Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

The central claim is that the sample covariance matrix of received echoes is a sufficient statistic for extended-target imaging, and that jointly optimizing grid positions with scattering intensities yields high-quality reflectivity maps ev

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 21:36 UTC pith:4M5RTW5U

load-bearing objection Covariance-based ML imaging with dynamic grid optimization is a promising combination, but the grid-gradient derivation omits path-loss derivatives and the numerical evidence is too thin to support the strong claims; still, the paper deserves serious peer review with major revision. the 3 major comments →

arxiv 2511.14490 v2 pith:4M5RTW5U submitted 2025-11-18 eess.SP cs.ITmath.IT

Multi-view imaging in networked sensing systems: A covariance-based approach

Junyuan Gao , Weifeng Zhu , Yanmo Hu , Shuowen Zhang , Jiannong Cao , Yongpeng Wu , Giuseppe Caire , Liang Liu This is my paper
classification eess.SP cs.ITmath.IT
keywords extended target imagingcovariance-based imagingmulti-view fusionintegrated sensing and communicationnetworked sensinggrid position optimizationnatural neighbor interpolationfield-of-view estimation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper sets out to show that in a 6G network where one base station transmits and several base stations listen, the sample covariance matrix of the received echo signals carries enough information to image extended targets—objects large enough that they cannot be treated as points. The authors propose a two-phase method: each receiver first builds a single-view reflectivity map by jointly estimating scattering intensities and the positions of the image grid, and then a central processor fuses the per-receiver maps, which live on different grids and cover different, unknown fields of view. The central claim is that working with second-order statistics instead of instantaneous channels, and letting grid points move to conform to target shape, avoids the resolution limits and coherence collapse that hurt FFT and compressed-sensing baselines. If true, high-quality environment reconstruction becomes feasible with narrowband, small-aperture, low-power, short-pilot resources. The reported simulations support the claim across five resource-limited regimes, with substantially better sidelobe suppression and shape overlap than the benchmarks.

Core claim

The discovery, on the paper's own terms, is that the map of effective scattering intensity—not the instantaneous channel—is the right object to estimate for extended-target imaging, and that the maximum-likelihood cost function based on the sample covariance matrix can be minimized jointly over intensities and grid positions. The covariance matrix Σk = Vk diag(γr,k ⊙ γβ,k) Vk^H + σ²I is the sufficient statistic; the likelihood follows from the Swerling-II assumption that each scatterer's attenuation stays constant within a frame and changes independently across frames. A graph-Laplacian penalty enforces spatial clustering, and an alternating optimization updates intensities by closed-form cu

What carries the argument

The load-bearing machinery is the sample covariance matrix treated as a sufficient statistic. Under the Swerling-II fluctuation model, the received columns are i.i.d. Gaussian with covariance Σk = Vk diag(γr,k ⊙ γβ,k) Vk^H + σ²I, so the negative log-likelihood becomes a function of the covariance alone and the imaging problem is to choose the grid-dependent steering matrix Vk and intensity vector γr,k that best match the sample covariance. Two devices carry the argument: the Sherman–Morrison rank-one update, which turns each intensity coordinate update into a cubic with closed-form roots, and a coherence bound on the dictionary Vk showing that increasing grid density Q drives the maximum col

Load-bearing premise

The whole construction rests on the Swerling-II assumption that each scatterer's attenuation coefficient is a circularly symmetric complex Gaussian that stays constant within a frame and changes independently across frames; if real extended targets are quasi-static or partially coherent across frames, the sample covariance is no longer a sufficient statistic and the maximum-likelihood estimate is not the reflectivity map.

