Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging

Andrea Sebastiani; Davide Evangelista; Elena Loli Piccolomini; Elena Morotti

REVIEW 4 major objections 4 minor 1 cited by

Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A neural-net weight guess gives few-view CT a provably stable TV model

desk verdict Neat learned-weights TV scheme for few-view CT, but the central uniqueness proof leans on a mischaracterization of the subdifferential of isotropic TV at zero-gradient pixels. read the letter →

arxiv 2501.09845 v1 pith:LWYZ5K5Q submitted 2025-01-16 math.NA cs.NAmath.OC

Elena Morotti , Davide Evangelista , Andrea Sebastiani , Elena Loli Piccolomini This is my paper · ORCID

classification math.NAcs.NAmath.OC MSC 65J2268T0794A0865K10

keywords WeightedTotalVariationSpatiallyAdaptiveRegularizationFew-viewTomographyNeuralNetworksImagereconstructionWell-posednessInverseproblems

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 proposes a way to make total-variation regularization spatially adaptive in few-view CT without recomputing weights at every iteration: a neural network turns a fast coarse reconstruction (filtered back projection) into an estimate of the true image or of its gradient magnitude, and that estimate fixes pixel-wise weights exactly once. The paper then proves that the resulting model $\Psi$-W $\ell^1$ is a well-posed regularization method: existence and uniqueness of the minimizer, noise stability as $\delta\to 0$, and stability with respect to the reconstructor $\Psi$, under the standard condition $\ker(K)\cap\ker(D)=\{0\}$. It further proves that if the network output converges to the ground truth in image norm or gradient-magnitude norm, the final solution converges to the solution obtained by using the ground truth itself to set the weights. A reader should care because the guarantee attaches the reliability of variational regularization to a learned intermediate image: the network only proposes weights, and the convex model supplies the stable reconstruction.

What carries the argument

The engine of the method is the fixed-weight functional with weights $w(\tilde{x})_i=(\eta^p/(\eta^2+|D\tilde{x}|_i^2))^{1-p}$ for $0<p<1$, computed once from a reconstructor $\tilde{x}=\Psi(y_\delta)$. The weights are high (equal to 1) in flat regions and smaller near edges, so the weighted total-variation term $\|w(\Psi(y_\delta))\odot |Dx|\|_1$ penalizes detail-rich pixels less. On the theoretical side, the proof machinery combines the primal-dual algorithm used to compute solutions with compactness of the functional's sublevel sets for stability and Bregman-distance comparisons (a convex-analytic separation measure) for uniqueness; Theorem 4 and Corollaries 1-2 make the map from the intermediate estimate to the final minimizer continuous, which is what converts network approximation error into reconstruction error.

What would settle it

Compute the subdifferential of the isotropic total variation term at an image with at least one zero-gradient pixel and check whether every vector in the unit disk is allowed; finding a subgradient outside the diagonal set $(c,c)$ would show the formula in equations (3.6)-(3.7) is false as stated, so the uniqueness theorem would need a different proof.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that the $\Psi$-W $\ell^1$ model (2.7) is a well-posed regularization method whose solution tracks the ideal ground-truth-weighted solution when the intermediate estimate is accurate. For every Lipschitz reconstructor $\Psi$, every $\lambda>0$, and every noise level $\delta\ge 0$, the objective admits at least one minimizer; under the condition $\ker(K)\cap S(x^*_{\Psi,\delta})=\{0\}$ and the subgradient condition $|c|<1$ stated in Theorem 2, the minimizer is unique. The solution depends continuously on the noise level and on the reconstructor, and Corollary 2 transfers convergence of $\Psi(y_\delta)$ to $x^{\mathrm{GT}}$ (in image or gradient-magnitude norm) into convergence of $x^*_{\Psi,\delta}$ to the oracle solution $x^*_{\mathrm{GT},\delta}$. The consequence the authors emphasize is that neural-network outputs become interpretable: the network never forms the final image, it only fixes the regularization weights, and the variational stage provides stability and convergence.

Load-bearing premise

The load-bearing premise is the paper's formula for the subdifferential of isotropic total variation at pixels where the image gradient is zero; if the true set of subgradients there is larger than the diagonal set $(c,c)$ used in equation (3.6), the Bregman-distance contradiction that proves uniqueness does not go through.

