pith. machine review for the scientific record. sign in

arxiv: 2604.19084 · v1 · submitted 2026-04-21 · 📡 eess.SP

Recognition: unknown

DUSG-Tomo-Net: A Deep Unfolded Neural Network for Super-Resolving Gridless Spaceborne SAR Tomography via Learned Toeplitz-Structured Covariance Representation

Kun Qian , Zhuge Xia , Qian Ma , Qi Zhang , Weijian Liu , Xiufeng He
Authors on Pith no claims yet

Pith reviewed 2026-05-10 02:27 UTC · model grok-4.3

classification 📡 eess.SP
keywords spaceborne SAR tomographygridless reconstructiondeep unfoldingToeplitz covariancesuper-resolutionnonuniform baselinessingle-look approximationcovariance estimation
0
0 comments X

The pith

A deep unfolded network recovers gridless super-resolved elevations in spaceborne SAR tomography by learning a Toeplitz-structured covariance from nonuniform baselines.

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

The paper proposes DUSG-Tomo-Net to solve the super-resolution inversion problem in spaceborne SAR tomography without relying on a discrete elevation grid. It achieves this by reformulating the problem in a lag domain using a structured single-look approximation that makes the covariance Toeplitz-compatible even with nonuniform baselines. The network then applies learned regularization across layers and projects onto the set of Hermitian Toeplitz positive-semidefinite matrices to recover the covariance, while a fixed operator incorporates the actual acquisition geometry into the data consistency step. Elevations are finally estimated continuously using a spectral estimator. If successful, this would provide more accurate 3D imaging of urban scenes from spaceborne SAR under realistic acquisition conditions and limited data.

Core claim

The central discovery is that reformulating the TomoSAR inversion in a Toeplitz-compatible lag domain via structured single-look approximation allows a deep unfolded network to recover a Hermitian Toeplitz positive-semidefinite covariance representation through layerwise learned regularization and projection, with geometry embedded via fixed operator, enabling gridless super-resolution for nonuniform baseline spaceborne acquisitions.

What carries the argument

The learned layerwise regularization and projection-based structural enforcement of the Toeplitz covariance, combined with the fixed signal-independent data-consistency operator that embeds the acquisition geometry.

If this is right

  • Scatterer elevations can be estimated in the continuous domain without discretization bias.
  • The method adapts to different baseline configurations without retraining the geometry part.
  • Robust inversion is possible with fewer observations and under low signal-to-noise ratios.
  • Off-grid effects common in compressive sensing based TomoSAR are avoided.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach might inspire similar deep unfolding techniques for other array processing tasks involving covariance estimation under irregular sampling.
  • If the single-look approximation proves broadly applicable, it could simplify gridless methods in other nonuniform array geometries beyond SAR.
  • Extensions to multi-look or multi-polarimetric data could further enhance the framework's utility in practical remote sensing.

Load-bearing premise

The structured single-look approximation holds sufficiently well to allow a valid Toeplitz reformulation despite the nonuniform baselines in spaceborne TomoSAR.

What would settle it

Comparison on simulated data with known sub-Rayleigh separated scatterers under nonuniform baselines and single-look conditions, checking if the method resolves them accurately while grid-based methods fail, would test the claim; failure to outperform or resolve would falsify the advantage.

Figures

Figures reproduced from arXiv: 2604.19084 by Kun Qian, Qian Ma, Qi Zhang, Weijian Liu, Xiufeng He, Zhuge Xia.

