Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Computational Imaging-Based ISAC Method with Large Pixel Division

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Replacing center-point propagation with a pixel-area integral lets millimeter-wave computational imaging use large pixels without the usual phase-error blowup, extending the method to large-scale sensing.

desk verdict The integral-form pixel model is a sensible and likely useful refinement for large-pixel ISAC, but the paper's own error analysis does not actually support the cancellation claim; the numerical results are suggestive, not conclusive. read the letter →

arxiv 2505.07355 v1 pith:OWTHEWVQ submitted 2025-05-12 eess.SP

Xin Tong , Zhaoyang Zhang , Zhaohui Yang , Yu Ge , Henk Wymeersch This is my paper · ORCID
classification eess.SP
keywords integratedsensingandcommunicationcomputationalimagingpixeldivisionphaseerrorcancellationcompressedmillimeter-wavegeneralizedapproximatemessagepassinglarge-scaleenvironment
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

Computational imaging in ISAC systems conventionally divides the environment into pixels and models each pixel's propagation by a single center-point distance. When pixels grow beyond a wavelength, that center-point approximation produces near-random phase errors and sensing fails. The paper proposes an integral-form model that averages the free-space Green's function over each pixel, which the authors show cancels the dominant phase error for extended objects. This makes it feasible to sense large-scale millimeter-wave environments with far fewer, larger pixels, reducing computational and antenna overhead. The claimed contribution is a first step toward applying computational imaging outside confined small-scale regions.

What carries the argument

The central object is the pixel-integrated propagation gain, defined in equation (2) as $\tilde{H}^{\mathrm{Tx}}_k(n_s,n_T) = \int_{x_0-l_s/2}^{x_0+l_s/2}\int_{y_0-w_s/2}^{y_0+w_s/2} \frac{\lambda_k}{4\pi d l_s w_s} e^{-2j\pi d/\lambda_k} \, dx\,dy$ with $d=\sqrt{(x_T-x)^2+(y_T-y)^2}$. This integral replaces the conventional center-point gain of equation (1), and it carries the argument by averaging the phase of the spherical wave over the entire pixel footprint, so that a target anywhere in the pixel is represented by a phase close to its own. The compressed sensing reconstruction, solved by the GAMP algorithm, then estimates the scattering coefficients from the jointly processed multi-carrier, multi-antenna measurements.

What would settle it

Run the proposed estimator on ray-traced or measured data from a scene with a strong second-bounce reflection off a large planar surface, or with a target whose reflectivity varies across a pixel; if reconstruction miss-detection and false-alarm rates rise well above the single-bounce uniform-pixel simulation, the integral-form cancellation is incomplete for that propagation regime.

Watch

Extended reading notes

Core claim

The paper argues that the severe sensing errors caused by large pixel division in traditional computational imaging are not inherent but arise from modeling each pixel as a point at its center. Replacing the discrete center-point propagation factor with a two-dimensional integral of the propagation phase over the pixel area removes the phase error that otherwise becomes effectively random when pixel size exceeds one wavelength. For planar targets that fill a pixel, the integral-form average distance converges to the true target distance, so the error drops sharply as the scatterer proportion within a pixel grows. The received signal model is then cast as a compressed sensing problem and solved with a Bernoulli-Gaussian sparse prior, yielding accurate environmental images even with pixel sizes of one meter at 30 GHz, where the baseline center-point method fails.

Load-bearing premise

The environment is modeled as a set of independent, uniformly scattering pixels whose received field is the single-bounce coherent sum of free-space waves integrated over each pixel; if significant multi-bounce, occlusion, or per-pixel reflectivity variation occurs, the claimed error cancellation does not carry over.

