Pith. sign in

REVIEW 4 major objections 7 minor 55 references

Hyperspectral Dual-Comb Compressive Imaging for Minimally-Invasive Video-Rate Endomicroscopy

T0 review · 4 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that hyperspectral dual-comb compressive imaging can deliver video-rate, high-fidelity images through a single-core optical fiber and a single-pixel detector, at sampling ratios as low as 0.3%, by encoding the image into…

desk verdict A real experimental proof-of-concept for dual-comb single-pixel ghost imaging through a fiber, but the endomicroscopy claim rests on a fixed-fiber calibration matrix that no experiment bends; send to review but push for tempered claims and a bending test. read the letter →

arxiv 2507.04157 v1 pith:GUYQNBG3 submitted 2025-07-05 physics.optics

Myoung-Gyun Suh , David Dang , Maodong Gao , Yucheng Jin , Byoung Jun Park , Beyonce Hu , Wilton J.M. Kort-Kamp , Ho Wai (Howard) Lee This is my paper · ORCID
classification physics.optics
keywords dual-combinterferometryhyperspectralimagingcomputationalghostsingle-pixelspecklepatternstransformerimagereconstructionendomicroscopycompressive
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 tries to establish a new architecture for minimally invasive endomicroscopy: hyperspectral dual-comb compressive imaging. Two optical frequency combs with slightly different line spacings generate many uncorrelated speckle patterns, one per comb line, which are projected through a single-core fiber onto the target. A single-pixel detector collects the total transmitted light, and a Fourier transform of the dual-comb interference signal yields one bucket measurement per comb line, so the imaging matrix is applied optically in parallel and no spatial or spectral scanning is needed. A transformer model, Ghost-GPT, then reconstructs a 256x256 image from these sparse measurements, achieving SSIM above 0.6 and MSE below 0.05 at a sampling ratio of 0.3%, with a reconstruction time of 8.39 ms. If correct, the result means endoscopes could shrink to a fiber plus one detector while retaining real-time imaging.

What carries the argument

The central mechanism is an optical multiply-accumulate: each comb line is mapped to a distinct speckle pattern, the target multiplies these patterns, and a single-pixel detector sums the transmitted light, so the hardware computes $y = H x$ in parallel. Two electro-optic combs whose line spacings differ by a small offset convert the bucket sums into a radio-frequency comb, so a fast Fourier transform of one interferogram recovers all measurements simultaneously. On the reconstruction side, Ghost-GPT, a twelve-block transformer, concatenates each flattened speckle pattern with its measured bucket value to form a token and outputs the reconstructed 256x256 image.

What would settle it

Mount the multimode fiber on a flexible arm, image the same target while bending the fiber into gentle curves, and compare the reconstructed SSIM and MSE against the fixed-fiber result; a clear drop would show that the calibration matrix does not survive bending.

Watch

Extended reading notes

Core claim

The central claim is that combining dual-comb interferometry with compressive ghost imaging lets a single-core fiber and a single-pixel detector act as a complete imaging system, with the encoding done optically and the decoding done by a transformer network. Each comb line carries an independent two-dimensional speckle pattern, so the object multiplies the stored pattern set and the bucket detector records the sum; the dual-comb readout maps each comb line to a distinct radio-frequency tone, giving parallel, scan-free acquisition. The authors demonstrate this on static USAF resolution targets and on a target moving at 2.4 mm/s, reconstructing images at 60 Hz with SSIM greater than 0.6 and MSE less than 0.05 at a sampling ratio of 0.3%. They state that the frame rate is set by the comb repetition-rate difference, not by computation, and that denser combs could enable higher-definition video-rate imaging.

Load-bearing premise

The load-bearing premise is that the speckle pattern matrix measured once with a 2D camera during calibration remains exactly valid during imaging; if the fiber bends or the setup drifts, the stored patterns no longer match the actual illumination and the reconstruction degrades.