What would settle it

Take a target with a static (Swerling-0) radar cross section across all M frames and run the proposed Phase I estimator. Under the paper's model, the covariance-based likelihood is misspecified for this target, so if the recovered reflectivity map still matches the true shape with IoU near the reported 0.82–0.89, the i.i.d.-across-frames assumption is not essential; if the map warps or fades, the method's success is tied to the fluctuation model it assumes.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • The number of estimated parameters drops from a full instantaneous channel matrix to an intensity vector, so imaging can operate with narrowband pilots and small antenna arrays where FFT and CS reconstructions break down.
  • Because the sample covariance is declared sufficient, receivers could in principle share only covariance information with the central processor, a natural fit for limited-capacity fronthaul links.
  • Optimizing grid positions lets a small number of adaptive grids match the performance of much denser fixed grids, and avoids the coherence collapse that appears when grid density is pushed beyond the resolution limit.
  • Fusing images with explicitly estimated fields of view removes the need for receivers to agree on a common visibility region, which earlier fusion methods assumed or handled with ad hoc clustering.
  • The reported numbers—P-ISLR from -7.65 to -11.37 dB and IoU from 0.82 to 0.89 across five resource-limited regimes—quantify the gain over FFT and CS benchmarks.

Where Pith is reading between the lines

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

  • A natural extension the paper leaves implicit: compress or quantize the sample covariance matrix before transmission to the CPU, and study the trade-off between fronthaul rate and image quality; the sufficient-statistic claim makes this a well-posed rate–distortion problem.
  • The dynamic-grid formulation sits between fixed-grid sparse recovery and fully gridless methods; one could push it to a continuum by letting grid points move without the d_max constraint, effectively an off-grid variant of the same likelihood.
  • The Swerling-II assumption is the most fragile point. If real 6G scattering is quasi-static over many frames, a robust likelihood that models partial temporal coherence would be needed; until such a model is tested, the performance gains should be read as conditional on the fluctuation model.
  • The fusion machinery—edge-preserving interpolation plus binary field-of-view selection—is not radio-specific; it could be lifted to other multi-sensor imaging problems where sensors see different, unknown subsets of a scene.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. The paper considers multi-view imaging of extended targets in a 6G networked sensing system with one transmit BS, K receive BSs, and a central processing unit. Under a Swerling-II target model, the received signal columns are i.i.d. Gaussian, and the paper proposes a covariance-based formulation in which the sample covariance matrix is treated as a sufficient statistic for imaging. Phase I formulates single-view imaging as penalized maximum-likelihood estimation of effective scattering intensities and grid positions, solved by alternating optimization: coordinate descent with Sherman–Morrison rank-one updates for the intensities, and gradient descent with projection for the grid positions. Phase II interpolates the heterogeneous single-view images onto a common fine grid using edge-preserving natural neighbor interpolation, then fuses them by an ADMM-based optimization that jointly estimates the fused intensity and the receivers' fields of view via a weighted least-squares term, an ℓ1 sparsity term, and a total-variation term. The paper includes a theoretical analysis of dynamic grids and numerical comparisons against FFT-IFFT and compressed-sensing baselines in five resource-limited regimes, reporting P-ISLR and IoU improvements.

Significance. If the proposed approach performs as claimed, the covariance-based single-view imaging framework is a useful contribution to ISAC/environment imaging: it reduces the number of estimated parameters from instantaneous channel coefficients to statistical scattering intensities, and the dynamic-grid idea is well motivated by the subspace-alignment analysis in Sec. IV-D. The coordinate-descent derivation in Sec. IV-B is detailed and algebraically consistent, and the EP-NNI fusion method is a novel adaptation of computer-vision tools. The ablation in Fig. 2 clearly illustrates the separate effects of the cluster penalty, grid optimization, and edge-preserving interpolation. However, the empirical validation is currently single-realization and fully in-sample, and the grid-position gradient omits the path-loss derivative, so the central 'joint ML estimation of intensities and grid positions' claim is not currently supported as stated.