Editorial extensions

If this is right

  • Pixel sizes can be many wavelengths without the phase-error collapse that plagues center-point computational imaging, enabling a drastic reduction in the number of pixels and the associated sensing overhead.
  • Computational imaging becomes practical for large-scale millimeter-wave wireless environments, such as outdoor cellular scenes, where fine pixel grids would be computationally prohibitive.
  • The residual phase error for planar targets is governed by the fraction of a pixel that the target fills, so errors concentrate at object edges rather than throughout the image.
  • The method naturally inherits the compressed sensing framework, so sparsity-promoting reconstruction algorithms and their convergence guarantees apply unchanged.
  • The integral computation per pixel can be precomputed and parallelized, so the added modeling cost does not scale with reconstruction complexity.

Reading between the lines

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

  • The same integral-form cancellation should extend to 3D volumetric pixels, suggesting a direct route to large-scale volumetric ISAC imaging with manageable pixel counts.
  • Because the paper shows the advantage is specifically for extended surface targets and not point targets, the practical gain is best stated for planar or large scatterers; point-like targets still require fine pixels.
  • A natural testable extension is to use nonuniform or boundary-adaptive pixel meshes that align with object edges, which could push the residual edge error further down.
  • The model implicitly assumes single-bounce scattering; if multi-bounce or occlusion is significant, the integral-form averaging alone will not recover the sensing accuracy shown in the single-bounce simulations.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a computational imaging-based integrated sensing and communication (ISAC) method that aims to overcome the severe phase errors caused by large pixel division in millimeter-wave cellular scenarios. The key idea is to replace the conventional point-based propagation model (free-space gain evaluated at the pixel center, Eq. (1)) with an integral-form model that coherently integrates the scattered field over each pixel area (Eq. (2)). The environment sensing problem is reformulated as a compressed sensing recovery of the scattering coefficient vector, solved with the GAMP algorithm. The paper claims that the integral-form model cancels the large-pixel phase errors, and provides a performance analysis in Section IV plus numerical simulations in Section V, comparing against the baseline center-point model in terms of missed detection and false alarm rates.

Significance. If the central claim is correct, the approach would be a meaningful step toward extending computational imaging to large-scale wireless sensing, potentially allowing pixel sizes much larger than the wavelength and thereby reducing the number of unknowns and computational cost. The core idea is appealing and the simulation results are suggestive. However, the paper's analytical support is not aligned with its actual model: the error analysis in Section IV is based on distance differences (phase arguments) and average distances, while the proposed forward model in Eq. (2) is a coherent integral of rapidly varying complex exponentials. The phase and amplitude behavior of that integral is not captured by the provided metrics. Since the cancellation claim is the central contribution, the current manuscript leaves this point unestablished. If the analysis were redone for the true integral model, and if the simulation methodology were clarified, the paper would likely be a valuable contribution to computational imaging for ISAC.

major comments (3)
  1. [Section IV.B, Eqs. (11)-(14)] The error analysis for the proposed model does not describe the model in Eq. (2). Equations (13) and (14) define the phase error in terms of the difference between the actual distance d_t and the average distance d_p, i.e., 2π|d_t - d_p|/λ. But the proposed forward model coherently integrates λ/(4π d) e^(-j2π d/λ) over the pixel; the phase of this integral is not equal to the phase at the average distance when the phase varies by many cycles across the pixel. For a 1 m pixel at 30 GHz, the phase changes by hundreds of cycles, and the integral magnitude can be orders of magnitude smaller than the center-point value. Therefore the claim in Section IV.B.2 that d_p is close to d_t and makes the error as small as possible does not establish cancellation of the phase error of the actual model. The analysis needs to be reworked in terms of the complex integral, or the paper should explicitly label the distance-based analysis as a heuristic that is not a bound on the true model error.
  2. [Section III.A, footnote 2 and accompanying text] The paper states that errors resulting from pixel area exceeding the target area or uneven scattering primarily affect amplitude and will be left to future work. For large pixels (e.g., 1 m at 30 GHz, i.e., 100 wavelengths), the amplitude of the coherent integral in Eq. (2) can be much smaller than the point-model amplitude due to phase cancellation across the pixel. Such amplitude loss is not a minor effect: it directly degrades detection and compressed sensing reconstruction. The paper should either incorporate amplitude effects into the analysis or provide a quantitative argument that they are negligible in the scenarios considered.
  3. [Section V.A, simulation setup] The simulation methodology is not fully specified. The text says ray tracing from [16] generates the received signals, but it is unclear whether the ground-truth targets are point scatterers or extended surfaces. If the targets are point-like, then the proposed integral model in Eq. (2) integrates over pixels that contain only a sparse point, which is not the same as the planar-target analysis in Section IV.B; this would misrepresent the source of the improvement. The authors should clarify the target geometry used in the simulations, specify how the sensing matrix A is constructed (integration rules, pixel sizes, etc.), and ideally include a case with an extended surface target to validate the claimed cancellation under exactly the conditions analyzed.
minor comments (5)
  1. [Author list] The author list contains "ning ming" with a space, which appears to be a placeholder or an error; please check the author names.
  2. [Section II.A, footnote 1] The phrase "It is worthy noted" should be "It is worth noting".
  3. [Algorithm 1, line 6] The update for sigma_x_ns(i+1) uses sigma_r_ns(i), which is not defined; this is likely a typo for sigma_u_ns(i).
  4. [Eq. (6)] The constraint is written as ||H_k - ...|| but the vectorized notation introduced in Eq. (5) suggests the constraint should be on the stacked vector; please check the indexing consistency.
  5. [Figure 5] The horizontal axis label "Scatterer proportion" should be defined in the caption as (l_t w_t)/(l_s w_s) to match the text in Section IV.B.