Editorial extensions

If this is right

  • Endomicroscope front ends could be built from a single-core fiber and a single-pixel detector, eliminating scanning mechanisms and multi-element optics.
  • Frame rate is limited by the comb spacing difference, not by reconstruction speed; at a difference of 3 kHz the authors demonstrate 60 Hz video with clear motion capture, and up to 1 kHz with some quality loss.
  • With denser combs, a 1 GHz spacing across a 100 nm span could support an effective fill rate of 12 gigapixels per second, enabling HD video-rate imaging.
  • Ghost-GPT outperforms classical algorithms such as differential ghost imaging, pseudoinverse, and FISTA on fidelity, and reconstructs roughly 430 times faster than FISTA.
  • The calibrated speckle pattern set remains stable for hours in the fixed-fiber laboratory setup, allowing repeated experiments without re-calibration.

Reading between the lines

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

  • A step the paper leaves implicit: a clinical probe will need a bending-robust calibration strategy, because the fixed-fiber speckle matrix is the load-bearing assumption and bending the fiber changes the speckle patterns.
  • Because the transformer was trained on synthetic handwritten-character images, deployment on biological tissue will likely require fine-tuning on clinical data, a domain gap the paper explicitly acknowledges.
  • The demonstrated resolution of about 0.077 mm is set by speckle grain size; engineering speckle patterns with different grain sizes across comb lines could push resolution well below the current limit.
  • The same architecture naturally extends to time-domain compressive sensing, which could simultaneously compress spatial and temporal information into a four-dimensional hyperspectral imaging modality.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper presents a hyperspectral dual-comb compressive imaging system in which wavelength-multiplexed speckle patterns from a 200-micron multimode fiber illuminate a transmissive target, and a single-pixel detector records dual-comb interferograms. A fast Fourier transform yields bucket-sum measurements, and a transformer-based model called Ghost-GPT reconstructs 256x256 images from roughly 188 measurements, corresponding to a sampling ratio of about 0.3%. Static USAF targets and a target translating at 2.4 mm/s are imaged, with claimed video-rate operation up to 60 Hz and, at reduced quality, up to 1000 Hz. The authors position the work as a step toward minimally invasive, video-rate endomicroscopy with drastically simplified front-end hardware.

Significance. If the claims hold, the combination of dual-comb illumination, single-pixel detection, and transformer-based reconstruction is a novel and potentially useful architecture for scan-free imaging through a single optical fiber, with the demonstrated forward-model agreement (SD 0.008 in Fig. 3a) lending credibility to the physical model. The study is not circular in the narrow sense: Ghost-GPT is trained on synthetic buckets generated from the measured speckle matrix and then evaluated on unseen experimental targets. The supplementary material is unusually detailed, including architecture, hyperparameter sweeps, noise robustness tests, and an explicit limitations section. However, the headline endomicroscopy claim is not yet supported by the evidence: the fiber is fixed during calibration and imaging, the resolution figure is obtained from simulation only, and the video-rate demonstration is qualitative.