major comments (3)
  1. [§IV-C, Eqs. (40)–(41)] The grid-position gradient is incomplete. In the model, γβ,k,q is an explicit function of p_k,q through the path-loss law in Sec. VI, γβ,k,q = β0² d_tx^{-2} d_rx^{-2}, so ∂γ_k,q/∂p = γr,k,q ∂γβ,k,q/∂p ≠ 0. Eqs. (40)–(41) differentiate only the steering-vector columns v_k,q; no ∂γβ/∂p term appears, even though Eq. (42) uses γ_k,q = γr,k,qγβ,k,q. Consequently, update (38) is not a gradient-descent step for objective (37)/(21), and the stopping criterion ∇ψ f ≤ ε2 in Algorithm 1 is not a stationarity condition for the actual problem. Please add the missing derivative terms, or explicitly state and justify that γβ is held fixed during grid optimization; under constraint (22d) the neglected terms can be significant at the simulated distances (order 10–20% per meter).
  2. [§VI, Table I and Figs. 2–5] All reported P-ISLR/IoU values and displayed images are single realizations; no Monte Carlo averaging or error bars are given, although the received signals are stochastic (Swerling-II fluctuations plus noise). In addition, the simulation setup does not specify the hyperparameters η, μ, σ_EP², ρ, d_max, or the Algorithm 1/2 tolerances and iteration counts. Without these values, the claimed consistent improvements (often 8–10 dB in P-ISLR) cannot be statistically assessed and the results are not reproducible. Please provide a full parameter table and Monte Carlo results with error bars or confidence intervals.
  3. [§II and §VI] The validation is entirely in-sample: the simulator uses exactly the Swerling-II statistical model for which the likelihood (18) is derived. The paper's practical claim for 6G environment imaging would be stronger if robustness to mismatched target statistics (e.g., quasi-static or partially coherent scatterers across frames) were demonstrated. As it stands, the reported gains may partly reflect that the estimator is given the exact model family used to generate the data.
minor comments (7)
  1. [Eq. (51)] Typo: the denominator should contain x_tx, not x_tx,k. Check Eq. (52) for consistency of notation.
  2. [Algorithm 1, line 4] The symbol Q is used both for the number of grids and for the randomly selected index set; rename one of them to avoid ambiguity.
  3. [Eq. (53)] In the second branch, p^(0)_q should be p^(0)_{k,q} for consistency with the first branch.
  4. [Eq. (35) and Algorithm 1] ℜ(d) is unnecessary because d is real. Also, the stopping criterion in Algorithm 1 line 20 should use a norm of the gradient, e.g., ‖∇ψ f‖₂ ≤ ε2, and it depends on the corrected gradient.
  5. [Eq. (79)] The discrete gradient operators D_x and D_y cannot have all entries in {−1,1}; they necessarily contain zero entries. Please correct the set notation.
  6. [Fig. 2] The color scales differ across panels (e.g., (c) and (d) vs. (e) and (f)); use a common color scale for a fair visual comparison.
  7. [Eq. (91) and IoU definition] The P-ISLR mainlobe/sidelobe cell determination and the 95% cumulative-intensity threshold rule should be described more precisely; state how the threshold is chosen and whether the reported values are sensitive to it.

Circularity Check

0 steps flagged

No significant circularity: the derivation chain is self-contained and the reported predictions are not fitted or definitionally forced.

full rationale

The paper's core derivation is not circular. The objective (21) is built from the Gaussian likelihood (18) of the observed data and depends only on the sample covariance matrix Δ^R_k, the model covariance Σ_k, and the regularization prior; no ground-truth reflectivity map is used in the estimator. The grid-position update (38) uses gradients of the same cost function, not of the target geometry. The reported P-ISLR and IoU in Table I are evaluation metrics defined against the true target (91)-(92), but these metrics are not fed back into the optimization. The paper's self-citations ([2], [3], [17]-[19], [22]) appear only in background or related-work discussion and do not carry the load of the main derivation. Standard external citations support the Born approximation, Swerling-II model, Sherman-Morrison identity, natural-neighbor interpolation, and steering-kernel weights. The in-sample synthetic evaluation uses the same statistical model family, which can be a validation limitation, but it is not circularity: the estimator does not consume the ground-truth values or metric outcomes as inputs. The skeptic's concern about omitting the path-loss derivative in the grid-position gradient is a stationarity/correctness issue rather than a reduction of a prediction to an input, so it does not count as circularity under the stated rules.

