the Optimisation inspired Deep Neural Network Prior

    Li, S.; Yaghoobi, M. Self-Supervised Hyperspectral Inpainting with the Optimisation inspired Deep Neural Network Prior. In Proceedings of the 2023 31st European Signal Processing Conference (EUSIPCO), Helsinki, Finland, 4–8 September 2023; pp. 471–475

    2023

  13. [21]

    Deep decoder: Concise image representations from untrained non-convolutional networks

    Heckel, R.; Hand, P . Deep decoder: Concise image representations from untrained non-convolutional networks. arXiv preprint arXiv:1810.03982 2018

    2018 arXiv

  14. [22]

    DeepRED: Deep image prior powered by RED

    Mataev, G.; Milanfar, P .; Elad, M. DeepRED: Deep image prior powered by RED. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops, Seoul, Republic of Korea, 27 October–2 November 2019

    2019

  15. [23]

    Image restoration using total variation regularized deep image prior

    Liu, J.; Sun, Y.; Xu, X.; Kamilov, U.S. Image restoration using total variation regularized deep image prior. In Proceedings of the ICASSP 2019—2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Brighton, UK, 12–17 May 2019; pp. 7715–7719

    2019

  16. [24]

    A plug-and-play deep image prior

    Sun, Z.; Latorre, F.; Sanchez, T.; Cevher, V . A plug-and-play deep image prior. In Proceedings of the ICASSP 2021—2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Toronto, Canada, 6–11 June 2021; pp. 8103–8107

    2021

  17. [25]

    Hyperspectral image classification using dictionary-based sparse representation

    Chen, Y.; Nasrabadi, N.M.; Tran, T.D. Hyperspectral image classification using dictionary-based sparse representation. IEEE Trans. Geosci. Remote Sens. 2011, 49, 3973–3985

    2011

  18. [26]

    Hyperspectral image restoration using low-rank matrix recovery

    Zhang, H.; He, W.; Zhang, L.; Shen, H.; Yuan, Q. Hyperspectral image restoration using low-rank matrix recovery. IEEE Trans. Geosci. Remote Sens. 2013, 52, 4729–4743

    2013

  19. [27]

    LRR-Net: An Interpretable Deep Unfolding Network for Hyperspectral Anomaly Detection

    Li, C.; Zhang, B.; Hong, D.; Yao, J.; Chanussot, J. LRR-Net: An Interpretable Deep Unfolding Network for Hyperspectral Anomaly Detection. IEEE Trans. Geosci. Remote Sens. 2023, 61, 1–12

    2023

  20. [28]

    Online dictionary learning for sparse coding

    Mairal, J.; Bach, F.; Ponce, J.; Sapiro, G. Online dictionary learning for sparse coding. In Proceedings of the 26th Annual International Conference on Machine Learning, Montreal, Canada, 14–18 June 2009; pp. 689–696

    2009

  21. [29]

    Self-paced joint sparse representation for the classification of hyperspectral images

    Peng, J.; Sun, W.; Du, Q. Self-paced joint sparse representation for the classification of hyperspectral images. IEEE Trans. Geosci. Remote Sens. 2018, 57, 1183–1194

    2018

  22. [30]

    Hyperspectral image denoising via sparse representation and low-rank constraint

    Zhao, Y.Q.; Yang, J. Hyperspectral image denoising via sparse representation and low-rank constraint. IEEE Trans. Geosci. Remote Sens. 2014, 53, 296–308

    2014

  23. [31]

    Sparse unmixing of hyperspectral data

    Iordache, M.D.; Bioucas-Dias, J.M.; Plaza, A. Sparse unmixing of hyperspectral data. IEEE Trans. Geosci. Remote Sens. 2011, 49, 2014–2039

    2011

  24. [32]

    Joint spatial and spectral low-rank regularization for hyperspectral image denoising

    Xue, J.; Zhao, Y.; Liao, W.; Kong, S.G. Joint spatial and spectral low-rank regularization for hyperspectral image denoising. IEEE Trans. Geosci. Remote Sens. 2017, 56, 1940–1958