major comments (4)
  1. [Methods, Experimental Setup; SI Limitations] The endomicroscopy claim depends on the speckle matrix H remaining valid during imaging, but the Methods state that "the MMF was fixed to the optical table," and the SI Limitations concede that "any deviations in experimental conditions may lead to distribution shift." An endomicroscope requires the fiber to bend and move during use, which changes the modal interference and hence H, invalidating the reconstruction for both the classical algorithms and Ghost-GPT. No experiment with fiber motion or bending is reported, so the paper's central application claim is unverified at its most load-bearing point. The authors should either demonstrate robustness to fiber perturbation or explicitly reframe the work as a fixed-fiber proof-of-principle and move the endomicroscopy claim to future work.
  2. [SI Section G; main text after Fig. 3d] The claimed resolution of approximately 0.077 mm is based solely on the simulated resolution target in Fig. S7, not on experimental images. The experimental Ghost-GPT reconstructions in Fig. 3d are of large features (group 0, 1.12-1.41 lp/mm), and the only experimental group-2 images are pseudoinverse reconstructions in Fig. 3c. Because the reconstruction model is trained on the specific speckle matrix and could in principle rely on training-set priors rather than true optical resolution, the text should clearly state that the 0.077 mm figure is a simulated estimate, or an experimental resolution test with Ghost-GPT should be provided.
  3. [Table I and main text after Fig. 3d] The main text claims "an MSE less than 0.05," but Table I reports MSE = 0.058 for the Number 2 target, which is greater than 0.05. Additionally, each row of Table I corresponds to a single experimental image with no replicate measurements or error bars, so the quantitative comparison among DGI, PI, FISTA, and Ghost-GPT has no statistical support. The authors should correct the numerical inconsistency and either provide replicate experimental images with variability estimates or label the table entries as single-trial results.
  4. [Figure 4 and video-rate discussion] The moving-target demonstration is only qualitative. No MSE, SSIM, or other quantitative metric is reported for the reconstructed frames at 60 Hz, and the higher-frame-rate results in Fig. 4c are shown without any fidelity measure. Since video-rate operation is a central claim, the 60 Hz reconstruction quality should be quantified on static targets under the same dual-comb conditions (Δf_FSR = 3 kHz), or the claim should be softened to "qualitative video-rate demonstration."
minor comments (7)
  1. [SI Section I, Eq. (1)] The text after Eq. (1) says "where K is the number of structured light patterns used," but the equation defines the sampling ratio using M; the variable names should be made consistent.
  2. [SI Section III.A] The sentence "a learning rate 184 of 0.0003" contains a stray number "184" and should read "a learning rate of 0.0003."
  3. [Figure 4c] The axis labels and frame-rate annotations in Fig. 4c are difficult to read at the printed size, and the panel does not include a scale bar or any quantitative image-quality metric.
  4. [Abstract and main text] The phrase "zero-dimensional hardware" and the claim of "completely eliminating" scanning are overstated, because the system still requires two EO combs, a Waveshaper, EDFAs, and a 2D camera for calibration; the compactness claim applies to the distal front-end only.
  5. [Data and Code Availability] The statement that data and code are available "from the corresponding authors upon reasonable request" is not sufficient for reproducibility; the authors should deposit the reconstruction code and representative datasets in a public repository.
  6. [References] Reference [23] is an incomplete web citation with only a URL; it should include the author, title, and access date.
  7. [SI Section G] The resolution analysis should clarify whether the reported "5-pixel gap" is the smallest resolved feature and how the conversion from pixels to physical millimeters is performed; a gap between bars is not the same as a line-pair resolution specification.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the forward model y = Hx is a standard measurement model, H is independently calibrated with a camera, and Ghost-GPT is evaluated on unseen experimental targets rather than fitted to them.

full rationale

The derivation chain is self-contained and non-circular. The speckle matrix H is measured independently with a 2D InGaAs camera during calibration ('The speckle pattern set (H) without the object is separately recorded using a 2D InGaAs camera by selecting individual comb lines with the Waveshaper'), while the bucket signals y are acquired separately through a single-pixel photodetector during imaging. The reconstruction therefore solves a genuine inverse problem y = Hx rather than recovering a quantity that was used to define the forward model. Ghost-GPT is trained on synthetic bucket sums generated by applying the same measured H to MNIST and Omniglot images, which is a standard supervised learning protocol, and its reported SSIM/MSE values are measured on experimental USAF targets that were not part of the training set. The agreement between simulated and experimental bucket sums in Figure 3a is a forward-model consistency check, not a circular input to the reconstruction claims. The paper's self-citations (e.g., Ref. 11 on dual-comb imaging and Ref. 47 on hyperspectral MAC operations) provide background and context, but no load-bearing uniqueness theorem is imported from the authors' prior work, no fitted parameter is renamed as a prediction, and no ansatz is smuggled in solely through a self-citation. The acknowledged limitation that calibration speckle stability under fiber motion is untested is a robustness/correctness concern for the endomicroscopy claim, not a circularity in the derivation.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The reconstruction and experimental claims rest on the stability and representativeness of the measured speckle matrix, the linear multiplicative forward model, and the transfer of a deep network trained on synthetic data. These are reasonable for a bench-top proof of concept but are unproven for flexible-fiber endomicroscopy.