Axiom & Free-Parameter Ledger

7 free parameters · 6 axioms · 0 invented entities

The method relies on standard ISAC modeling assumptions (Born, Swerling-II, Gaussian noise, known geometry) and on several hand-chosen regularizers with undisclosed hyperparameters. No new physical entities are introduced. The covariance-based formulation is not circular, but its validation is entirely within the same model family used to derive it.

free parameters (7)
  • η (Phase I cluster penalty) = not specified
    Weights the graph-Laplacian smoothness term in (21); no value or selection rule is given, and results may be sensitive to it.
  • μ (fusion sparsity penalty) = not specified
    Weights ||γ'_r||_1 in (76); controls how aggressively low-intensity grids are zeroed; value not stated.
  • η (fusion TV penalty) = not specified
    Weights ||Dγ'_r||_1 in (76); balances boundary sharpness against false-alarm suppression; value not stated.
  • σ_EP^2 (edge-sensitive decay) = not specified
    Sets the anisotropy width in the EP-NNI weight (74); no value given.
  • ρ (ADMM penalty factor) = not specified
    ADMM penalty in (81); affects convergence rate and final solution; not specified.
  • d_max (grid displacement bound) = not specified
    Euclidean-ball constraint (22d) preventing grid swaps; value not given.
  • Algorithm tolerances/iterations (ǫ1, ǫ2, Iter_max, Iter1, Iter2) = not specified
    Stopping criteria in Algorithms 1–2 are not given, so the reported reconstructions are not exactly reproducible.
axioms (6)
  • domain assumption Born approximation / single-scattering linear model (Eq. (2))
    The channel is a linear superposition of independent point scatterers; multiple scattering is ignored.
  • domain assumption Swerling-II fluctuation: scatterer amplitudes are i.i.d. zero-mean CSCG and change independently across frames (Sec. II, after Eq. (6))
    Makes sample covariance sufficient and justifies likelihood (18); static/coherent targets break the model.
  • domain assumption Far-field plane-wave steering vectors (Eqs. (5)–(6))
    Steering vectors depend only on angle of departure/arrival, not range; with the 15 m RoI and 16-antenna arrays, some targets may be in the near field, which is not modeled.
  • domain assumption Known path-loss and perfect synchronization/positions
    The paper explicitly assumes BS locations are known and fronthaul synchronization is perfect; path-loss model γβ is known.
  • domain assumption Grid-discretization CLT approximation (Sec. III)
    Treats the aggregate scattering of a grid cell as a zero-mean CSCG variable with variance γ_{k,q}; assumes enough scatterers per cell for the CLT.
  • ad hoc to paper Spatial smoothness prior (graph-Laplacian penalty (20))
    Introduced as a physical prior; no evidence provided that targets satisfy this smoothness, and the penalty weight is undisclosed.

pith-pipeline@v1.3.0-alltime-deepseek · 24311 in / 19932 out tokens · 189673 ms · 2026-08-03T21:36:24.042126+00:00 · methodology

0 comments
read the original abstract

This paper considers multi-view imaging in a sixth-generation (6G) integrated sensing and communication network, which consists of a transmit base-station (BS), multiple receive BSs connected to a central processing unit (CPU), and multiple extended targets. Our goal is to devise an effective multi-view imaging technique that can jointly leverage the targets' echo signals at all the receive BSs to precisely construct the image of these targets. To achieve this goal, we propose a two-phase approach. In Phase I, each receive BS recovers an individual image based on the sample covariance matrix of its received signals. Specifically, we propose a novel covariance-based imaging framework to jointly estimate effective scattering intensity and grid positions, which reduces the number of estimated parameters leveraging channel statistical properties and allows grid adjustment to conform to target geometry. In Phase II, the CPU fuses the individual images of all the receivers to construct a high-quality image of all the targets. Specifically, we design edge-preserving natural neighbor interpolation (EP-NNI) to map individual heterogeneous images onto common and finer grids, and then propose a joint optimization framework to estimate fused scattering intensity and BS fields of view. Extensive numerical results show that the proposed scheme significantly enhances imaging performance, facilitating high-quality environment reconstruction for future 6G networks.

Figures

Figures reproduced from arXiv: 2511.14490 by Giuseppe Caire, Jiannong Cao, Junyuan Gao, Liang Liu, Shuowen Zhang, Weifeng Zhu, Yanmo Hu, Yongpeng Wu.

