pith. sign in

arxiv: 1906.11408 · v1 · pith:D4SVDWWLnew · submitted 2019-06-27 · 📡 eess.IV · physics.optics

Mean Gradient Descent: An optimization approach for single-shot interferogram analysis

Sunaina , Mansi Butola , Kedar Khare This is my paper

Pith reviewed 2026-05-25 14:42 UTC · model grok-4.3

classification 📡 eess.IV physics.optics
keywords single-shot interferogram analysismean gradient descentdigital holographyphase recoveryoptimizationdata consistencyconstraint balancingfull pixel resolution
0
0 comments X

The pith

Mean gradient descent recovers full-resolution object waves from single-shot interferograms by balancing data consistency and constraint terms along their gradient bisector.

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

The paper presents Mean Gradient Descent (MGD) as a new optimization method for recovering the complex object wave from a single interferogram. Traditional Fourier filtering loses resolution while standard optimization requires tuning free parameters which is tedious. MGD instead progresses the solution along the bisector of the descent directions from the data consistency term and constraint terms to reach a balance point rather than minimizing a cost. Simulations for step phase objects from off-axis and on-axis interferograms with noise demonstrate full pixel resolution and good accuracy to ground truth. The approach is parameter-free and may apply to other optimization tasks.

Core claim

The MGD iteration does not try to achieve minimization of any cost function but instead aims to reach a solution point where the data consistency and the constraint terms balance each other. This is achieved by iteratively progressing the solution in the direction that bisects the descent directions associated with the error and constraint terms. Numerical illustrations are shown for recovery of a step phase object from its corresponding off-axis as well as on-axis interferograms simulated with multiple noise levels with full pixel resolution and excellent rms phase accuracy relative to the ground truth phase map.

What carries the argument

The MGD iteration that bisects the descent directions associated with the error and constraint terms to reach a balance between data consistency and constraints.

Load-bearing premise

Iteratively moving along the bisector of the data-consistency and constraint gradients will converge to the physically correct object wave without free parameters or additional regularization choices.

What would settle it

A simulation or experiment where the recovered phase map shows significant deviation from the known ground truth phase despite low noise levels would falsify the claim of accurate full-resolution recovery.

Figures

Figures reproduced from arXiv: 1906.11408 by Kedar Khare, Mansi Butola, Sunaina.

Figure 1
Figure 1. Figure 1: (a) Phase map of square shaped object with a step phase of [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (a) Phase and (b) amplitude of solution for object wave after [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Behaviour of (a) logarithm of Cerr with iteration number, and (b) logarithm of CT V with iteration number for three light levels of 103 ,104 and 105 photons/pixel.(c) Variation of angle between uˆ1 and uˆ2 with number of iterations corresponding to the average light levels of 103 ,104 and 105 photons/pixel. 5 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Progression of solution with MGD algorithm for Poisson noise realization with light level of [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Hologram of step phase object with (a) on-axis and (b) near on-axis spherical reference wave simulated with [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

Complex object wave recovery from single-shot interference pattern is an important practical problem in interferometry and digital holography. The most popular single-shot interferogram analysis method involves Fourier filtering of cross-term but this method suffers from poor resolution. For obtaining full pixel resolution, it is necessary to model the object wave recovery as an optimization problem. The optimization approach typically involves minimizing a cost function consisting of a data consistency term and one or more constraint terms. Despite its potential performance advantages, this method is not used widely due to several tedious and difficult tasks such as empirical tuning of free parameters. We introduce a new optimization approach Mean gradient descent (MGD) for single-shot interferogram analysis that is simple to implement, robust and does not require any free parameters. The MGD iteration does not try to achieve minimization of any cost function but instead aims to reach a solution point where the data consistency and the constraint terms balance each other. This is achieved by iteratively progressing the solution in the direction that bisects the descent directions associated with the error and constraint terms. Numerical illustrations are shown for recovery of a step phase object from its corresponding off-axis as well as on-axis interferograms simulated with multiple noise levels. Our results show full pixel resolution as evident from the recovery of phase step and excellent rms phase accuracy relative to the ground truth phase map. The concept of MGD as presented here can potentially find applications to wider class of optimization problems.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 0 minor

Summary. The paper proposes Mean Gradient Descent (MGD) as a parameter-free optimization method for recovering the complex object wave from single-shot off-axis or on-axis interferograms. Instead of minimizing a composite cost function, MGD iteratively advances the solution along the bisector of the descent directions of the data-consistency (error) term and the constraint term(s) until those terms balance. Numerical results on simulated step-phase objects at multiple noise levels are reported to achieve full pixel resolution with low RMS phase error relative to ground truth.