free parameters (3)
  • TV loss weight beta = 0.4
    Chosen by hyperparameter sweep on a noisy validation dataset (SI Section III H, Table S2) to maximize SSIM; affects reconstruction smoothing and reported fidelity.
  • Ghost-GPT architecture hyperparameters = heads 8-32, embedding dim 1-129, AdamW lr 3e-4, weight decay 0.1
    Selected by validation sweeps in SI Section III E; these determine the reconstruction quality reported in Table I and Table S1.
  • Number of comb lines used for reconstruction = 150-200 (188 reported in SI)
    Sampling ratio 0.3% depends on this number; it is set by available comb lines and Waveshaper resolution, not optimized.
assumptions (6)
  • domain assumption The target is a planar multiplicative mask: bucket sum y(m) = sum_{i,j} S(m)_{i,j} x_{i,j}.
    Assumed in the forward model, Eq. (2) in SI; holds for a thin transmissive USAF target but not for scattering tissue or 3D objects.
  • domain assumption Speckle patterns from different comb lines are uncorrelated and form a valid sensing basis.
    Supported by measured Pearson correlations (Fig. 2c), but residual correlations near pattern edges are acknowledged and could limit conditioning of H.
  • domain assumption The MMF speckle patterns are stable between calibration and imaging.
    Methods: 'MMF was fixed to the optical table... stable enough to support multiple imaging experiments over several hours.' Critical for treating H as known.
  • domain assumption Dual-comb interferometry maps each optical comb line to a resolved RF comb line with negligible crosstalk.
    The RF comb FFT readout (Fig. 2d) is assumed to yield one bucket value per speckle pattern; this requires N x delta_f_FSR < photodetector bandwidth and mutual coherence.
  • ad hoc to paper The transformer model trained on synthetic MNIST/Omniglot buckets generalizes to experimental data.
    The model is trained on synthetic data generated from the same H; experimental targets resemble the training distribution; SI limitations note the domain gap.
  • standard math Standard linear algebra and compressed sensing results (pseudoinverse, FISTA) are used without proof.
    Uses y = Psi x and pseudoinverse/FISTA; these are standard results, not questioned.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hyperspectral Dual-Comb Compressive Imaging for Minimally-Invasive Video-Rate Endomicroscopy." pith.science (2026). https://pith.science/paper/GUYQNBG3

@misc{pith2026250704157,
  author       = {Pith},
  title        = {Pith review of: Hyperspectral Dual-Comb Compressive Imaging for Minimally-Invasive Video-Rate Endomicroscopy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GUYQNBG3}},
  note         = {Machine review of arXiv:2507.04157}
}
read the original abstract

Endoscopic imaging is essential for real-time visualization of internal organs, yet conventional systems remain bulky, complex, and expensive due to their reliance on large, multi-element optical components. This limits their accessibility to delicate or constrained anatomical regions. Achieving real-time, high-resolution endomicroscopy using compact, low-cost hardware at the hundred-micron scale remains an unsolved challenge. Optical fibers offer a promising route toward miniaturization by providing sub-millimeter-scale imaging channels; however, existing fiber-based methods typically rely on raster scanning or multicore bundles, which limit the resolution and imaging speed. In this work, we overcome these limitations by integrating dual-comb interferometry with compressive ghost imaging and advanced computational reconstruction. Our technique, hyperspectral dual-comb compressive imaging, utilizes optical frequency combs to generate wavelength-multiplexed speckle patterns that are delivered through a single-core fiber and detected by a single-pixel photodetector. This parallel speckle illumination and detection enable snapshot compression and acquisition of image information using zero-dimensional hardware, completely eliminating the need for both spatial and spectral scanning. To decode these highly compressed signals, we develop a transformer-based deep learning model capable of rapid, high-fidelity image reconstruction at extremely low sampling ratios. This approach significantly outperforms classical ghost imaging methods in both speed and accuracy, achieving video-rate imaging with a dramatically simplified optical front-end. Our results represent a major advance toward minimally invasive, cost-effective endomicroscopy and provide a generalizable platform for optical sensing in applications where hardware constraints are critical.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages

  1. [1]

    & author Forde, K

    author Berci, G. & author Forde, K. A. title History of endoscopy . journal Surgical endoscopy volume 14 , pages 5--15 ( year 2000 )

    work page 2000

  2. [2]

    author Boese, A. et al. title Endoscopic imaging technology today . journal Diagnostics volume 12 , pages 1262 ( year 2022 )

    work page 2022

  3. [3]

    author Kiesslich, R. et al. title Confocal laser endoscopy for diagnosing intraepithelial neoplasias and colorectal cancer in vivo . journal Gastroenterology volume 127 , pages 706--713 ( year 2004 )

    work page 2004

  4. [4]

