Pith. sign in

REVIEW 1 major objections 5 minor 24 references

Deconvolution of 3-D Gaussian kernels

T0 review · 1 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read The paper derives an explicit inverse kernel for three-dimensional Gaussian deconvolution.

desk verdict Main 3D inverse-kernel series is correct and the scalar-Hermite identity is neat, but the paper's alternative delta-form has a coefficient error and is wrong as printed. read the letter →

arxiv 1908.07259 v1 pith:4IY65Y7O submitted 2019-08-20 physics.data-an hep-ph

classification physics.data-anhep-ph MSC 44A3533C45 PACS 02.30.Zz02.30.Uu
keywords deconvolutionGaussiankernelsmultivariateHermitepolynomialsscalarLaguerreinverseintegralkernelGrad
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

The paper derives explicit inverse kernels for the three-dimensional Gaussian convolution, extending the one-dimensional formulas that this work generalizes. The central result is that the inverse of the blurring operator can be written as an infinite series of ordinary Gaussians multiplied by scalar Hermite polynomials, with an equivalent delta-plus-series form. The argument writes the Gaussian kernel as a differential operator acting on a delta function, moves the inverse operator onto the signal, and expresses the resulting scalar Hermite polynomials through ordinary Hermite polynomials. If correct, the construction gives a direct, rotationally invariant way to recover volumetric signals blurred by an isotropic Gaussian resolution without relying on Fourier-domain inversion.

What carries the argument

The load-bearing object is the multivariate Hermite polynomial $H^{(2n)}_{i_1\cdots i_{2n}}(r)=(-1)^{2n}e^{r^2}\nabla_{i_1}\cdots\nabla_{i_{2n}}e^{-r^2}$, fully contracted to the scalar $H_{2n}(r^2)$. From the commutation relations of the operators $E=r^2/2$, $F=-\Delta/2$, $H=r\cdot\nabla+3/2$, the paper obtains the identity $e^{r^2}\Delta e^{-r^2}=-2F-4H+8E$, and hence $H_{2n}(r^2)=e^{-\Delta/4}(2r)^{2n}$. Expanding this with $\Delta^m r^{2n}$ and comparing with associated Laguerre polynomials gives $H_{2n}(r^2)=H_{2n+1}(r)/(2r)$. That reduction is what turns the inverse kernel into a computable series of ordinary Hermite polynomials times Gaussians.

What would settle it

Take a smooth compactly supported signal $\rho$, form $\varphi=K*\rho$ numerically, apply the truncated series in the claimed inverse kernel with increasing $n$, and check whether the error to $\rho$ decreases to zero; failure on a smooth band-limited input would show the series is not an inverse in the claimed sense.

Watch

Extended reading notes

Core claim

The paper establishes that $K^{-1}(r-\xi;\sigma)=\sum_{n=0}^{\infty}\frac{(-1)^n}{2^n n!}H_{2n}\left(\frac{(r-\xi)^2}{\sigma^2}\right)K(r-\xi;\sigma)$, where $K$ is the three-dimensional Gaussian kernel and $H_{2n}$ is the completely contracted scalar multivariate Hermite polynomial, recovers $\rho$ from $\varphi=K*\rho$ through $\rho(r)=\int K^{-1}(r-\xi;\sigma)\varphi(\xi)\,d\xi$. A second form adds a delta function and starts the series at $n=1$ with an extra factor $(2n-1)$. The paper further shows that the scalar Hermite polynomials reduce to ordinary Hermite polynomials via $H_{2n}(r^2)=H_{2n+1}(r)/(2r)$, so the inverse kernel is a series of explicit functions rather than an abstract object.

Load-bearing premise

The derivation assumes the blurred signal is smooth enough and decays fast enough that integrating by parts leaves no boundary terms, and that the infinite exponential-operator series converges when applied to the signal.

Editorial extensions

If this is right

  • Volumetric deconvolution with an isotropic Gaussian point-spread function can be performed by direct series evaluation instead of solving the integral equation or using Fourier transforms.
  • Truncating the series at finite $n$ gives a natural approximate inverse, so the formula opens a route to controlled approximation errors in dose or image reconstruction.
  • The delta-plus-series form separates the singular part of the inverse from a smooth correction, which can help when correcting for finite detector or chamber volume.
  • The identity $H_{2n}(r^2)=H_{2n+1}(r)/(2r)$ supplies a practical evaluation scheme for contracted multivariate Hermite polynomials in kinetic-theory moment expansions.