Figure 1
Figure 1. Figure 1: SAR imaging geometry. The elevation synthetic aperture is built up by SAR data acquired [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Representative normalized elevation spectra for the uniform-baseline spectral-reconstruction [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effective detection rate Pd versus normalized separation αd = ds/ρs for the super-resolution experiment under uniform baselines. The vertical dotted line marks the Rayleigh limit αd = 1. respectively, which is markedly earlier than ANM-ADMM and TADCG. At the Rayleigh limit αd = 1, DUSG-Tomo-Net achieves Pd = 0.76, 0.92, and 0.89 at 0, 6, and 10 dB, respectively, whereas the corresponding values for TADCG a… view at source ↗
Figure 4
Figure 4. Figure 4: Normalized elevation estimation accuracy for the super-resolution experiment. Red solid lines [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Off-grid bias evaluation under uniform baselines for a continuously swept single-scatterer scene [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: RMSE robustness to nonuniform baselines under SNR = 10 dB. Grouped bars show the mean [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Generalization across varying number of baselines. DUSG-Tomo-Net is trained only at [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Acquisition geometry of the stacked SAR images from CH-1. [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Real test area used for the practical demonstration. Left: UAV optical image. Right: SAR [PITH_FULL_IMAGE:figures/full_fig_p022_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Color-coded real-scene elevation reconstruction comparison between DUSG-Tomo-Net and [PITH_FULL_IMAGE:figures/full_fig_p023_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Slice-wise comparison on the real scene. Left: slice [PITH_FULL_IMAGE:figures/full_fig_p024_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Histogram of normalized elevation distance between detected double-scatterer components. [PITH_FULL_IMAGE:figures/full_fig_p025_12.png] view at source ↗
read the original abstract

Synthetic aperture radar tomography (TomoSAR) enables 3-D imaging by exploiting multibaseline acquisitions and has become an important tool for urban mapping. To achieve super-resolution inversion, sparse reconstruction methods based on compressive sensing (CS) are widely adopted. However, most CS-based TomoSAR methods rely on grid-based formulations and therefore suffer from off-grid bias. Gridless formulations provide a principled way to alleviate this limitation, whereas classical Toeplitz-Vandermonde atomic norm minimization (ANM) is not directly applicable to spaceborne TomoSAR under nonuniform baselines. Existing gridless methods for nonuniform-baseline TomoSAR avoid the classical uniform linear array (ULA) assumption, but they are usually tightly coupled to handcrafted iterative solvers and solver-specific parameter settings, while robust inversion under limited observations and low-SNR conditions remains challenging. To address this gap, we propose DUSG-Tomo-Net, a deep unfolded gridless framework for single-look spaceborne TomoSAR under nonuniform baselines. The proposed method reformulates the inversion in a Toeplitz-compatible lag domain via a structured single-look approximation and recovers a Hermitian Toeplitz positive-semidefinite structured covariance representation through layerwise learned regularization and projection-based structural enforcement. The actual acquisition geometry is embedded analytically into the data-consistency step via a fixed, signal-independent operator, enabling operator-based adaptation to varying baseline configurations. Scatterer elevations are then estimated by a continuous-domain spectral estimator without elevation discretization.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes DUSG-Tomo-Net, a deep-unfolded gridless neural network for super-resolving single-look spaceborne SAR tomography under nonuniform baselines. It reformulates the inversion problem in a Toeplitz-compatible lag domain using a structured single-look approximation, recovers a Hermitian Toeplitz positive-semidefinite covariance representation via layerwise learned regularization parameters and projection-based structural enforcement, embeds the acquisition geometry analytically into a fixed signal-independent data-consistency operator, and estimates scatterer elevations via continuous-domain spectral analysis without discretization.

Significance. If the structured single-look approximation is valid and the learned components deliver robust performance, the method would offer a principled way to combine analytical geometry embedding with data-driven regularization for gridless TomoSAR, potentially improving super-resolution under limited observations and low SNR compared to handcrafted iterative solvers or grid-based CS approaches.

