{"id":"3c08275b-dd8b-4ca8-ad84-b9838ef8bf7b","arxiv_id":"2501.10022","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"By subtracting cepstra (log-domain spectra) of one-pixel-shifted off-axis holograms, the authors recover three complex amplitude fields and triple the usable field of view in quantitative phase microscopy.","lead":"This paper introduces Cepstrum-based Interferometric Microscopy (CIM), a method that recovers quantitative phase images from four off-axis holograms without requiring a clean reference beam, and retrieves three overlapping fields of view at once. If it holds up, it triples the useful image area per sensor and could make quantitative phase imaging easier to retrofit onto ordinary microscopes.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Section 2.3 Eq. (3) misapplies the cepstrum identity: log of a convolution is not the sum of logarithms; the formulas actually used rely on a pointwise log of U=O1O2* and therefore inherit nonzero-field and phase-unwrapping restrictions that contradict the claimed 'no restrictions' universality.","rationale":"The central promise of CIM is that two arbitrary complex fields can be separated without a clean reference or other a priori information. The only place where that separation is derived is Sec. 2.3, and Eq. (3) is where the logarithm is split. If Eq. (3) is false, the derivation as written does not establish the algorithm; if it is replaced by the pointwise-log construction implicit in Eqs. (6)-(7), the method is restricted to non-vanishing, phase-unwrappable fields with stable coherent noise across four sequential recordings. This is the most load-bearing concern because every recovered FOV—static, horizontal, vertical—depends on this step. I am not arguing the experimental results are fabricated or that the method cannot work; the calibrated phase-target thicknesses and the comparison with common-path and conventional DHM are meaningful evidence that something along these lines can work on static samples. But those experiments do not test the universality claim, since the phase targets and fixed cells are phase objects with roughly constant amplitude and no controlled phase wraps or zero-amplitude regions. The paper also does not release code or data, so the discrepancy between the stated cepstrum identity and the implemented log-domain calculation cannot be checked from the text alone. The reader's conditional verdict is appropriate: the method deserves revision and re-review, not rejection, and a synthetic noiseless test plus a corrected derivation would resolve the remaining doubt.","tokens_in":18392,"tokens_out":14122,"duration_ms":145263,"concrete_test":"Run a noiseless 2D synthetic recovery with known complex fields: choose O1 and O2 with non-zero amplitude everywhere, form U=O1O2* and Ux=O1(x-x0,y)O2*(x,y) with x0 exactly one pixel, apply Eqs. (6)-(7), and confirm exact recovery. Then repeat the same recovery with (a) one zero-amplitude pixel inserted in O1 and (b) a 2π phase wrap added to U. If the recovered O1 shows localized artifacts in cases (a)/(b), the method does not separate arbitrary fields and the 'no restrictions' claim fails. Separately, for the same O1,O2 compute both sides of Eq. (3) numerically: F^{-1}log(F(U)) versus cepstrum(O1)+cepstrum(O2*); a non-negligible difference demonstrates the factorization error directly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (3) is the load-bearing step in the SSC derivation. Starting from Eq. (1), U~(u,v)=O1~(u,v)⊗O2~*(u,v) is the Fourier transform of the spatial cross-correlation U(x,y)=O1(x,y)O2*(x,y), so the spectrum is a convolution. Applying the complex cepstrum defined in Eq. (2) gives F^{-1} log[O1~⊗O2~*], and log of a convolution is not the sum of the individual logs. Thus U~=O1~+O2~* does not follow by any standard cepstrum property. The working formulas in Eqs. (6)-(7) are not actually this cepstrum step: they use F{log|U|-log|Ux|} and F{φ_U-φ_Ux}, i.e., they take the complex logarithm pointwise in the spatial domain before Fourier transformation. That operation is legitimate only where O1 and O2 are nonzero, and it requires a consistent unwrapped phase branch for the difference φ_U-φ_Ux. These are exactly the 'a priori assumptions' the abstract claims are unnecessary. The pointwise-log reading also makes the 1-pixel shift requirement literal: O1(x-x0,y) must be an exact translated copy with O2 and the coherent noise unchanged between the two recordings; the Discussion concedes coherent noise is not constant between interferograms. As written, Sec. 2.3 does not support the central claim; a corrected derivation must state the pointwise-product/log-domain assumptions and the resulting restrictions.