    , author Kozarek, R

    author Urayama, S. , author Kozarek, R. & author Raltz, S. title Evaluation of per-procedure equipment costs in an outpatient endoscopy center . journal Gastrointestinal endoscopy volume 44 , pages 129--132 ( year 1996 )

    work page 1996

  5. [5]

    author Ren, H. et al. title An achromatic metafiber for focusing and imaging across the entire telecommunication range . journal nature communications volume 13 , pages 4183 ( year 2022 )

    work page 2022

  6. [6]

    author Porat, A. et al. title Widefield lensless imaging through a fiber bundle via speckle correlations . journal Optics express volume 24 , pages 16835--16855 ( year 2016 )

    work page 2016

  7. [7]

    author Lee, M. et al. title Confocal laser endomicroscope with distal mems scanner for real-time histopathology . journal Scientific Reports volume 12 , pages 20155 ( year 2022 )

    work page 2022

  8. [8]

    author Lee, C. M. , author Engelbrecht, C. J. , author Soper, T. D. , author Helmchen, F. & author Seibel, E. J. title Scanning fiber endoscopy with highly flexible, 1 mm catheterscopes for wide-field, full-color imaging . journal Journal of biophotonics volume 3 , pages 385--407 ( year 2010 )

    work page 2010

Show all 55 references
  1. [9]

    author Choi, W. et al. title Flexible-type ultrathin holographic endoscope for microscopic imaging of unstained biological tissues . journal Nature communications volume 13 , pages 4469 ( year 2022 )

    2022

  2. [10]

    , author Newbury, N

    author Coddington, I. , author Newbury, N. & author Swann, W. title Dual-comb spectroscopy . journal Optica volume 3 , pages 414--426 ( year 2016 )

    2016

  3. [11]

    , author Suh, M.-G

    author Bao, C. , author Suh, M.-G. & author Vahala, K. title Microresonator soliton dual-comb imaging https://opg.optica.org/optica/fulltext.cfm?uri=optica-6-9-1110&id=417129 . journal Optica volume 6 , pages 1110--1116 ( year 2019 )

    2019

  4. [12]

    , author Wang, Z

    author Vicentini, E. , author Wang, Z. , author Van Gasse, K. , author H \"a nsch, T. W. & author Picqu \'e , N. title Dual-comb hyperspectral digital holography . journal Nature Photonics volume 15 , pages 890--894 ( year 2021 )

    2021

  5. [13]

    , author Bromberg, Y

    author Katz, O. , author Bromberg, Y. & author Silberberg, Y. title Compressive ghost imaging . journal Applied Physics Letters volume 95 ( year 2009 )

    2009

  6. [14]

    author Padgett, M. J. & author Boyd, R. W. title An introduction to ghost imaging: quantum and classical . journal Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences volume 375 , pages 20160233 ( year 2017 )

    2017

  7. [15]

    & author Baumann, E

    author Fortier, T. & author Baumann, E. title 20 years of developments in optical frequency comb technology and applications . journal Communications Physics volume 2 , pages 153 ( year 2019 )

    2019

  8. [16]

    author Vaswani, A. et al. title Attention is all you need . journal Advances in neural information processing systems volume 30 ( year 2017 )

    2017

  9. [17]

    , author Nie, X

    author Ren, W. , author Nie, X. , author Peng, T. & author Scully, M. O. title Ghost translation: an end-to-end ghost imaging approach based on the transformer network . journal Optics Express volume 30 , pages 47921--47932 ( year 2022 )

    2022

  10. [18]

    author Chen, Y. et al. title Large model enhanced computational ghost imaging . journal arXiv preprint arXiv:2503.08710 ( year 2025 )

    2025 arXiv

  11. [19]

    & author Davies, M

    author Blumensath, T. & author Davies, M. E. title Iterative hard thresholding for compressed sensing . journal Applied and computational harmonic analysis volume 27 , pages 265--274 ( year 2009 )

    2009

  12. [20]

    & author Teboulle, M

    author Beck, A. & author Teboulle, M. title A fast iterative shrinkage-thresholding algorithm for linear inverse problems . journal SIAM journal on imaging sciences volume 2 , pages 183--202 ( year 2009 )

    2009

  13. [21]

    author Boyd, S. et al. title Distributed optimization and statistical learning via the alternating direction method of multipliers . journal Foundations and Trends in Machine learning volume 3 , pages 1--122 ( year 2011 )