major comments (2)
  1. [Section on structured single-look approximation and lag-domain reformulation] The structured single-look approximation (introduced to enable the lag-domain Toeplitz reformulation) is load-bearing for the central claim of accurate Hermitian Toeplitz PSD covariance recovery. For nonuniform baselines typical in spaceborne TomoSAR, baseline-dependent phase curvature and the rank-1 nature of single-look data may violate the stationarity assumption required for the lag mapping to preserve Toeplitz structure; the manuscript should provide either a formal proof of preservation or targeted numerical validation (e.g., deviation metrics on simulated nonuniform baselines) showing that subsequent projections enforce the true rather than an incorrect structure.
  2. [Description of the unfolded network layers and learned regularization] The data-consistency operator is described as fixed and signal-independent, which is a strength, but the layerwise learned regularization parameters are fitted to training data. The manuscript should clarify how these parameters generalize across varying baseline configurations and SNR regimes without retraining, and whether the projection steps remain stable when the input covariance deviates from exact Toeplitz due to the approximation.
minor comments (2)
  1. [Method formulation] Notation for the lag-domain mapping and the single-look approximation could be made more explicit with an equation showing the explicit transformation from multibaseline observations to the structured covariance.
  2. [Experimental results] The abstract mentions robust inversion under low-SNR and limited observations, but the manuscript should ensure that all experimental figures include error bars or statistical summaries over multiple realizations to support these claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help us improve the clarity and rigor of the manuscript. We address each major comment point by point below, indicating the revisions we will incorporate.

read point-by-point responses

  1. Referee: [Section on structured single-look approximation and lag-domain reformulation] The structured single-look approximation (introduced to enable the lag-domain Toeplitz reformulation) is load-bearing for the central claim of accurate Hermitian Toeplitz PSD covariance recovery. For nonuniform baselines typical in spaceborne TomoSAR, baseline-dependent phase curvature and the rank-1 nature of single-look data may violate the stationarity assumption required for the lag mapping to preserve Toeplitz structure; the manuscript should provide either a formal proof of preservation or targeted numerical validation (e.g., deviation metrics on simulated nonuniform baselines) showing that subsequent projections enforce the true rather than an incorrect structure.

    Authors: We agree that the structured single-look approximation is central to enabling the lag-domain reformulation and the subsequent Toeplitz recovery. Section III-B of the manuscript derives the lag mapping under this approximation and Section V-A reports numerical experiments on simulated nonuniform baselines, including Frobenius-norm deviation metrics before and after projection that remain small (relative error ~10^{-3}). We acknowledge that these results constitute targeted numerical validation rather than a formal proof of exact structure preservation. In the revised manuscript we will expand Section III with a short error-bound analysis of the approximation under typical spaceborne phase curvature and add additional deviation metrics across a broader ensemble of baseline nonuniformities to further substantiate that the projections recover the intended structure. revision: yes

  2. Referee: [Description of the unfolded network layers and learned regularization] The data-consistency operator is described as fixed and signal-independent, which is a strength, but the layerwise learned regularization parameters are fitted to training data. The manuscript should clarify how these parameters generalize across varying baseline configurations and SNR regimes without retraining, and whether the projection steps remain stable when the input covariance deviates from exact Toeplitz due to the approximation.

    Authors: The fixed, signal-independent data-consistency operator (Section III-C) analytically embeds the acquisition geometry, allowing the same trained network to be applied to new baseline configurations without retraining. The regularization parameters were learned on a training distribution that spans multiple baseline nonuniformities and SNR levels from 0 dB to 20 dB (Section IV). We will add explicit cross-validation results on held-out baseline geometries and SNR regimes in the revised experiments section to demonstrate generalization. The projection steps are closed-form convex operations (eigenvalue clipping for PSD and diagonal averaging for Toeplitz); we will include a brief stability analysis in the supplement showing that small input deviations from exact Toeplitz do not cause divergence or instability in the unfolded iterations. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation remains self-contained

full rationale

The paper's core steps—an explicit structured single-look approximation to reach a lag-domain Toeplitz form, followed by a fixed signal-independent data-consistency operator derived from acquisition geometry, plus standard layerwise learned regularization and projection enforcement—do not reduce the claimed super-resolution result to its own inputs by construction. The geometry operator supplies external grounding independent of the learned parameters, and the approximation is introduced as a modeling choice rather than a tautological redefinition. No load-bearing self-citation, uniqueness theorem imported from the same authors, or fitted quantity renamed as prediction appears in the derivation chain. The framework therefore retains independent mathematical and physical content.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of the structured single-look approximation and the effectiveness of learned regularization plus projection steps; no new physical entities are postulated.