Figure 1
Figure 1. Figure 1: System model. [k] = {1, . . . , k}. We use 1[·] to denote the indicator function. The distribution of a Bernoulli random variable x with mean λ is denoted as Ber(λ). The distribution of a circularly symmetric complex Gaussian (CSCG) random vector x with mean µ and covariance matrix Σ is denoted as CN (µ, Σ). II. SYSTEM MODEL We consider a multi-view imaging scenario in 6G cellu￾lar network as illustrated i… view at source ↗
Figure 2
Figure 2. Figure 2: Without loss of generality, we assume that the RoI is [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 2
Figure 2. Figure 2: Covariance-based single-view images at receiver [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Covariance-based single-view images at receiver [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Covariance-based single-view images and multi-vie [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Imaging performance comparison between different s [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Networked Tracking of Multiple Moving Targets in 6G Network

    eess.SP 2026-04 unverdicted novelty 4.0

    Networked Kalman filter with PCRB-minimizing beamforming enables dynamic target-BS association and lowers multi-target tracking MSE in 6G systems.

Reference graph

Works this paper leans on

35 extracted references · 1 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,

    F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, an d S. Buzzi, “Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,” IEEE J. Sel. Areas Commun. , vol. 40, no. 6, pp. 1728–1767, Jun. 2022

    2022

  2. [2]

    Leveraging a variety of anch ors in cellular network for ubiquitous sensing,

    L. Liu, S. Zhang, and S. Cui, “Leveraging a variety of anch ors in cellular network for ubiquitous sensing,” IEEE Commun. Mag. , vol. 62, no. 9, pp. 98–104, Sep. 2024

    2024

  3. [3]

    Joint transmission and comp ression optimization for networked sensing with limited-capacity fronthaul links,

    W. Zhu, S. Zhang, and L. Liu, “Joint transmission and comp ression optimization for networked sensing with limited-capacity fronthaul links,” to appear in IEEE Trans. Wireless Commun. , 2025

    2025

  4. [4]

    Spatial diversity in radars–Models and d etection performance,

    E. Fishler, A. Haimovich, R. S. Blum, L. J. Cimini, D. Chiz hik, and R. A. V alenzuela, “Spatial diversity in radars–Models and d etection performance,” IEEE Trans. Signal Process. , vol. 54, no. 3, pp. 823–838, 2006

    2006

  5. [5]

    A direct imaging algorithm for extended targets,

    S. Hou, K. Solna, and H. Zhao, “A direct imaging algorithm for extended targets,” Inverse Problems, vol. 22, no. 4, pp. 1151–1178, Jun. 2006

    2006

  6. [6]

    Distributed multi-sen sor fusion of PHD filters with different sensor fields of view,

    W. Yi, G. Li, and G. Battistelli, “Distributed multi-sen sor fusion of PHD filters with different sensor fields of view,” IEEE Trans. Signal Process. , vol. 68, pp. 5204–5218, 2020

    2020

  7. [7]

    Distributed multi-view sparse vector recovery,

    Z. Tian, Z. Zhang, and L. Hanzo, “Distributed multi-view sparse vector recovery,” IEEE Trans. Signal Process. , vol. 71, pp. 1448–1463, 2023

    2023

  8. [8]

    Multistatic parameter estimation in the near/far field for integrated sensing and communication,

    S. K. Dehkordi, L. Pucci, P . Jung, A. Giorgetti, E. Paolin i, and G. Caire, “Multistatic parameter estimation in the near/far field for integrated sensing and communication,” IEEE Trans. Wireless Commun. , vol. 23, no. 12, pp. 17 929–17 944, Dec. 2024

    2024

  9. [9]

    Synthetic -aperture radar processing using fast factorized back-projection,

    L. M. Ulander, H. Hellsten, and G. Stenstrom, “Synthetic -aperture radar processing using fast factorized back-projection,” IEEE Trans. Aerosp. Electron. Syst. , vol. 39, no. 3, pp. 760–776, Jul. 2003

    2003

  10. [10]