Circularity Check

1 steps flagged · score 6.0 of 10

The analytical error cancellation in Section IV is definitional: e2,p is defined relative to the model's own pixel-averaged distance dp, so the claimed cancellation is built into the metric rather than derived from the coherent integral forward model.

  1. self definitional [Section IV-B-2, Eqs. (12)-(14)]
    "The proposed method effectively cancels the phase errors for planar targets. In (11), there is a certain difference between d0 calculated by the conventional algorithm and dt in the planar target model. The dp in (13) of the proposed model is close to dt and makes the error as small as possible, especially when the planar target completely fills the pixel."

    Equation (13) defines the proposed-model phase error e2,p as the pixel average of |dt(x,y)-dp|. Equation (12) defines dt as the average distance over the target patch and Eq. (14) defines dp as the average distance over the pixel; both are averages of the same distance function d(u,v). In the full-fill case highlighted by the paper, dt and dp are the same averaged quantity by construction, so e2,p vanishes. The claimed cancellation is therefore an identity embedded in the error metric, not a property derived from the coherent integral model (2), whose phase is the argument of ∫∫ (λ/(4πd ls ws)) e^{-2jπd/λ} dxdy. The analysis never computes that phase, so the central analytical support reduces to a definition.

full rationale

The forward model in Eq. (2) is an independent reformulation, the CS/GAMP reconstruction is standard, and the numerical results use an external ray-tracing simulator (Ref. [16]) as the benchmark, so the main numerical validation is not fitted to the proposed model. The self-citations to the authors' prior computational-imaging work are used only to motivate the conventional model and are not load-bearing for the new integral form. The significant defect is the analytical error analysis: the proposed-model 'phase error' is defined relative to dp, the pixel-averaged distance, so that the claimed cancellation for planar targets follows from the definition of the metric rather than from the coherent integral in Eq. (2). This makes the analytical cancellation claim partial circular even though the numerical results provide independent evidence.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method relies on the single-bounce uniform-pixel propagation model, sparsity assumptions, and an unverified analytical error analysis. No new physical entities are introduced. The main free parameters are reconstruction-algorithm settings whose values are not specified.