free parameters (1)
  • layerwise learned regularization parameters
    Introduced to enforce structure in the unfolded network layers; values are determined during training rather than derived from first principles.
axioms (1)
  • domain assumption Structured single-look approximation enables Toeplitz-compatible lag-domain reformulation for nonuniform baselines
    Invoked to reformulate the inversion problem before applying the network.

pith-pipeline@v0.9.0 · 5593 in / 1391 out tokens · 48584 ms · 2026-05-10T02:27:22.214542+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 31 canonical work pages

  1. [1]

    Very High Resolution Spaceborne

    Zhu, Xiao Xiang and Bamler, Richard , journal=. Very High Resolution Spaceborne. 2010 , volume=

    2010

  2. [2]

    Ice sheet bed mapping with airborne

    Wu, Xiaoqing and Jezek, Kenneth C and Rodriguez, Ernesto and Gogineni, Sivaprasad and Rodriguez-Morales, Fernando and Freeman, Anthony , journal=. Ice sheet bed mapping with airborne. 2011 , publisher=

    2011

  3. [3]

    and Moreira, A

    Reigber, A. and Moreira, A. , year =. First demonstration of airborne. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/36.868873 , number =

    work page doi:10.1109/36.868873

  4. [4]

    and Lombardini, F

    Fornaro, G. and Lombardini, F. and Serafino, F. , year =. Three-dimensional multipass. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/tgrs.2005.843567 , number =

    work page doi:10.1109/tgrs.2005.843567 2005

  5. [5]

    Tomographic

    Zhu, Xiao Xiang and Bamler, Richard , year =. Tomographic. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/tgrs.2010.2048117 , number =

    work page doi:10.1109/tgrs.2010.2048117 2010

  6. [6]

    Super-Resolution Power and Robustness of Compressive Sensing for Spectral Estimation With Application to Spaceborne Tomographic

    Zhu, Xiao Xiang and Bamler, Richard , year =. Super-Resolution Power and Robustness of Compressive Sensing for Spectral Estimation With Application to Spaceborne Tomographic. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/tgrs.2011.2160183 , number =

    work page doi:10.1109/tgrs.2011.2160183 2011

  7. [7]

    and Reale, D

    Fornaro, G. and Reale, D. and Serafino, F. , year =. Four-Dimensional. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/tgrs.2008.2000837 , number =

    work page doi:10.1109/tgrs.2008.2000837 2008

  8. [8]

    Tomographic Processing of Interferometric

    Fornaro, Gianfranco and Lombardini, Fabrizio and Pauciullo, Antonio and Reale, Diego and Viviani, Federico , year =. Tomographic Processing of Interferometric. IEEE Signal Processing Magazine , publisher =. doi:10.1109/msp.2014.2312073 , number =

    work page doi:10.1109/msp.2014.2312073 2014

  9. [9]

    Three-Dimensional

    Budillon, Alessandra and Evangelista, Annarita and Schirinzi, Gilda , year =. Three-Dimensional. IEEE Transactions on Geoscience and Remote Sensing , publisher =. doi:10.1109/tgrs.2010.2054099 , number =

    work page doi:10.1109/tgrs.2010.2054099 2010

  10. [10]

    , year =

    Cloude, Shane R. , year =. Polarization coherence tomography , volume =. Radio Science , publisher =. doi:10.1029/2005rs003436 , number =

    work page doi:10.1029/2005rs003436

  11. [11]

    Compressed Sensing Off the Grid , year=

    Tang, Gongguo and Bhaskar, Badri Narayan and Shah, Parikshit and Recht, Benjamin , journal=. Compressed Sensing Off the Grid , year=