Reading between the lines

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

  • For real data with noise, the series inverse will amplify high spatial frequencies, so practical use would require truncation or regularization; the paper notes it does not test robustness.
  • Because the three-dimensional Gaussian kernel is separable, the inversion could be applied as three successive one-dimensional inverses; the scalar form is preferable when rotational invariance and a single radial variable matter.
  • The same operator method should extend to anisotropic Gaussian kernels by replacing $\sigma^2\Delta$ with a weighted Laplacian, and to mixtures of Gaussians by summing the corresponding inverse series.
  • A numerical benchmark on synthetic volumes with known true signals would settle how many terms of the series are needed for a given resolution, a testable consequence not explored in the paper.
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

1 major / 5 minor

Summary. This short paper generalizes the one-dimensional Gaussian deconvolution formulas of Ulmer and Kaissl to three dimensions. Starting from the representation K = exp(σ²Δ/4)δ, the author writes the inverse kernel as an exponential operator, expands it, and expresses powers of the Laplacian acting on the 3D Gaussian in terms of fully contracted Grad multivariate Hermite polynomials. The main result is Eq. (24): K^{-1}(r-ξ;σ) = Σ_{n=0}^∞ (-1)^n/(2^n n!) H_{2n}((r-ξ)^2/σ^2) K(r-ξ;σ). A substantial part of the paper proves that the scalar Hermite polynomials satisfy H_{2n}(r²)=H_{2n+1}(r)/(2r), using Laguerre polynomials and a generating-function argument. The paper explicitly states that no numerical testing or practical algorithm is proposed.

Significance. The derivations are self-contained and the special-function identities check out against standard formulas. The final relation (44) is a clean and useful byproduct, and the generating-function proof of the three-dimensional identity (46) is elegant. The main formula (24) is explicit and parameter-free, and the paper is honest about not testing numerical robustness. If the coefficient error in the alternative delta-containing formula is corrected, this is a valid contribution to the theory of Gaussian deconvolution, with potential applications in medical physics, plasma physics, and nonlocal field theory.

major comments (1)
  1. [Section 3, Eq. (26); also Section 2, Eq. (17)] The coefficient (2n-1) in the delta-containing inverse kernel is incorrect. With A=σ²Δ/4, Eq. (15) gives e^{-A}δ = δ + (e^{-2A}-e^{-A})K. Expanding e^{-2A}-e^{-A} = Σ_{n=1}^∞ (-1)^n(2^n-1)A^n/n! and using A^nK = 4^{-n}H_{2n}K yields the coefficient (-1)^n(2^n-1)/(4^n n!), not (-1)^n(2n-1)/(4^n n!). For n=3 the printed formula gives -5/384 instead of the correct -7/384. Because Eq. (26) is presented as a main generalization alongside Eq. (24), this is a load-bearing error; the same correction is needed in the one-dimensional formula (17).
minor comments (5)
  1. [Section 2, Eq. (6)] The integration by parts is formal; please state explicitly that φ is assumed smooth and sufficiently decaying at infinity so that boundary terms vanish and the exponential-operator series converges on the relevant class of functions.
  2. [Section 3, Eq. (25)] The scaling of the Grad polynomials in H^{(2n)}(r-ξ;σ) is not defined, since Eq. (23) defines H^{(n)} for dimensionless variables. Please state the identity Δ^nK = σ^{-2n}H_{2n}((r-ξ)^2/σ^2)K explicitly to make the notation unambiguous.
  3. [Section 4, Eq. (30)] The Baker–Hausdorff formula should read [A,[A,[A,B]]]; a comma is missing before B in the printed expression.
  4. [Section 5] The sentence 'The work of is supported by the Ministry of Education and Science of the Russian Federation' appears to be missing the author's name.
  5. [Section 4, Eq. (44)] Equation (44) has a removable singularity at r=0; this is harmless because H_{2n+1}(r) is divisible by r, but a brief remark would help readers.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the inverse 3-D Gaussian kernel is obtained from operator identities and standard Hermite/Laguerre relations; the only self-citation is a non-load-bearing closing remark.