    2011

  14. [22]

    author Kim, Y. et al. title A 1/2.8-inch 24mpixel cmos image sensor with 0.9 m unit pixels separated by full-depth deep-trench isolation . In booktitle 2018 IEEE International Solid-State Circuits Conference-(ISSCC) , pages 84--86 ( organization IEEE , year 2018 )

    2018

  15. [23]

    note Https://www.ovt.com/wp-content/uploads/2023/02/OH0TA-PB-v1.1-WEB.pdf

    2023

  16. [24]

    author Turtaev, S. et al. title High-fidelity multimode fibre-based endoscopy for deep brain in vivo imaging . journal Light: Science & Applications volume 7 , pages 92 ( year 2018 )

    2018

  17. [25]

    author Gmitro, A. F. & author Aziz, D. title Confocal microscopy through a fiber-optic imaging bundle . journal Optics letters volume 18 , pages 565--567 ( year 1993 )

    1993

  18. [26]

    author Jabbour, J. M. , author Saldua, M. A. , author Bixler, J. N. & author Maitland, K. C. title Confocal endomicroscopy: instrumentation and medical applications . journal Annals of biomedical engineering volume 40 , pages 378--397 ( year 2012 )

    2012

  19. [27]

    author Flusberg, B. A. et al. title Fiber-optic fluorescence imaging . journal Nature methods volume 2 , pages 941--950 ( year 2005 )

    2005

  20. [28]

    , author Kuschmierz, R

    author Sun, J. , author Kuschmierz, R. , author Katz, O. , author Koukourakis, N. & author Czarske, J. W. title Lensless fiber endomicroscopy in biomedicine . journal PhotoniX volume 5 , pages 18 ( year 2024 )

    2024

  21. [29]

    , author Scharf, E

    author Kuschmierz, R. , author Scharf, E. , author Orteg \'o n-Gonz \'a lez, D. F. , author Glosemeyer, T. & author Czarske, J. W. title Ultra-thin 3d lensless fiber endoscopy using diffractive optical elements and deep neural networks . journal Light: Advanced Manufacturing v...

    2021

  22. [30]

    author Fr \"o ch, J. E. et al. title Real time full-color imaging in a meta-optical fiber endoscope . journal elight volume 3 , pages 13 ( year 2023 )

    2023

  23. [31]

    author Hase, E. et al. title Scan-less confocal phase imaging based on dual-comb microscopy . journal Optica volume 5 , pages 634--643 ( year 2018 )

    2018

  24. [32]

    , author Khan, F

    author Mart \' n-Mateos, P. , author Khan, F. U. & author Bonilla-Manrique, O. E. title Direct hyperspectral dual-comb imaging . journal Optica volume 7 , pages 199--202 ( year 2020 )

    2020

  25. [33]

    author Sterczewski, L. A. et al. title Terahertz hyperspectral imaging with dual chip-scale combs . journal Optica volume 6 , pages 766--771 ( year 2019 )

    2019

  26. [34]

    author Giorgetta, F. R. et al. title Broadband dual-comb hyperspectral imaging and adaptable spectroscopy with programmable frequency combs . journal APL Photonics volume 9 ( year 2024 )

    2024

  27. [35]

    author Diddams, S. A. , author Hollberg, L. & author Mbele, V. title Molecular fingerprinting with the resolved modes of a femtosecond laser frequency comb . journal Nature volume 445 , pages 627--630 ( year 2007 )

    2007

  28. [36]

    , author Newbury, N

    author Coddington, I. , author Newbury, N. R. & author Swann, W. C. title Dual-comb spectroscopy . journal Optica volume 3 , pages 414--426 ( year 2016 )

    2016

  29. [37]

    , author Yang, Q.-F

    author Suh, M.-G. , author Yang, Q.-F. , author Yang, K. Y. , author Yi, X. & author Vahala, K. J. title Microresonator soliton dual-comb spectroscopy . journal Science volume 354 , pages 600--603 ( year 2016 )

    2016

  30. [38]

    author Suh, M.-G. et al. title Searching for exoplanets using a microresonator astrocomb . journal Nature photonics volume 13 , pages 25--30 ( year 2019 )