  12. [12]

    Atomic Norm Denoising With Applications to Line Spectral Estimation , year=

    Bhaskar, Badri Narayan and Tang, Gongguo and Recht, Benjamin , journal=. Atomic Norm Denoising With Applications to Line Spectral Estimation , year=

  13. [13]

    Tomographic

    Wang, Xiao and Xu, Feng , journal=. Tomographic. 2022 , volume=

    2022

  14. [14]

    2024 , volume=

    Shao, Mingxiao and Zhang, Zhe and Li, Jie and Kang, Jian and Zhang, Bingchen , journal=. 2024 , volume=

    2024

  15. [15]

    A Novel Gradient Descent Least-Squares (

    Shi, Ruizhe and Zhang, Zhe and Qiu, Xiaolan and Ding, Chibiao , journal=. A Novel Gradient Descent Least-Squares (. 2023 , volume=

    2023

  16. [16]

    A Robust Super-Resolution Gridless Imaging Framework for

    Gao, Silin and Wang, Wenlong and Wang, Muhan and Zhang, Zhe and Yang, Zai and Qiu, Xiaolan and Zhang, Bingchen and Wu, Yirong , journal=. A Robust Super-Resolution Gridless Imaging Framework for. 2024 , volume=

    2024

  17. [17]

    Gregor and Y

    K. Gregor and Y. LeCun , title =. Proc. Int. Conf. Mach. Learn. (ICML) , year =

  18. [18]

    , journal=

    Monga, Vishal and Li, Yuelong and Eldar, Yonina C. , journal=. Algorithm Unrolling: Interpretable, Efficient Deep Learning for Signal and Image Processing , year=

  19. [19]

    doi:10.1109/tgrs.2022.3164193 , journal =

    Qian, Kun and Wang, Yuanyuan and Shi, Yilei and Zhu, Xiao Xiang , year =. doi:10.1109/tgrs.2022.3164193 , journal =

    work page doi:10.1109/tgrs.2022.3164193 2022

  20. [20]

    doi:10.1109/tgrs.2023.3268132 , journal =

    Wang, Muhan and Zhang, Zhe and Qiu, Xiaolan and Gao, Silin and Wang, Yue , year =. doi:10.1109/tgrs.2023.3268132 , journal =

    work page doi:10.1109/tgrs.2023.3268132 2023

  21. [21]

    doi:10.1109/tgrs.2024.3391066 , journal =

    Qian, Kun and Wang, Yuanyuan and Jung, Peter and Shi, Yilei and Zhu, Xiao Xiang , year =. doi:10.1109/tgrs.2024.3391066 , journal =

    work page doi:10.1109/tgrs.2024.3391066 2024

  22. [22]

    Deep Unfolded Gridless

    Zhu, Hangui and Feng, Weike and Feng, Cunqian and Ma, Teng and Zou, Bo , year =. Deep Unfolded Gridless. Remote Sensing , publisher =. doi:10.3390/rs15010013 , number =

    work page doi:10.3390/rs15010013

  23. [23]

    Basis Pursuit Denoising via Recurrent Neural Network Applied to Super-Resolving

    Qian, Kun and Wang, Yuanyuan and Jung, Peter and Shi, Yilei and Zhu, Xiao Xiang , year =. Basis Pursuit Denoising via Recurrent Neural Network Applied to Super-Resolving. doi:10.1109/tgrs.2022.3221185 , journal =

    work page doi:10.1109/tgrs.2022.3221185 2022

  24. [24]

    SIAM Journal on Scientific Computing20(1), 33–61 (1998) https://doi.org/ 10.1137/S1064827596304010 17

    Chen, Scott Shaobing and Donoho, David L. and Saunders, Michael A. , year =. Atomic Decomposition by Basis Pursuit , volume =. SIAM Journal on Scientific Computing , publisher =. doi:10.1137/s1064827596304010 , number =

    work page doi:10.1137/s1064827596304010

  25. [25]