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript proposes Cepstrum-based Interferometric Microscopy (CIM), a quantitative phase imaging (QPI) method based on four off-axis holograms in which one interferometric beam is shifted by one pixel in two orthogonal directions. The authors introduce a Spatial-Shifting Cepstrum (SSC) algorithm intended to recover the complex amplitude distributions of three fields without requiring a clean reference beam, thereby tripling the field of view. The method is validated on a phase resolution test target and on fixed biological samples (LnCaP and PC3 prostate cancer cells, cheek cells), with thickness measurements compared to AFM and conventional and common-path digital holographic microscopy (DHM).","tokens_in":18728,"tokens_out":6463,"duration_ms":63079,"significance":"If the claims are upheld, CIM would be a practically attractive QPI approach that removes the sparse-sample or clean-reference constraints of many common-path interferometers and extends the usable field of view. The experimental validation is substantial: measured thicknesses agree with AFM values within uncertainty (316±26 nm vs 329.5 nm for the static FOV; 400±24 nm vs 388 nm and 287±24 nm vs 278 nm for the shifted FOVs), and the phase comparison with common-path DHM shows a mean difference of -0.02±0.24 rad. The authors also honestly discuss limitations, including the zero-ambiguity at DC and the instability of coherent noise between recordings. However, the central mathematical derivation in Sec. 2.3 is flawed, and the claimed universality (no assumptions on either beam) is not supported as written. The paper is a promising proof-of-concept but requires a corrected derivation and appropriately qualified claims before publication.","major_comments":[{"comment":"Equation (3) is not a valid consequence of applying the complex cepstrum to Eq. (1). From Eq. (1), U~(u,v) = O1~(u,v) ⊗ O2~*(u,v), so the cepstrum gives F^{-1} log[O1~ ⊗ O2~*], and the logarithm of a convolution is not the sum of logarithms. The standard cepstrum property applies to convolutions in the spatial domain, not to the Fourier-domain convolution of two spectra. The working formulas (6)-(7) are indeed based on a different, pointwise logarithmic operation on U(x,y)=O1(x,y)O2*(x,y). The derivation should be rewritten to state the pointwise-product/log-domain assumptions and to remove the incorrect statement that applying Eq. (2) to Eq. (1) yields Eq. (3).","section":"Sec. 2.3, Eq. (3)"},{"comment":"The amplitude and phase formulas require that, for each direction, the second recording contains exactly the same O2(x,y) (and O1 shifted by x0), with no change in coherent noise or illumination between the two frames. This condition is not a mere practical detail: it is essential to the subtraction step, and it is also a restriction on the beams that contradicts the abstract's claim that no assumptions are required for the two interferometric beams. The Discussion (Sec. 4) itself concedes that coherent noise is not constant between interferograms, causing high-frequency mismatches. These restrictions, together with the nonzero-field and phase-unwrapping requirements of the pointwise complex logarithm, should be stated explicitly in the derivation and in the statements of generality.","section":"Sec. 2.3, Eqs. (6)-(7)"},{"comment":"The reconstruction divides by 1-exp(-i2πu x0), which is singular at u=0 for x0=1, and the resulting zero-ambiguity is removed by bow-tie masking followed by Gaussian high-pass filtering with a manually chosen FWHM of 9 pixels. This operation discards the DC and low-frequency content of the retrieved field, so the method yields relative phase/thickness variations rather than an absolute phase offset. The paper should state this limitation clearly; in particular, the choice of the Gaussian FWHM is a free parameter that affects the low-frequency content of the final QPI image, and the claim of a parameter-free or fully assumption-free reconstruction should be tempered accordingly.","section":"Sec. 2.3, Eq. (5) and Sec. 4"}],"minor_comments":[{"comment":"The in-text citations [48-49] for Kramers-Kronig approaches do not match the reference list, where Ref. [48] is a transport-of-intensity tutorial and Ref. [49] is the Oppenheim and Schafer cepstrum paper. Please correct the citation numbering.","section":"Sec. 4 (Discussion)"},{"comment":"The notation for the retrieved quantities is inconsistent: Õ0̃1, Õ1, and Re{Õ1} are used without clear definitions. In particular, Eqs. (6)-(7) should explicitly define the logarithm of the spectrum (or the cepstrum) that is being inverse-Fourier-transformed, rather than writing Re{Õ1(u,v)} where Õ1 is