    Two dimensional array imaging with beam steered data,

    S. Patole and M. Torlak, “Two dimensional array imaging with beam steered data,” IEEE Trans. Image Process. , vol. 22, no. 12, pp. 5181– 5189, Dec. 2013

    2013

  11. [11]

    Fourie r transform- based wavenumber domain 3D imaging in RIS-aided communicat ion systems,

    Y . Huang, J. Y ang, W. Tang, C.-K. Wen, and S. Jin, “Fourie r transform- based wavenumber domain 3D imaging in RIS-aided communicat ion systems,” IEEE Trans. Wireless Commun. , vol. 23, no. 10, pp. 13 872– 13 888, Oct. 2024

    2024

  12. [12]

    Join t multi- user communication and sensing exploiting both signal and e nvironment sparsity,

    X. Tong, Z. Zhang, J. Wang, C. Huang, and M. Debbah, “Join t multi- user communication and sensing exploiting both signal and e nvironment sparsity,” IEEE J. Sel. Topics Signal Process. , vol. 15, no. 6, pp. 1409– 1422, Nov. 2021

    2021

  13. [13]

    L(1/2) regularization for ISAR imaging and target enhancement of complex image,

    A. Wu, Y . Wu, Y . Jin, Y . Wang, and Z. Guo, “ L(1/2) regularization for ISAR imaging and target enhancement of complex image,” IEEE Trans. Geosci. Remote Sens. , vol. 60, Dec. 2022, Art. no. 5200410

    2022

  14. [14]

    Total variat ion compressive sensing for 3D shape estimation in short-range imaging rada rs,

    I. Roldan, F. Fioranelli, and A. Y arovoy, “Total variat ion compressive sensing for 3D shape estimation in short-range imaging rada rs,” IEEE Trans. Radar Systems , vol. 1, pp. 583–592, 2023

    2023

  15. [15]

    N ear-field integrated imaging and communication in distributed MIMO n etworks,

    K. Zhi, T. Y ang, S. Li, Y . Song, A. Rezaei, and G. Caire, “N ear-field integrated imaging and communication in distributed MIMO n etworks,” arXiv preprint arXiv:2508.17526 , 2025. 14

    Pith/arXiv arXiv 2025

  16. [16]

    On spatial smoothin g for direction- of-arrival estimation of coherent signals,

    T.-J. Shan, M. Wax, and T. Kailath, “On spatial smoothin g for direction- of-arrival estimation of coherent signals,” IEEE Trans. Acoust., Speech, Signal Process. , vol. 33, no. 4, pp. 806–811, Aug. 1985

    1985

  17. [17]

    Device-free sensin g in OFDM cellular network,

    Q. Shi, L. Liu, S. Zhang, and S. Cui, “Device-free sensin g in OFDM cellular network,” IEEE J. Sel. Areas Commun. , vol. 40, no. 6, pp. 1838– 1853, Jun. 2022

    2022

  18. [18]

    Joint LOS identification and data asso ciation for 6G- enabled networked device-free sensing,

    Q. Shi and L. Liu, “Joint LOS identification and data asso ciation for 6G- enabled networked device-free sensing,” IEEE Trans. Commun., vol. 72, no. 8, pp. 5117–5129, Aug. 2024

    2024

  19. [19]

    Integrated massive communication and trget localization in 6G cell- free networks,

    J. Gao, W. Zhu, S. Zhang, Y . Wu, J. Cao, G. Caire, and L. Liu , “Integrated massive communication and trget localization in 6G cell- free networks,” arXiv preprint arXiv:2510.14281 , 2025

    arXiv 2025

  20. [20]

    Wavefield networ ked sens- ing: Principles, algorithms and applications,

    M. Manzoni, D. Tagliaferri, S. Tebaldini, M. Mizmizi, A . V . Monti- Guarnieri, C. M. Prati, and U. Spagnolini, “Wavefield networ ked sens- ing: Principles, algorithms and applications,” IEEE Open J. Commun. Soc., vol. 6, pp. 181–197, Jan. 2025

    2025

  21. [21]

    Cooperative coh erent multi- static imaging and phase synchronization in networked sens ing,