    2017

  25. [33]

    Local low-rank and sparse representation for hyperspectral image denoising

    Ma, G.; Huang, T.Z.; Huang, J.; Zheng, C.C. Local low-rank and sparse representation for hyperspectral image denoising. IEEE Access 2019, 7, 79850–79865

    2019

  26. [34]

    Distributed optimization and statistical learning via the alternating direction method of multipliers

    Boyd, S.; Parikh, N.; Chu, E.; Peleato, B.; Eckstein, J. Distributed optimization and statistical learning via the alternating direction method of multipliers. Found. Trends® Mach. Learn. 2011, 3, 1–122

    2011

  27. [35]

    Proximal algorithms

    Parikh, N.; Boyd, S.; et al. Proximal algorithms. Found. Trends® Optim. 2014, 1, 127–239

    2014

  28. [36]

    Plug-and-play priors for model based reconstruction

    Venkatakrishnan, S.V .; Bouman, C.A.; Wohlberg, B. Plug-and-play priors for model based reconstruction. In Proceedings of the 2013 IEEE Global Conference on Signal and Information Processing, Austin, TX, USA, 3–5 December 2013; pp. 945–948

    2013

  29. [37]

    An iterative thresholding algorithm for linear inverse problems with a sparsity constraint

    Daubechies, I.; Defrise, M.; De Mol, C. An iterative thresholding algorithm for linear inverse problems with a sparsity constraint. Commun. Pure Appl. Math. J. Issued Courant Inst. Math. Sci. 2004, 57, 1413–1457

    2004

  30. [38]

    A singular value thresholding algorithm for matrix completion

    Cai, J.F.; Candès, E.J.; Shen, Z. A singular value thresholding algorithm for matrix completion. SIAM J. Optim. 2010, 20, 1956–1982

    2010

  31. [39]

    Plug-and-play priors for bright field electron tomography and sparse interpolation

    Sreehari, S.; Venkatakrishnan, S.V .; Wohlberg, B.; Buzzard, G.T.; Drummy, L.F.; Simmons, J.P .; Bouman, C.A. Plug-and-play priors for bright field electron tomography and sparse interpolation. IEEE Trans. Comput. Imaging 2016, 2, 408–423

    2016

  32. [40]

    Fixed-point and objective convergence of plug-and-play algorithms

    Nair, P .; Gavaskar, R.G.; Chaudhury, K.N. Fixed-point and objective convergence of plug-and-play algorithms. IEEE Trans. Comput. Imaging 2021, 7, 337–348

    2021

  33. [41]

    Convex Analysis and Variational Problems; SIAM: Philadelphia, PA, USA, 1999

    Ekeland, I.; Temam, R. Convex Analysis and Variational Problems; SIAM: Philadelphia, PA, USA, 1999. Remote Sens. 2025, 1, 0 31 of 31

    1999

  34. [42]

    Plug-and-play methods provably converge with properly trained denoisers

    Ryu, E.; Liu, J.; Wang, S.; Chen, X.; Wang, Z.; Yin, W. Plug-and-play methods provably converge with properly trained denoisers. In Proceedings of the International Conference on Machine Learning, Long Beach, CA, USA, 9–15 June 2019; pp. 5546–5557

    2019

  35. [43]

    Airborne Hyperspectral Data over Chikusei; SAL-2016-05-27; University of Tokyo: Tokyo, Japan, 2016

    Yokoya, N.; Iwasaki, A. Airborne Hyperspectral Data over Chikusei; SAL-2016-05-27; University of Tokyo: Tokyo, Japan, 2016

    2016

  36. [44]

    220 band AVIRIS hyperspectral image data set: June 12, 1992 Indian Pine test site 3

    Baumgardner, M.F.; Biehl, L.L.; Landgrebe, D.A. 220 band AVIRIS hyperspectral image data set: June 12, 1992 Indian Pine test site 3. Purdue Univ. Res. Repos. 2015, 10, 991

    1992

  37. [45]

    Orbital edge computing: Nanosatellite constellations as a new class of computer system

    Denby, B.; Lucia, B. Orbital edge computing: Nanosatellite constellations as a new class of computer system. In Proceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems, Lausanne, Switzerland, 16–20 March...