the ordinary Fourier transform of O1.","section":"Sec. 2.3, Eqs. (5)-(7)"},{"comment":"The sentence \"forty frames integrate a displacement of a single pixel\" is ambiguous: it is not explained how the forty frames are used to determine or ensure the exact one-pixel shift, nor how the 25-thousandths-of-a-pixel confidence was obtained from the cross-correlation registration. Please elaborate.","section":"Sec. 2.2 (Calibration Procedure)"},{"comment":"The workflow figure is dense, and the numbered step labels are small relative to the panels. Consider enlarging the step numbers or splitting the figure into two panels for readability.","section":"Fig. 2"},{"comment":"The noise comparison for the biosample background reports only the mean and SD of the phase; the authors could also provide a phase-noise map or an RMS value to separate low-frequency background variations from high-frequency phase noise, which would better support the comparison between CIM and the two DHM alternatives.","section":"Table 2"}],"recommendation":"major_revision","confidential_remarks":"The derivation gap in Sec. 2.3 is serious but appears to be fixable: the authors should replace the erroneous cepstrum identity with the correct pointwise-logarithm derivation and then qualify the universality claims. The experimental data are credible and the method is conceptually novel, so I believe the manuscript can be made acceptable after a major revision. I also note that the reference numbering in the Discussion needs correction before the paper is resubmitted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Rubio-Oliver et al. report a way to do off-axis holographic QPI without a clean reference beam. They record four holograms (two orthogonal directions, each with a 1-pixel shift of one beam), take the pointwise complex logarithm of the filtered cross-correlation, subtract shifted versions, and divide by the known shift transfer function. The result is a parameter-free recovery of three complex fields, tripling the field of view. That is genuinely new relative to the cited cepstrum-DHM work (Refs 45-47), which either suppress diffraction orders or assume a known or dominant reference. The experiments back the proof of concept: thicknesses 316±26 nm vs 329.5 nm AFM, 400±24 vs 388, 287±24 vs 278; phase difference -0.02±0.24 rad vs common-path DHM; resolution limit preserved. Those numbers are credible.\n\nThe main soft spot is the derivation. Eq. (3) claims the cepstrum of a convolution is the sum of the individual cepstra. That is not true: the Fourier transform of the cross-correlation is a convolution in the spectral domain, and the log of a convolution does not separate. The working formulas (6)-(7) are not actually that step; they take log|U| and the phase of U pointwise in the spatial domain, then Fourier transform. That is a legitimate but different operation, and it carries real restrictions: both fields must be nonzero wherever data are used, the phase branch must be unwrapped consistently, the 1-pixel shift must be an exact translation with unchanged illumination and coherent noise, and the sample must be static across the four sequential frames. Those are a priori assumptions the abstract claims are unnecessary. The Discussion does concede that coherent noise varies between interferograms, and the static-sample validation does not address drift. So the universal/no-restrictions framing overreaches. The fix is straightforward: rewrite Sec. 2.3 as a pointwise-log subtraction, state the nonzero/unwrapping/stationarity conditions explicitly, and soften the universality claim.\n\nOther issues are minor: no code or data release, which would help independent verification; the tripled FOV is non-contiguous in this implementation, which limits some applications; and the high-pass filter (FWHM 9 px) removes low-frequency phase, though the comparison plots suggest the effect is mild for the tested targets. The citation pattern looks fair; the prior cepstrum-DHM and tripled-FOV literature is properly acknowledged.