Editorial extensions

If this is right

Because the global isotropic TV model is the special case with all weights equal to one, the uniqueness theorem also supplies an extension of uniqueness results from anisotropic to isotropic TV under the stated kernel conditions.
Practitioners can fix weights after a single network pass and still receive a reconstruction that is stable as measurement noise tends to zero, which is exactly the property that iterative reweighting schemes often sacrifice for speed.
The final image inherits variational stability even when the neural network is fed out-of-distribution noisy sinograms, because the network is not the terminal reconstructor.
Using the gradient-magnitude loss is at least as well justified as using the image loss, since Corollaries 1 and 2 require only closeness of gradient magnitudes to transfer accuracy.
If a reconstructor approximates the ground truth ever better, the resulting solutions converge to the oracle-weighted solution, giving a formal sense in which learned intermediate images can be trusted.

Reading between the lines

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

The theory does not require the intermediate estimator to be trained with paired ground truths; it only needs a Lipschitz estimator of the image or its gradient magnitude, so unsupervised or classical edge-preserving preprocessors should inherit the same stability guarantees if they can control the Lipschitz error.
The stability estimates suggest a quantitative calibration experiment: measure the gradient-magnitude error of a network on held-out noise levels and check whether the final reconstruction error scales linearly with that error, as Corollary 2 would predict.
The same fixed-weight trick could be applied to wavelet or shearlet sparsity regularizers, provided the network estimates the transform-domain magnitudes; the convexity and Lipschitz-continuity properties used in the proofs are not specific to the gradient operator.

Share X Bluesky LinkedIn Reddit HN

Formalized claims in Lean

Claim #1: On its own terms, the paper's central claim is that the $\Psi$-W $\ell^1$ model (2.7) is a well-posed regularization method whose solution tracks the ideal ground-truth-weighted solution when the intermediate estimate is accurate. For every Lipschitz reconstructor $\Psi$, every $\lambda>0$, and every noise level $\delta\ge 0$, the objective admits at least one minimizer; under the condition $\ker(

/-- @claim 1 On its own terms, the paper's central claim is that the $\Psi$-W $\ell^1$ model (2.7) is a well-posed regularization method whose solution tracks the ideal ground-truth-weighted solution when the intermediate estimate is accurate. For every Lipschitz reconstructor $\Psi$, every $\lambda>0$, and every noise level $\delta\ge 0$, the objective admits at least one minimizer; under the condition $\ker( -/
def central_claim : Prop :=

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Thanks for the report. I think the reader's basic verdict is right: the paper has a neat computational idea, but the theoretical proof collapses at Lemma 3. I checked the proof in Appendix A.2. The subdifferential of the Euclidean norm at zero is the whole unit disk, so restricting the subgradient to (c,c) is false. More importantly, that diagonal restriction is what makes the strict inequality in Eq. (3.15) work. Without it, the Bregman-distance argument in Theorem 2 does not go through, and since uniqueness feeds directly into Theorems 3 and 4 and Corollary 2, the well-posedness claims are not established. This is a load-bearing error, not a cosmetic one.

That said, the underlying idea is worth attention. Fixing the weights from a learned intermediate image, rather than iteratively reweighting, is a simple and sensible way to get adaptivity in few-view CT. The experiments on COULE and Mayo show consistent gains over global TV and the two IR baselines, and the out-of-distribution noise test is a reasonable robustness check. The authors also deserve credit for being explicit that the gradient loss can work even when the intermediate image itself is poor.

The soft spots beyond the theory: in Section 4.1 they tune lambda, p, and eta on the test image and then report results on that same image. That makes the gains look stronger than they are. And Corollary 2 is mostly a consistency check: the training losses in (2.12) and (2.13) minimize exactly the quantities the corollary assumes go to zero, so the empirical agreement is expected, not an independent prediction.

Who is this for? People working on learned regularizers for inverse problems, especially CT. The empirical framework is useful and the writing is clear. The theory is currently wrong, though. I would send this to peer review — a good convex-analysis referee will catch the subdifferential error immediately — but my recommendation to the editor would be major revision at best. If the authors can fix Lemma 3, or scale back the claims to avoid uniqueness, the paper becomes a solid applied contribution.