full rationale

The central claim, Eq. (24), is derived by applying the inverse operator exp(-sigma^2/2 Delta) to the Gaussian kernel and expressing the resulting derivatives through Grad's multivariate Hermite polynomials via their defining Rodrigues formula (Eq. (23)) and the scalar contraction (Eq. (25)). This is a direct application of an independently stated definition, not an input disguised as an output. Equation (44), which converts scalar Hermite polynomials into ordinary ones, is proved from the sl(2,C) commutation relations (Eq. (28)), the Baker-Hausdorff conjugation (Eqs. (31)-(34)), and the standard Laguerre-Hermite identity (Eq. (43)) cited to Bell [14]; none of these steps presuppose Eq. (24). The alternative delta-form Eq. (26) is asserted rather than fully derived, and its printed coefficient appears algebraically inconsistent with Eq. (15) (expanding e^{-2a}-e^{-a} gives (-1)^n(2^n-1)/n! a^n, not the printed (2n-1) denominator), but that is a correctness flaw, not circularity: it is not a fitted parameter being renamed as a prediction. The only self-reference is the closing remark that the method was used in [22] (Maziashvili & Silagadze); that citation is post-derivation, non-load-bearing, and does not supply any premise of the calculation. The paper also explicitly disclaims robustness testing in Section 5, which is a limitation rather than a circular step. Overall, the derivation is self-contained and would stand or fall on standard mathematical identities rather than on self-citation.

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

No free parameters or invented physical entities are introduced. The derivation relies on standard distribution identities, the definition of Grad polynomials, and standard special-function relations, all cited from the literature. None of these axioms encodes the target result.

assumptions (4)
  • standard math The Gaussian kernel equals the heat operator applied to the Dirac delta: K(x;sigma)=exp(sigma^2/4 d^2/dx^2) delta(x).
    Used at Eq. (4) as the starting point; proved via the Fourier representation in Section 2.
  • domain assumption Formal operator calculus with unbounded differential operators and Baker-Hausdorff expansions is valid on the relevant function space.
    Used in Section 4, Eqs. (29)-(34), to derive H_{2n}(r^2)=exp(-Delta/4)(2r)^{2n}; no convergence or domain conditions are stated.
  • domain assumption Grad's multivariate Hermite polynomials and their scalar contraction are defined as in [7,8].
    Definitions (23)-(25) supply the scalar Hermite polynomials used in the 3D inverse-kernel formulas.
  • standard math Known special-function identities: Rodrigues formula, the Laguerre-Hermite relation in Eq. (43), and the Legendre duplication formula.
    Used in Section 4 to reduce scalar Hermite polynomials to ordinary Hermite polynomials; cited to [6,14,17].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deconvolution of 3-D Gaussian kernels." pith.science (2026). https://pith.science/paper/4IY65Y7O

@misc{pith2026190807259,
  author       = {Pith},
  title        = {Pith review of: Deconvolution of 3-D Gaussian kernels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4IY65Y7O}},
  note         = {Machine review of arXiv:1908.07259}
}
read the original abstract

Ulmer and Kaissl formulas for the deconvolution of one-dimensional Gaussian kernels are generalized to the three-dimensional case. The generalization is based on the use of the scalar version of the Grad's multivariate Hermite polynomials which can be expressed through ordinary Hermite polynomials.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 22 canonical work pages

  1. [1]

    Ulmer and W

    W. Ulmer and W. Kaissl, The inverse problem of a Gaussian convo- lution and its application to the finite size of the measurement cham- bers/detectors in photon and proton dosimetry, Phys. Med. Biol. 48, 707–727 (2003)

  2. [2]

    W. Ulmer, Inverse problem of linear combinations of Gaussian conv o- lution kernels (deconvolution) and some applications to proton/pho ton dosimetry and image processing, Inverse Prob. 26, 085002 (2010)

  3. [3]

    W. D. Newhauser and R. Zhang, The physics of proton therapy, Phys. Med. Biol. 60, R155–R209 (2015). 10

  4. [4]

    Ulmer and A

    W. Ulmer and A. Carabe-Fernandez, Aspects of the physical pr inciples of the proton therapy with inclusion of nuclear interactions, J. Pro ton Ther. 3, 311 (2017)

  5. [5]

    R. G. Hohlfeld, J. I. F. King, T. W. Drueding and G. V. H. Sandri, Solution of Convolution Integral Equations by the Method of Differe ntial Inversion, SIAM J. Appl. Math. 53, 154–167 (1993)