\n\nWho should read this: anyone working on compact QPI add-ons, off-axis holography, or FOV-multiplexed interferometry. It deserves a serious referee: the experimental result is useful and the core idea is sound once the derivation is corrected. I would send it to review, with the expectation that the authors fix Eq. (3) and the claims. My verdict is a conditional accept after major revision.","headline":"CIM is a genuinely new way to do QPI without a clean reference and triples the FOV, but the written derivation misuses the cepstrum and the universal/no-restrictions claim overreaches.","tokens_in":19333,"tokens_out":4409,"would_cite":true,"duration_ms":35596,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Four holograms with one-pixel shifts give three quantitative phase images.","keywords":["digital holographic microscopy","quantitative phase imaging","phase retrieval","complex cepstrum","Spatial-Shifting Cepstrum","field of view tripling","off-axis holography","Michelson interferometer"],"falsifier":"Record a CIM sequence on a phase target while a small opaque absorber obscures part of only one of the three fields, then compare the recovered static field with a conventional off-axis hologram of the same region: at pixels where the shifted beam has near-zero amplitude, the complex logarithm becomes undefined and the reconstruction should show localized artifacts that mark the exact failure of the nonzero-field assumption.","tokens_in":1702,"feed_emoji":"🔬","tokens_out":5646,"duration_ms":94801,"temperature":0.7,"pith_summary":"Conventional off-axis digital holographic microscopy assumes one of the two interfering beams is a known clean reference. This paper claims that assumption is unnecessary: two arbitrary unknown complex fields can be separated from their cross-correlation. The proposed Cepstrum-based Interferometric Microscopy (CIM) records four off-axis holograms, applying a one-pixel shift to one beam in horizontal and vertical directions, and then uses the Spatial-Shifting Cepstrum (SSC) algorithm to recover the static field. The two shifted fields are obtained by subtraction, yielding three quantitative phase images from four recordings and tripling the usable field of view. Validation on calibrated phase targets and fixed biological cells shows agreement with conventional digital holographic microscopy, with slightly more noise.","feed_headline":"Four holograms, two shifts, three phase images","feed_subtitle":"Cepstrum subtraction splits two unknown optical fields, yielding three phase images from four holograms.","key_machinery":"The central object is the Spatial-Shifting Cepstrum (SSC) algorithm, built on the complex cepstrum, defined as the inverse Fourier transform of the logarithm of a Fourier-transformed field. The algorithm takes the filtered cross-correlation terms from two off-axis holograms in the same direction, one containing an exact one-pixel shift, computes their complex cepstra, and subtracts them. The shift enters as the transfer function $1 - \\exp(-i 2\\pi u x_0)$ in Eq. (5); division by this factor recovers the static field's spectrum. The companion field is then obtained by subtracting the static spectrum from the original cross-correlation and exponentiating. Complementary bow-tie masks in the horizontal and vertical Fourier domains combine the two partial spectra so that only the DC term remains ambiguous, and a high-pass filter removes the resulting background inhomogeneity.","core_discovery":"The central discovery asserted is that the cross-correlation of two unknown fields, $\\tilde U(u,v) = \\tilde O_1(u,v) \\otimes \\tilde O_2^*(u,v)$, can be inverted without knowing either field, provided a second hologram in the same direction includes an exact one-pixel shift of one of the fields. In the complex-cepstrum domain the product becomes a sum, and subtracting the shifted hologram cancels the unknown companion. Equation (5) then yields the static object spectrum divided by $1 - \\exp(-i 2\\pi u x_0)$, leaving a null line in the Fourier domain. Repeating in an orthogonal direction and combining with complementary bow-tie masks collapses the null lines to the DC term, and the two shifted fields are recovered by complex subtraction. The result is a full quantitative phase image of each of the three fields involved in the two orthogonal interferometric recordings.","pith_inferences":["Editorial inference: the same cepstrum-subtraction logic should work with shifts larger than one pixel, but at the cost of additional null lines in the Fourier domain; the paper chooses one pixel to minimize these singularities.","Editorial inference: the coherent-noise mismatch that the Discussion concedes should appear as a systematic high-frequency error in the recovered static field; averaging multiple interferograms per shift could test whether the mismatch is random speckle or slow drift between frames.","Editorial extension: because the method records four frames from a static sample, applying CIM to live or flowing cells would require rapid motorized switching between shift directions; the paper mentions video-rate implementation as future work but does not demonstrate it."],"forward_implications":["A compact Michelson module placed after a standard microscope tube lens can provide quantitative phase imaging without pinholes, sparse-sample constraints, or spatially reserved