Significance. If the central construction can be shown to be free of implicit scaling parameters and to converge reliably to the physically correct wave, the approach would address a practical bottleneck in digital holography by eliminating empirical regularization tuning. The reported numerical accuracy on simulated data is encouraging, but the absence of convergence analysis, comparison against established Fourier or optimization baselines, and explicit implementation details limits the immediate impact.

major comments (2)
  1. [Abstract] Abstract: the claim that MGD 'does not require any free parameters' and reaches a balance point by 'iteratively progressing the solution in the direction that bisects the descent directions' is load-bearing for the central contribution. Descent directions are raw gradient vectors; without an explicit statement of whether the vectors are used un-normalized (so the bisector is dominated by the larger-norm term) or first scaled to unit length (introducing an implicit weighting), the landing point of the iteration depends on an unstated relative scaling between the error and constraint gradients. This directly contradicts the parameter-free assertion and must be resolved with a precise algorithmic definition (e.g., an equation for the update direction).
  2. [Abstract] Abstract / numerical illustrations: the reported 'excellent rms phase accuracy' and 'full pixel resolution' rest on simulated data only, with no quantitative tables, no comparison to Fourier filtering or conventional regularized optimization, and no demonstration on experimental interferograms. Because the method's advantage is claimed to be robustness without tuning, the lack of such controls undermines the practical-significance claim.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback on our manuscript. Below we respond point-by-point to the major comments, indicating where revisions will be made.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that MGD 'does not require any free parameters' and reaches a balance point by 'iteratively progressing the solution in the direction that bisects the descent directions' is load-bearing for the central contribution. Descent directions are raw gradient vectors; without an explicit statement of whether the vectors are used un-normalized (so the bisector is dominated by the larger-norm term) or first scaled to unit length (introducing an implicit weighting), the landing point of the iteration depends on an unstated relative scaling between the error and constraint gradients. This directly contradicts the parameter-free assertion and must be resolved with a precise algorithmic definition (e.g., an equation for the update direction).

    Authors: We agree that the manuscript requires a precise statement of the update rule to substantiate the parameter-free claim. The current description does not specify normalization of the descent directions. In the revision we will add an explicit equation defining the MGD step as the normalized average of the unit descent directions of the error and constraint terms. This removes any implicit scaling and will be placed in the methods section with a corresponding update to the abstract wording if needed. revision: yes

  2. Referee: [Abstract] Abstract / numerical illustrations: the reported 'excellent rms phase accuracy' and 'full pixel resolution' rest on simulated data only, with no quantitative tables, no comparison to Fourier filtering or conventional regularized optimization, and no demonstration on experimental interferograms. Because the method's advantage is claimed to be robustness without tuning, the lack of such controls undermines the practical-significance claim.

    Authors: We accept that quantitative benchmarks against established methods would strengthen the presentation. The revised manuscript will include tables of RMS phase errors for MGD versus Fourier filtering and a standard regularized optimizer on the same simulated step-phase objects at each noise level. The study deliberately uses controlled simulations to enable direct comparison with ground truth; experimental data are outside the present scope and will be addressed separately. revision: partial

standing simulated objections not resolved
  • Demonstration on real experimental interferograms (the work is restricted to simulated data with known ground truth)

Circularity Check

0 steps flagged

MGD presented as direct algorithmic construction; no reduction to fitted parameters or self-citations

full rationale

The paper defines MGD explicitly as an iteration that advances along the bisector of the descent directions of the data-consistency and constraint terms, with the explicit claim that it does not minimize a cost function and requires no free parameters. No equation in the provided text equates a derived quantity to a fitted parameter taken from the same work, nor does any central premise rest on a self-citation chain. The construction is presented as an independent algorithmic choice rather than a tautological renaming or normalization of its own inputs. This satisfies the default expectation of no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach assumes an interferometric forward model and the existence of suitable constraint terms whose gradients can be computed; these are standard domain assumptions rather than new entities or fitted constants.

axioms (2)
  • domain assumption An accurate forward model exists that maps the object wave to the measured interferogram intensity.
    Invoked implicitly when stating that data consistency can be evaluated.
  • domain assumption Suitable constraint terms (e.g., smoothness or support) can be defined whose gradients are available.
    Required for the bisecting step to be well-defined.

pith-pipeline@v0.9.0 · 5789 in / 1396 out tokens · 24782 ms · 2026-05-25T14:42:46.618163+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    The MGD iteration does not try to achieve minimization of any cost function but instead aims to reach a solution point where the data consistency and the constraint terms balance each other. This is achieved by iteratively progressing the solution in the direction that bisects the descent directions associated with the error and constraint terms.