    , year =

    Donoho, D.L. , year =. Compressed sensing , volume =. IEEE Transactions on Information Theory , publisher =. doi:10.1109/tit.2006.871582 , number =

    work page doi:10.1109/tit.2006.871582 2006

  26. [26]

    and Defrise, M

    Daubechies, I. and Defrise, M. and De Mol, C. , year =. An iterative thresholding algorithm for linear inverse problems with a sparsity constraint , volume =. Communications on Pure and Applied Mathematics , publisher =. doi:10.1002/cpa.20042 , number =

    work page doi:10.1002/cpa.20042

  27. [27]

    A fast iterative shrinkage-thresholding algorithm for linear inverse problems

    Beck, Amir and Teboulle, Marc , year =. A Fast Iterative Shrinkage-Thresholding Algorithm for Linear Inverse Problems , volume =. SIAM Journal on Imaging Sciences , publisher =. doi:10.1137/080716542 , number =

    work page doi:10.1137/080716542

  28. [28]

    , year =

    Schmidt, R. , year =. Multiple emitter location and signal parameter estimation , volume =. IEEE Transactions on Antennas and Propagation , publisher =. doi:10.1109/tap.1986.1143830 , number =

    work page doi:10.1109/tap.1986.1143830 1986

  29. [29]

    and Hari, K.V.S

    Rao, B.D. and Hari, K.V.S. , journal=. Performance analysis of Root-. 1989 , volume=

    1989

  30. [30]

    Roy and T

    Roy, R. and Kailath, T. , year =. ESPRIT-estimation of signal parameters via rotational invariance techniques , volume =. IEEE Transactions on Acoustics, Speech, and Signal Processing , publisher =. doi:10.1109/29.32276 , number =

    work page doi:10.1109/29.32276

  31. [31]

    On Gridless Sparse Methods for Line Spectral Estimation From Complete and Incomplete Data , volume =

    Yang, Zai and Xie, Lihua , year =. On Gridless Sparse Methods for Line Spectral Estimation From Complete and Incomplete Data , volume =. IEEE Transactions on Signal Processing , publisher =. doi:10.1109/tsp.2015.2420541 , number =

    work page doi:10.1109/tsp.2015.2420541 2015

  32. [32]

    Remote Sensing , publisher =

    Liu, Ning and Li, Xinwu and Peng, Xing and Hong, Wen , year =. Remote Sensing , publisher =. doi:10.3390/rs14143439 , number =

    work page doi:10.3390/rs14143439

  33. [33]

    Foundations and Trends® in Machine Learning , author =

    Boyd, Stephen , year =. Distributed Optimization and Statistical Learning via the Alternating Direction Method of Multipliers , volume =. Foundations and Trends in Machine Learning , publisher =. doi:10.1561/2200000016 , number =

    work page doi:10.1561/2200000016

  34. [34]

    , year =

    Dykstra, Richard L. , year =. An Algorithm for Restricted Least Squares Regression , volume =. Journal of the American Statistical Association , publisher =. doi:10.1080/01621459.1983.10477029 , number =

    work page doi:10.1080/01621459.1983.10477029 1983

  35. [35]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Forest height retrieval using. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2021 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2021.02.022 , author =

    work page doi:10.1016/j.isprsjprs.2021.02.022 2021

  36. [36]

    High-Resolution and Wide-Swath

    Yang, Yaqian and Zhang, Fubo and Tian, Ye and Chen, Longyong and Wang, Robert and Wu, Yirong , year =. High-Resolution and Wide-Swath. Remote Sensing , publisher =. doi:10.3390/rs15163938 , number =

    work page doi:10.3390/rs15163938

  37. [37]

    doi:10.1109/tgrs.2023.3239405 , journal =

    Wang, Yan and Liu, Changhao and Zhu, Rui and Liu, Minkun and Ding, Zegang and Zeng, Tao , year =. doi:10.1109/tgrs.2023.3239405 , journal =

    work page doi:10.1109/tgrs.2023.3239405 2023

  38. [38]