reference regions.","Four time-sequential holograms supply three complex images, making the temporal footprint comparable to phase-shifting holography while tripling the imaged area.","The one-pixel shift direction need not be horizontal and vertical; any two non-parallel directions work, making the geometry adaptable to different mirror stages and optical layouts.","The DC-only ambiguity left after bow-tie masking can be removed by high-pass filtering without erasing low-frequency sample content, as confirmed by comparison with conventional off-axis holography on a calibrated phase target.","The three recovered fields need not be adjacent in the current implementation, but adjusting the off-axis angle and mirror distance could make the extended field of view contiguous."],"supporting_citations":[{"why":"Defines the complex cepstrum, the transform at the core of the SSC algorithm.","marker":"[49]"},{"why":"Introduces two-step phase retrieval from the logarithmic transformation of a hologram's Fourier transform, the cepstrum precursor that SSC extends.","marker":"[45]"},{"why":"Provides experimental validation of cepstrum-based holographic reconstruction with two cameras, prior art that the paper claims to improve on.","marker":"[46]"},{"why":"Applies cepstrum filtering to slightly off-axis holography, one of the reference-beam-dependent approaches that SSC removes the constraint for.","marker":"[47]"},{"why":"Demonstrates tripled imaging area in off-axis interferometric phase microscopy, the baseline that the paper's field-of-view claim extends.","marker":"[41]"},{"why":"Supplies the calibrated phase resolution target used for quantitative thickness comparison against AFM and conventional digital holographic microscopy.","marker":"[50]"}],"fun_headline_variants":["Cepstrum subtraction yields three phase images","Four holograms, two shifts, triple phase map","Two shifts unlock three quantitative phase fields","Cepstrum-based microscope triples phase view"],"cache_read_input_tokens":21248,"weakest_assumption_plain":"For the math to work, the one-pixel shift must be an exact rigid translation of one complex field while the other stays fixed, both fields must be nonzero wherever the algorithm uses them, and the four recordings must share the same coherent noise with no sample motion between frames.","fun_headline_variants_meta":{"raw":{"variants":["Cepstrum subtraction yields three phase images","Four holograms, two shifts, triple phase map","Two shifts unlock three quantitative phase fields","Cepstrum-based microscope triples phase view"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00028,"raw_usage":{"total_tokens":1680,"prompt_tokens":987,"completion_tokens":693,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":603,"completion_tokens_details":{"reasoning_tokens":634}},"tokens_in":603,"tokens_out":693,"duration_ms":7387,"temperature":1.0,"reasoning_tokens":634,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:24:18.608916+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Record a CIM sequence on a phase target while a small opaque absorber obscures part of only one of the three fields, then compare the recovered static field with a conventional off-axis hologram of the same region: at pixels where the shifted beam has near-zero amplitude, the complex logarithm becomes undefined and the reconstruction should show localized artifacts that mark the exact failure of the nonzero-field assumption.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines the complex cepstrum, the transform at the core of the SSC algorithm."},{"cited_title":"Zhang, G","cited_arxiv_id":null,"evidence_quote":"Introduces two-step phase retrieval from the logarithmic transformation of a hologram's Fourier transform, the cepstrum precursor that SSC extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides experimental validation of cepstrum-based holographic reconstruction with two cameras, prior art that the paper claims to improve on."},{"cited_title":"Pavillon, C.S","cited_arxiv_id":null,"evidence_quote":"Applies cepstrum filtering to slightly off-axis holography, one of the reference-beam-dependent approaches that SSC removes the constraint for."},{"cited_title":"Girshovitz, N.T","cited_arxiv_id":null,"evidence_quote":"Demonstrates tripled imaging area in off-axis interferometric phase microscopy, the baseline that the paper's field-of-view claim extends."},{"cited_title":"https://www.benchmarktech.com/index.php?q=quantitativephasemicroscop","cited_arxiv_id":null,"evidence_quote":"Supplies the calibrated phase resolution target used for quantitative thickness comparison against AFM and conventional digital holographic microscopy."}],"review_version":1}