    2019

  31. [39]

    author Ideguchi, T. et al. title Coherent raman spectro-imaging with laser frequency combs . journal Nature volume 502 , pages 355--358 ( year 2013 )

    2013

  32. [40]

    & author H \"a nsch, T

    author Picqu \'e , N. & author H \"a nsch, T. W. title Frequency comb spectroscopy . journal Nature Photonics volume 13 , pages 146--157 ( year 2019 )

    2019

  33. [41]

    author Li, D. et al. title Enhancing robustness of ghost imaging against environment noise via cross-correlation in time domain . journal Optics Express volume 29 , pages 31068--31077 ( year 2021 )

    2021

  34. [42]

    author Liu, X. et al. title Photon-limited single-pixel imaging . journal Optics express volume 28 , pages 8132--8144 ( year 2020 )

    2020

  35. [43]

    , author Zhao, M

    author Li, F. , author Zhao, M. , author Tian, Z. , author Willomitzer, F. & author Cossairt, O. title Compressive ghost imaging through scattering media with deep learning . journal Optics Express volume 28 , pages 17395--17408 ( year 2020 )

    2020

  36. [44]

    , author Cao, J

    author Lin, L.-X. , author Cao, J. , author Zhou, D. , author Cui, H. & author Hao, Q. title Ghost imaging through scattering medium by utilizing scattered light . journal Optics Express volume 30 , pages 11243--11253 ( year 2022 )

    2022

  37. [45]

    title An introduction to computational ghost imaging with example code

    author Don, M. title An introduction to computational ghost imaging with example code . journal CCDC Army Research Laboratory: Aberdeen Proving Ground, MD, USA ( year 2019 )

    2019

  38. [46]

    , author Tran, T

    author Kilic, V. , author Tran, T. D. & author Foster, M. A. title Compressed sensing in photonics: tutorial . journal Journal of the Optical Society of America B volume 40 , pages 28--52 ( year 2022 )

    2022

  39. [47]

    author Latifpour, M. H. , author Park, B. J. , author Yamamoto, Y. & author Suh, M.-G. title Hyperspectral in-memory computing with optical frequency combs and programmable optical memories . journal Optica volume 11 , pages 932--939 ( year 2024 )

    2024

  40. [48]

    title Imaging reconstruction comparison of different ghost imaging algorithms

    author Liu, H.-C. title Imaging reconstruction comparison of different ghost imaging algorithms . journal Scientific Reports volume 10 , pages 14626 ( year 2020 )

    2020

  41. [49]

    , author Alam, M

    author Redding, B. , author Alam, M. , author Seifert, M. & author Cao, H. title High-resolution and broadband all-fiber spectrometers . journal Optica volume 1 , pages 175--180 ( year 2014 )

    2014

  42. [50]

    , author Khan, S

    author Facchin, M. , author Khan, S. N. , author Dholakia, K. & author Bruce, G. D. title Determining intrinsic sensitivity and the role of multiple scattering in speckle metrology . journal Nature Reviews Physics volume 6 , pages 500--508 ( year 2024 )

    2024

  43. [51]

    author Baumann, E. et al. title Comb-calibrated laser ranging for three-dimensional surface profiling with micrometer-level precision at a distance . journal Optics express volume 22 , pages 24914--24928 ( year 2014 )

    2014

  44. [52]

    author Herman, D. I. et al. title Squeezed dual-comb spectroscopy . journal Science pages eads6292 ( year 2025 )

    2025

  45. [53]

    author Caldwell, E. D. , author Sinclair, L. C. , author Newbury, N. R. & author Deschenes, J.-D. title The time-programmable frequency comb and its use in quantum-limited ranging . journal Nature volume 610 , pages 667--673 ( year 2022 )

    2022

  46. [54]

    author Giorgetta, F. R. et al. title Free-form dual-comb spectroscopy for compressive sensing and imaging . journal Nature Photonics volume 18 , pages 1312--1319 ( year 2024 )

    2024

  47. [55]

    , author Bosworth, B

    author Shin, J. , author Bosworth, B. T. & author Foster, M. A. title Single-pixel imaging using compressed sensing and wavelength-dependent scattering . journal Optics letters volume 41 , pages 886--889 ( year 2016 )

    2016

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.