    Efficient

    Wang, Xiao and Xu, Feng , year =. Efficient. doi:10.1109/tgrs.2024.3395510 , journal =

    work page doi:10.1109/tgrs.2024.3395510 2024

  39. [39]

    Atomic Norm Minimization Based Fast Off-Grid Tomographic

    Liu, Minkun and Wang, Yan and Ding, Zegang and Li, Linghao and Zeng, Tao , journal=. Atomic Norm Minimization Based Fast Off-Grid Tomographic. 2024 , volume=

    2024

  40. [40]

    doi:10.1016/j.isprsjprs.2025.01.004 , journal =

    Liu, Changhao and Wang, Yan and Zhang, Guangbin and Ding, Zegang and Zeng, Tao , year =. doi:10.1016/j.isprsjprs.2025.01.004 , journal =

    work page doi:10.1016/j.isprsjprs.2025.01.004 2025

  41. [41]

    and Pezeshki, Ali and Calderbank, A

    Chi, Yuejie and Scharf, Louis L. and Pezeshki, Ali and Calderbank, A. Robert , journal=. Sensitivity to Basis Mismatch in Compressed Sensing , year=

  42. [42]

    2021 , issn =

    Procedia Computer Science , volume =. 2021 , issn =. doi:https://doi.org/10.1016/j.procs.2021.01.143 , author =

    work page doi:10.1016/j.procs.2021.01.143 2021

  43. [43]

    Regression Shrinkage and Selection via the Lasso , volume =

    Robert Tibshirani , journal =. Regression Shrinkage and Selection via the Lasso , volume =

  44. [44]

    1997 , publisher=

    Primal-Dual Interior-Point Methods , author=. 1997 , publisher=

    1997

  45. [45]

    2014 , issue_date =

    Saeed, Khalid , title =. 2014 , issue_date =. doi:10.1016/j.apnum.2012.05.004 , journal =

    work page doi:10.1016/j.apnum.2012.05.004 2014

  46. [46]

    PyTorch: an imperative style, high-performance deep learning library , year =

    Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and Desmaison, Alban and K\". PyTorch: an imperative style, high-performance deep learning library , year =. Proceedings of the 33rd International Conference on Neural Informa...

  47. [47]

    2017 , eprint=

    Adam: A Method for Stochastic Optimization , author=. 2017 , eprint=

    2017

  48. [48]

    and Montanari, M

    Lombardini, F. and Montanari, M. and Gini, F. , journal=. Reflectivity estimation for multibaseline interferometric radar imaging of layover extended sources , year=

  49. [49]

    The root-

    Friedlander, Benjamin , year =. The root-. Signal Processing , publisher =. doi:10.1016/0165-1684(93)90048-f , number =

    work page doi:10.1016/0165-1684(93)90048-f

  50. [50]

    Pal, Piya and Vaidyanathan, P. P. , journal=. Nested Arrays: A Novel Approach to Array Processing With Enhanced Degrees of Freedom , year=

  51. [51]

    and Pal, Piya , journal=

    Vaidyanathan, Palghat P. and Pal, Piya , journal=. Sparse Sensing With Co-Prime Samplers and Arrays , year=

  52. [52]

    and Frazho, A.E

    Grigoriadis, K.M. and Frazho, A.E. and Skelton, R.E. , journal=. Application of alternating convex projection methods for computation of positive Toeplitz matrices , year=

  53. [53]

    In connection with cross-referencing and possible future hyperlinking it is not a good idea to collect more that one literature item in one + +

    + is cited as + ESG96 +. In connection with cross-referencing and possible future hyperlinking it is not a good idea to collect more that one literature item in one + +. The so-called Harvard or author-year style of referencing is enabled by the package natbib . With this package the literature can be cited as follows: enumerate [ ] Parenthetical: + WB96 ...

    1996