  • IndisputableMonolith/Foundation/BranchSelection.lean branch_selection unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    û1 = ∇O* Cerr / ||∇O* Cerr||2 , û2 = ∇O* CTV / ||∇O* CTV||2 , u = (û1 + û2)/2

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages

  1. [1]

    Hariharan, Optical Holography, (Cambridge U

    P. Hariharan, Optical Holography, (Cambridge U. P.,1984)

    work page 1984

  2. [2]

    Fourier-transform method of fringe-pattern analysis for computer based topography and interferometry,

    M. Takeda, H. Ina, and S. Kobayashi, “Fourier-transform method of fringe-pattern analysis for computer based topography and interferometry,” J. Opt. Soc. Am. A 72, 156–160 (1982). 8 A PREPRINT - J UNE 28, 2019 Figure 5: Hologram of step phase object with (a) on-axis and (b) near on-axis spherical reference wave simulated with Poisson noise with an average...

    work page 1982

  3. [3]

    Digital holographic interference-phase measurement using the Fourier-transform method,

    T. Kreis, Thomas,“Digital holographic interference-phase measurement using the Fourier-transform method,” J. Opt. Soc. Am. A 3, 847-855 (1986)

    work page 1986

  4. [4]

    Phase shifting interferometry techniques,

    K. Creath, “Phase shifting interferometry techniques,” in Progress in Optics, E. Wolf, ed. (Elsevier, 1988), V ol. 26, pp. 357–373

    work page 1988

  5. [5]

    Phase Shifting Digital Holography,

    I. Yamaguchi and T. Zhang, “ Phase Shifting Digital Holography,” Opt. Lett. 22, 1268-1270 (1997)

    work page 1997

  6. [6]

    Demodulation of a single interferogram by use of a two-dimensional regularized phase-tracking technique,

    M. Servin, J. L. Marroquin and F. J. Cuevas, “Demodulation of a single interferogram by use of a two-dimensional regularized phase-tracking technique,” Appl. Opt. 36(19), 4540-4548 (1997)

    work page 1997

  7. [7]

    Servin, J

    M. Servin, J. A. Quiroga, and J. M. Padilla, Fringe pattern analysis for optical metrology: theory, algorithms, and applications, (John Wiley & Sons, 2014)

    work page 2014

  8. [8]

    Complex-wave retrieval from a single off-axis hologram,

    M. Liebling, T. Blu, and M. Unser, “Complex-wave retrieval from a single off-axis hologram,” J. Opt. Soc. Am. A 21(3), 367-377 (2004)

    work page 2004

  9. [9]

    A complex gradient operator and its application in adaptive array theory,

    D. H. Brandwood, “A complex gradient operator and its application in adaptive array theory,” Proc. IEE 130, 11–16 (1983)

    work page 1983

  10. [10]

    Single shot high resolution digital holography,

    K. Khare, P.T.S. Ali, and J. Joseph, “Single shot high resolution digital holography,” Opt. Exp. 21, 2581-2591 (2013)

    work page 2013

  11. [11]

    Accurate multipixel phase measurement with classical-light interferometry,

    M. Singh, K. Khare, A.K. Jha, S. Prabhakar, and R.P. Singh, “Accurate multipixel phase measurement with classical-light interferometry,” Phys. Rev. A 91, 021802 (2015)

    work page 2015

  12. [12]

    Single shot Digital Holography Using Iterative Reconstruction with Alternating updates of Amplitude and Phase,

    D. J. Lee, C. A. Bouman, and A. M. Weiner, “Single shot Digital Holography Using Iterative Reconstruction with Alternating updates of Amplitude and Phase,” Electronic imaging 19, 1-6 (2016)

    work page 2016

  13. [13]

    Single-shot interferogram analysis for accurate reconstruction of step phase objects,

    M. Singh and K. Khare, “Single-shot interferogram analysis for accurate reconstruction of step phase objects,” J. Opt. Soc. Am. A 34, 349–355 (2017). 9 A PREPRINT - J UNE 28, 2019

    work page 2017

  14. [14]

    Efficient algorithm for isotropic and anisotropic total variation deblurring and denoising,

    Y . Shi, and Q. Chang, “Efficient algorithm for isotropic and anisotropic total variation deblurring and denoising,” J. Appl. Math. 2013, (2013)

    work page 2013

  15. [15]

    Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization,

    E. Y . Sidky. and X. Pan, “Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization,” Phys. Med. Biol. 53(17), 4777 (2008)

    work page 2008

  16. [16]

    Total variation regularization cost function for demodulating phase discontinuities,

    R. Legarda-Saenz, C. Brito-Loeza, and A. Espinosa-Romero, “Total variation regularization cost function for demodulating phase discontinuities,” Appl. Opt. 53, 2297–2301 (2014)

    work page 2014

  17. [17]

    Second-order robust regularization cost function for detecting and reconstructing phase discontinuities,

    C. Galvan, and M. Rivera, “Second-order robust regularization cost function for detecting and reconstructing phase discontinuities,” Appl. Opt. 45, 353-359 (2006)

    work page 2006

  18. [18]

    Kramers-Kronig holographic imaging for high-space-bandwidth product,

    Y . Baek, K. Lee, S. Shin, and Y . Park,“ Kramers-Kronig holographic imaging for high-space-bandwidth product,” Optica 6, 45-51 (2019)

    work page 2019

  19. [19]

    D. F. Walls and G. J. Milburn, Quantum optics, 2nd ed. (Springer Science & Business Media, 2007). 10

    work page 2007