free parameters (4)
  • epsilon (CS slack variable) = not specified
    Appears in the compressed sensing constraint (6); must be chosen to balance model error and noise, but no value or selection rule is given.
  • Bernoulli-Gaussian prior parameters alpha, theta_x, sigma_x = not specified
    Appear in equation (7); the GAMP algorithm requires them, and the paper does not state how they are set in simulations.
  • Noise variances sigma_w and sigma_z in GAMP output = not specified
    Used in equations (18) and (19); values are required to run the algorithm but are not provided.
  • Detection threshold for MD and FA = 0.5
    Section V.A sets threshold 0.5 on the normalized scattering coefficient to decide presence; results depend on this choice.
assumptions (5)
  • domain assumption Single-bounce, free-space propagation with no occlusion, diffraction, or multiple scattering.
    Used to construct the measurement model (3)-(5); real mmWave environments violate this.
  • domain assumption Uniform scattering coefficient over each pixel.
    The integral model (2) assumes uniformity; uneven surfaces or point targets are acknowledged as error sources in Section IV and left to future work.
  • domain assumption Perfect LOS estimation and cancellation.
    Section III.B removes HLOS using a point-to-point model; residual LOS errors would bias the compressed sensing measurement.
  • domain assumption Sparse environment with Bernoulli-Gaussian prior.
    Required to set up the compressed sensing problem (6) and the GAMP prior (7).
  • standard math Numerical integration of (2) is accurate.
    The paper relies on numerical quadrature for the integral model, but details of the quadrature are not given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Computational Imaging-Based ISAC Method with Large Pixel Division." pith.science (2026). https://pith.science/paper/OWTHEWVQ

@misc{pith2026250507355,
  author       = {Pith},
  title        = {Pith review of: Computational Imaging-Based ISAC Method with Large Pixel Division},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OWTHEWVQ}},
  note         = {Machine review of arXiv:2505.07355}
}
read the original abstract

One of the key points in designing an integrated sensing and communication (ISAC) system using computational imaging is the division size of imaging pixels. If the size is too small, it leads to a high number of pixels that need processing. On the contrary, it usually causes large processing errors since each pixel is no longer uniformly coherent. In this paper, a novel method is proposed to address such a problem in environment sensing in millimeter-wave wireless cellular networks, which effectively cancels the severe errors caused by large pixel division as in conventional computational imaging algorithms. To this end, a novel computational imaging model in an integral form is introduced, which leverages the continuous characteristics of object surfaces in the environment and takes into account the different phases associated with the different parts of the pixel. The proposed algorithm extends computational imaging to large wireless communication scenarios for the first time. The performance of the proposed method is then analyzed, and extensive numerical results verify its effectiveness.

Figures

Figures reproduced from arXiv: 2505.07355 by the authors.

Figure 1
Figure 1. The considered 2D ISAC scenario and pixel division [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Two cases of pixel division errors caused by incomplete [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The comparison sensing results between the baseline method (top row) and proposed method (bottom row). [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The relationship between the pixel size and environment sensing performance. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The relationship between the scatterer proportion and [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Near-Field Integrated Imaging and Communication in Distributed MIMO Networks

    eess.SP 2025-08 unverdicted novelty 5.0 of 10

    Distributed MIMO networks can image near-field targets by converting spatial-domain signals to the Fourier domain, using RMA for small objects and sparse Bayesian learning for large scenes.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages · cited by 1 Pith paper

  1. [11]

    X. Tong et al., “Joint multi-user communication and sensing exploiting Algorithm 1 The GAMP Algorithm [15] Input: Given the estimated propagation gain ˆH∈ CM×1 and the measurement matrix A∈ CM×Ns. 1: Initialization: Set environmental information prior param- eter p. Defined gin (·) and gout (·) from (15), (18). Set i = 0, ˆs (−1) = 0 , ˆzm(i) = 0 , ˆxns (...

    work page 2021

  2. [12]

    Environment sensing considering the occlusion effect: A multi-view approach,

    X. Tong et al., “Environment sensing considering the occlusion effect: A multi-view approach,” IEEE Transactions on Signal Processing , vol. 70, pp. 3598–3615, 2022

    work page 2022

  3. [16]

    Physics-inspired target shape detection and reconstruc- tion in mmWave communication systems,

    Z. Xing et al., “Physics-inspired target shape detection and reconstruc- tion in mmWave communication systems,” in IEEE Global Communi- cations Conference, 2023, pp. 3910–3915

    work page 2023

  4. [1]