    2020

  38. [46]

    Regularisation of neural networks by enforcing lipschitz continuity

    Gouk, H.; Frank, E.; Pfahringer, B.; Cree, M.J. Regularisation of neural networks by enforcing lipschitz continuity. Mach. Learn. 2021, 110, 393–416

    2021

  39. [47]

    Early Stopping for Deep Image Prior

    Wang, H.; Li, T.; Zhuang, Z.; Chen, T.; Liang, H.; Sun, J. Early Stopping for Deep Image Prior. arXiv 2021, arXiv:2112.06074

    2021 arXiv

  40. [48]

    Deep sparse and low-rank prior for hyperspectral image denoising

    Nguyen, H.V .; Ulfarsson, M.O.; Sigurdsson, J.; Sveinsson, J.R. Deep sparse and low-rank prior for hyperspectral image denoising. In Proceedings of the IGARSS 2022–2022 IEEE International Geoscience and Remote Sensing Symposium, Kuala Lumpur, Malaysia, 17–22 July 2022; pp. 1217–1220

    2022

  41. [49]

    Spectral normalization for generative adversarial networks

    Miyato, T.; Kataoka, T.; Koyama, M.; Yoshida, Y. Spectral normalization for generative adversarial networks. arXiv 2018, arXiv:1802.05957

    2018 arXiv

  42. [50]

    Unsupervised deep hyperspectral inpainting using a new mixing model

    Rasti, B.; Ghamisi, P .; Gloaguen, R. Unsupervised deep hyperspectral inpainting using a new mixing model. In Proceedings of the IGARSS 2022–2022 IEEE International Geoscience and Remote Sensing Symposium, Kuala Lumpur, Malaysia, 17–22 July 2022; pp. 1221–1224

    2022

  43. [51]

    A single model CNN for hyperspectral image denoising

    Maffei, A.; Haut, J.M.; Paoletti, M.E.; Plaza, J.; Bruzzone, L.; Plaza, A. A single model CNN for hyperspectral image denoising. IEEE Trans. Geosci. Remote Sens. 2019, 58, 2516–2529

    2019

  44. [52]

    ADMM DIP-TV: Combining Total Variation and Deep Image Prior for image restoration

    Cascarano, P .; Sebastiani, A.; Comes, M.C. ADMM DIP-TV: Combining Total Variation and Deep Image Prior for image restoration. arXiv 2020, arXiv:2009.11380

    2020 arXiv

  45. [53]

    Transformer meets remote sensing video detection and tracking: A comprehensive survey

    Jiao, L.; Zhang, X.; Liu, X.; Liu, F.; Yang, S.; Ma, W.; Li, L.; Chen, P .; Feng, Z.; Guo, Y.; et al. Transformer meets remote sensing video detection and tracking: A comprehensive survey. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2023, 16, 1–45

    2023

  46. [54]

    Lyapunov stability theory of nonsmooth systems

    Shevitz, D.; Paden, B. Lyapunov stability theory of nonsmooth systems. IEEE Transactions on automatic control 1994, 39, 1910–1914

    1994

  47. [55]

    Lyapunov theory for discrete time systems

    Bof, N.; Carli, R.; Schenato, L. Lyapunov theory for discrete time systems. arXiv 2018, arXiv:1809.05289

    2018 arXiv

  48. [56]

    The stability of nonlinear dissipative systems

    Hill, D.; Moylan, P . The stability of nonlinear dissipative systems. IEEE Trans. Autom. Control 1976, 21, 708–711

    1976

  49. [57]

    A fundamental proof of convergence of alternating direction method of multipliers for weakly convex optimization

    Zhang, T.; Shen, Z. A fundamental proof of convergence of alternating direction method of multipliers for weakly convex optimization. J. Inequalities Appl. 2019, 2019, 1–21. Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely...

    2019

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.