    D. Tagliaferri, M. Manzoni, M. Mizmizi, S. Tebaldini, A . V . Monti- Guarnieri, C. M. Prati, and U. Spagnolini, “Cooperative coh erent multi- static imaging and phase synchronization in networked sens ing,” IEEE J. Sel. Areas Commun. , vol. 42, no. 10, pp. 2905–2921, Oct. 2024

    2024

  22. [22]

    A 6G-based multi-view reco nstruction approach,

    Q. Shi, S. Zhang, and L. Liu, “A 6G-based multi-view reco nstruction approach,” in Proc. IEEE Wireless Commun. Netw. Conf. (WCNC) , Milan, Italy, 2025

    2025

  23. [23]

    Sampling and reco nstructing angular domains with uniform arrays,

    S. Mandelli, M. Henninger, and J. Du, “Sampling and reco nstructing angular domains with uniform arrays,” IEEE Trans. Wireless Commun. , vol. 22, no. 6, pp. 3628–3642, Jun. 2022

    2022

  24. [24]

    Radio tomographic imaging with reconfigurable intelligent surfa ces,

    Z. Li, A. Dubey, S. Shen, N. K. Kundu, J. Rao, and R. Murch, “Radio tomographic imaging with reconfigurable intelligent surfa ces,” IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 15 784–15 797, Nov. 2024

    2024

  25. [25]

    P . M. Morse and H. Feshbach, Methods of Theoretical Physics . New Y ork: McGraw-Hill, 1953

    1953

  26. [26]

    Skolnik, Introduction to Radar Systems

    M. Skolnik, Introduction to Radar Systems . McGraw-hill, 2001

    2001

  27. [27]

    Adjustment of an inverse matrix corresponding to a change in one element of a given matrix,

    J. Sherman and W. J. Morrison, “Adjustment of an inverse matrix corresponding to a change in one element of a given matrix,” Ann. Math. Statist. , vol. 21, no. 1, pp. 124–127, 1950

    1950

  28. [28]

    New edge-directed interpolati on,

    X. Li and M. T. Orchard, “New edge-directed interpolati on,” IEEE Trans. Image process. , vol. 10, no. 10, pp. 1521–1527, 2001

    2001

  29. [29]

    Fan and I

    J. Fan and I. Gijbels, Local Polynomial Modelling and Its Applications . Boca Raton, FL, USA: Routledge, 2018

    2018

  30. [30]

    Weickert, Anisotropic Diffusion in Image Processing

    J. Weickert, Anisotropic Diffusion in Image Processing . Stuttgart, Germany: Teubner Stuttgart, 1998, vol. 1

    1998

  31. [31]

    A brief description of natural neighbour in terpolation,

    R. Sibson, “A brief description of natural neighbour in terpolation,” Interpreting multivariate data , pp. 21–36, 1981

    1981

  32. [32]

    Kernel regressi on for image processing and reconstruction,

    H. Takeda, S. Farsiu, and P . Milanfar, “Kernel regressi on for image processing and reconstruction,” IEEE Trans. Image Process. , vol. 16, no. 2, pp. 349–366, 2007

    2007

  33. [33]

    Fast gradient-based algorith ms for con- strained total variation image denoising and deblurring pr oblems,

    A. Beck and M. Teboulle, “Fast gradient-based algorith ms for con- strained total variation image denoising and deblurring pr oblems,” IEEE Trans. Image process. , vol. 18, no. 11, pp. 2419–2434, Nov. 2009

    2009

  34. [34]

    An efficient augmen ted Lagrangian method with applications to total variation min imization,

    C. Li, W. Yin, H. Jiang, and Y . Zhang, “An efficient augmen ted Lagrangian method with applications to total variation min imization,” Comput. Optim. Appl. , vol. 56, no. 3, pp. 507–530, Dec. 2013

    2013

  35. [35]

    Successive linea r approxima- tion VBI for joint sparse signal recovery and dynamic grid pa rameters estimation,

    W. Xu, A. Liu, B. Zhou, and M.-j. Zhao, “Successive linea r approxima- tion VBI for joint sparse signal recovery and dynamic grid pa rameters estimation,” to appear in IEEE Trans. Wireless Commun. , 2025

    2025