    A vision of 6G wireless systems: Applications, trends, technologies, and open research problems,

    W. Saad et al., “A vision of 6G wireless systems: Applications, trends, technologies, and open research problems,” IEEE Network , vol. 34, no. 3, pp. 134–142, 2020

    work page 2020

  5. [2]

    Exploiting diffuse multipath in 5G SLAM,

    Y . Ge et al., “Exploiting diffuse multipath in 5G SLAM,” in GLOBECOM 2020 - 2020 IEEE Global Communications Conference , 2020, pp. 1–6

    work page 2020

  6. [3]

    Edge learning for B5G networks with distributed signal processing: Semantic communication, edge computing, and wireless sensing,

    W. Xu et al., “Edge learning for B5G networks with distributed signal processing: Semantic communication, edge computing, and wireless sensing,” IEEE Journal of Selected Topics in Signal Processing , vol. 17, no. 1, pp. 9–39, 2023. 10-2 10-1 100 Pixel size (m) 10-4 10-3 10-2 10-1 100 MD SNR = 5dB, baseline SNR = 10dB, baseline SNR = 20dB, baseline SNR = 5...

    work page 2023

  7. [4]

    Integrated sensing and communications: Towards dual- functional wireless networks for 6G and beyond,

    F. Liu et al., “Integrated sensing and communications: Towards dual- functional wireless networks for 6G and beyond,” IEEE Journal on Selected Areas in Communications , 2022

    work page 2022

  8. [5]

    A computationally efficient EK-PMBM filter for bistatic mmwave radio SLAM,

    Y . Ge et al., “A computationally efficient EK-PMBM filter for bistatic mmwave radio SLAM,” IEEE Journal on Selected Areas in Communi- cations, vol. 40, no. 7, pp. 2179–2192, 2022

    work page 2022

Show all 16 references
  1. [6]

    Multi-view sensing for wireless communications: Archi- tectures, designs, and opportunities,

    X. Tong et al., “Multi-view sensing for wireless communications: Archi- tectures, designs, and opportunities,” IEEE Communications Magazine , vol. 61, no. 5, pp. 40–46, 2023

    2023

  2. [7]

    3D environment sensing with channel state infor- mation based on computational imaging,

    Y . Zhang et al., “3D environment sensing with channel state infor- mation based on computational imaging,” in 2022 IEEE International Conference on Communications Workshops (ICC Workshops) , 2022, pp. 842–847

    2022

  3. [8]

    Holographic imaging with XL-MIMO and RIS: Illumination and reflection design,

    G. Torcolacci et al., “Holographic imaging with XL-MIMO and RIS: Illumination and reflection design,” IEEE Journal of Selected Topics in Signal Processing, pp. 1–16, 2024

    2024

  4. [9]

    Radar coincidence imaging with random microwave source,

    S. Zhu et al., “Radar coincidence imaging with random microwave source,” IEEE Antennas and Wireless Propagation Letters , vol. 14, pp. 1239–1242, 2015

    2015

  5. [10]

    3D computational imaging with single-pixel detectors,

    B. Sun et al., “3D computational imaging with single-pixel detectors,” Science, vol. 340, no. 6134, pp. 844–847, 2013

    2013

  6. [13]

    Compressed sensing,

    D. L. Donoho, “Compressed sensing,” IEEE Transactions on Informa- tion Theory , vol. 52, no. 4, pp. 1289–1306, 2006

    2006

  7. [14]

    On channel estimation in OFDM systems,

    J.-J. van de Beek et al., “On channel estimation in OFDM systems,” in IEEE 45th V ehicular Technology Conference., 1995, vol. 2, pp. 815–819

    1995

  8. [15]

    Generalized approximate message passing for estimation with random linear mixing,

    S. Rangan, “Generalized approximate message passing for estimation with random linear mixing,” in 2011 IEEE International Symposium on Information Theory Proceedings , 2011, pp. 2168–2172

    2011

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.