Referee Report

4 major / 4 minor

Summary. The paper proposes a spatially adaptive weighted total variation model, denoted Ψ-Wℓ1, for few-view CT reconstruction. The weights are computed once from an intermediate image produced by a neural network (or from the magnitude of its gradient), avoiding iterative reweighting. The authors state and prove four theoretical properties—existence, uniqueness, noise stability, and reconstructor stability—and derive Corollary 2 linking the accuracy of the learned reconstructor to the accuracy of the final reconstruction. Numerical experiments on a synthetic COULE image and on Mayo Clinic chest data compare the method with global TV and with two iterative reweighted ℓ1 strategies.

Significance. Adaptive weighted TV is a practically important topic, and the idea of fixing the weights in advance from a learned intermediate reconstruction is computationally attractive. If the theoretical results were correct, the paper would provide a useful bridge between learned priors and classical regularization theory. The authors are also transparent in pointing out that the assumptions of Corollary 2 coincide with the training losses. The out-of-distribution noise experiment is a valuable addition. However, the central theoretical contribution is undermined by an incorrect subdifferential characterization, and the numerical evaluation contains a test-set tuning issue. As a result, the main claims are not presently supported.

major comments (4)

[Appendix A.2, Lemma 3 (Eqs. (3.6)–(3.7))] The claimed characterization of the subdifferential of the isotropic TV term is incorrect. For the term ∥U_i Dx∥_2 with U_iDx = ((D_h x)_i, (D_v x)_i), the subdifferential of the Euclidean norm at the origin is the full unit disk {v ∈ R^2 : ∥v∥_2 ≤ 1}, not the diagonal set {(c,c) : c ∈ [-1,1]}. For example, (1,0) is an admissible subgradient at a zero-gradient pixel but is not of the form (c,c), while (c,c) with |c| > 1/√2 has norm larger than 1 and is not admissible. Consequently Eq. (3.7) does not describe ∂TV(x), and all subsequent uses of this formula are unsupported.
[Theorem 2 (Section 3), Eq. (3.15)] Theorem 2 is load-bearing for Theorems 3 and 4 and Corollary 2, since those results identify limit points with the unique minimizer of the limit problem. The uniqueness proof depends on the strict inequality in Eq. (3.15), which uses the assumption that at a zero-gradient pixel the subgradient has the form (c,c) with |c| < 1. With the correct subdifferential, a subgradient at such a pixel can have Euclidean norm exactly one, for instance (1,0), so the strict inequality R(x') < R(x') is not guaranteed. The paper does not prove that an optimality subgradient satisfying Eq. (3.9) can always be chosen with norm strictly less than one at every pixel where x' has nonzero gradient and x* has zero gradient. Thus the uniqueness claim, and consequently the noise stability and reconstructor stability theorems, are unproven.
[Section 4.1] All model and algorithmic parameters, including λ, p, η, and the comparison methods' parameters, are tuned to minimize the RE metric on the final solutions, and the same test image is then used to report the RE values in Table 2. This means the reported improvements over global TV and the IRℓ1 baselines are partially the result of test-set optimization. No independent validation set or cross-validation procedure is described. The empirical claim that the proposed method outperforms competing approaches is therefore weakened, independently of the theoretical issues.
[Corollary 2 (Section 3)] The paragraph immediately following Corollary 2 states that the quantities minimized in assumptions 1 and 2 are the same, up to the norm, as those minimized by the training losses (2.12) and (2.13). This makes the corollary a consistency statement about the training objective rather than an independent predictive guarantee for a trained network. Moreover, assumptions 1 and 2 use the ℓ1 norm while the losses use squared ℓ2 norms, so the equivalence is only qualitative. This limitation should be stated explicitly; the current framing overstates what the theorem adds beyond the training loss itself.

minor comments (4)

[Section 2.1] The text says the relative error plots are 'consistent with the behavior discussed in Table 3', but the synthetic experiment is reported in Table 1; Table 3 concerns the Mayo dataset.
[Section 4.2] The final paragraph contains the incomplete sentence 'FBP-GNet-W ℓ1 outperforms FBP-Net-W ℓ1, fewer noise artifacts'; it should read '... with fewer noise artifacts' or be rephrased.
[Section 3, Eqs. (3.4) and (3.20)] The letter M is used both for the set of minimizers in Eq. (3.4) and for the constant bounding the objective in the proofs of Theorems 3 and 4; using different symbols would avoid confusion.
[Appendix A.2] In the proof of Lemma 3, the notation w is reused for the vector U_i^T v, although w already denotes the adaptive weight vector from Section 2; a different symbol, such as g^{(i)}, would be clearer.