  6. [6]

    Szeg¨ o,Orthogonal Polynomials, 4th edition (American Mathematical Society, Providence, 1975)

    G. Szeg¨ o,Orthogonal Polynomials, 4th edition (American Mathematical Society, Providence, 1975)

  7. [7]

    Grad, Note on N-dimensional hermite polynomials, Commun

    H. Grad, Note on N-dimensional hermite polynomials, Commun. Pur e Appl. Math. 2 (1949) 325–330

  8. [8]

    Holmquist, The d-Variate Vector Hermite Polynomial of Order k , Linear Algebra Appl

    B. Holmquist, The d-Variate Vector Hermite Polynomial of Order k , Linear Algebra Appl. 237-238, 155–190 (1996)

Show all 24 references
  1. [9]

    Balescu, Transport Processes in Plasmas(Elsevier Science Publishers, Amsterdam, 1988)

    R. Balescu, Transport Processes in Plasmas(Elsevier Science Publishers, Amsterdam, 1988)

  2. [10]

    Pfefferl´ e, E

    D. Pfefferl´ e, E. Hirvijoki and M. Lingam, Exact collisional mome nts for plasma fluid theories, Phys. Plasmas 24, 042118 (2017)

  3. [11]

    De Bie, An Alternative Definition of the Hermite Polynomials Relat ed to the Dunkl Laplacian, SIGMA 4, 093 (2008)

    H. De Bie, An Alternative Definition of the Hermite Polynomials Relat ed to the Dunkl Laplacian, SIGMA 4, 093 (2008)

  4. [12]

    W¨ unsche, Generating Functions for Products of Special L aguerre 2D and Hermite 2D Polynomials, Appl

    A. W¨ unsche, Generating Functions for Products of Special L aguerre 2D and Hermite 2D Polynomials, Appl. Math. 6, 61552 (2015)

  5. [13]

    J. D. Bjorken and S. D. Drell, Relativistic Quantum Mechanics (McGraw-Hill, New York, 1964)

  6. [14]

    W. W. Bell, Special Functions for Scientists and Engineers (D. Van Nostrand Company, London, 1968)

  7. [15]

    F. M. Fern´ andez, Generating functions for Hermite polynomia ls of ar- bitrary order, Phys. Lett. A 237, 189–191 (1998)

  8. [16]

    W¨ unsche, Hermite and Laguerre 2D polynomials, J

    A. W¨ unsche, Hermite and Laguerre 2D polynomials, J. Comput. Appl. Math. 133, 665–678 (2001). 11

  9. [17]

    G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists, 6th edition (Elsevier Academic Press, Amsterdam, 2005)

  10. [18]

    H-Yi and Z

    F. H-Yi and Z. D.-Hui, New generating function formulae of even - and odd-Hermite polynomials obtained and applied in the context of quan- tum optics, Chinese Phys. B 23, 060301 (2014)

  11. [19]

    Saitoh and Y

    S. Saitoh and Y. Sawano, Theory of Reproducing Kernels and Applica- tions (Springer, Singapure, 2016)

  12. [20]

    J. L. Mueller and S. Siltanen, Linear and Nonlinear Inverse Problems with Practical Applications (SIAM, Philadelphia, 2012)

  13. [21]

    Kosugi, Inverse Integral Kernel for Diffusion in a Harmonic P otential, J

    T. Kosugi, Inverse Integral Kernel for Diffusion in a Harmonic P otential, J. Phys. Soc. Jap. 83, 054002 (2014)

  14. [22]

    Maziashvili and Z

    M. Maziashvili and Z. K. Silagadze, Non-local imprints of gravity o n quantum theory, arXiv:1812.03993 [quant-ph]

  15. [23]

    Calcagni, L

    G. Calcagni, L. Modesto and G. Nardelli, Non-perturbative spec trum of non-local gravity, Phys. Lett. B 795, 391-397 (2019)

  16. [24]

    Calcagni, Taming the Beast: Diffusion Method in Nonlocal Gravit y, Universe 4, no

    G. Calcagni, Taming the Beast: Diffusion Method in Nonlocal Gravit y, Universe 4, no. 9, 95 (2018); Erratum: [Universe 5, no. 5, 121 (2019)] 12

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.