Circularity Check

1 steps flagged · score 3.0 of 10

Corollary 2's guarantee is keyed to the network training losses, so the corresponding empirical confirmation is a consistency check; the variational well-posedness proof itself is not circular.

self definitional [Section 3, Corollary 2 and the paragraph immediately following; Section 2.2, Eqs. (2.12)-(2.13)]
"We emphasize that the quantities minimized in assumptions 1 and 2 of the previous corollary are the same, up to the norm, as those minimized in the loss functions (2.12) and 2.13, respectively."

Corollary 2 assumes either ||Psi_k(y^delta) - x_GT||_1 -> 0 or || |D Psi_k(y^delta)| - |D x_GT| ||_1 -> 0. In finite dimensions and up to equivalent norms, these are exactly the residuals minimized by the supervised training losses (2.12) and (2.13). The network is therefore fitted to satisfy the corollary's antecedent. The later experimental demonstration that an accurate network yields x*_{Psi,delta} close to x*_{GT,delta} verifies that the training objective was met and that the conditional stability theorem applies; it is not an independent test of a first-principles prediction. The theorem itself is a genuine continuity statement, so the circularity is partial and confined to the learning-theoretic narrative.

full rationale

Theorems 1-4 are derived by convex analysis (coercivity, subdifferential calculus, Bregman distances) and do not reduce to the fitted network outputs; the well-posedness claim has independent mathematical content. The one near-circular element is Corollary 2, whose assumptions are the same quantities minimized by the network losses; the authors themselves call this out, and the numerical confirmation of that corollary is therefore a consistency check. I did not score the apparent error in Lemma 3's subdifferential formula for isotropic TV at zero-gradient pixels as circularity: it is a correctness risk that would undermine the proof of Theorem 2, not a reduction of the result to its inputs. Similarly, tuning lambda, p, and eta on the test images to minimize the reported RE is an evaluation weakness, not a circular derivation. No load-bearing self-citation chain was found; citations to the authors' prior work are used for terminology or supporting observations.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

No new physical or mathematical entities are introduced beyond the computational devices already listed. The neural network and the weight rule are tools, not new ontological objects. The main load-bearing assumptions are the kernel intersection condition, the ad hoc weight formula, and the unverified uniqueness hypotheses.

free parameters (3)

lambda (regularization weight) = 0.8 to 12 depending on dataset and noise
Controls trade-off between fidelity and regularization; tuned on the test image to minimize RE (Section 4.1).
p = 0.3
Exponent in weight rule (2.6); set heuristically; controls how quickly weights decrease with gradient magnitude.
eta = 2e-5 or 2e-3
Stabilizer in weight rule (2.6); avoids division by zero and sets gradient scale; chosen heuristically per experiment.

assumptions (5)

domain assumption ker(K) intersect ker(D) = {0}
Assumed for coercivity, existence, and uniqueness (Assumption A.1, Section 3). The authors argue it holds for CT with non-negative images since K has non-negative entries.
ad hoc to paper Weight formula (2.6) with eta and p is the chosen model for adaptivity
The weights are introduced as an analog of iterative reweighting (2.5) but fixed; no derivation from first principles, only desired properties (Proposition 1).
domain assumption Psi in Rec_{m,n} is Lipschitz continuous
All theoretical results assume the reconstructor is Lipschitz (Section 2, definition of Rec_{m,n}); not verified for the trained U-Net.
domain assumption Theorem 2 hypotheses: ker(K) intersect S(x*) = {0} and |c| < 1
The uniqueness theorem only holds under these conditions on the minimizer; the paper does not verify them in experiments.
domain assumption Uniform convergence of reconstructors on Y^delta in Theorem 4
Reconstructor stability assumes sup over Y^delta goes to 0; Y^delta may be unbounded for a cone-shaped KX, making this strong.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging." pith.science (2026). https://pith.science/paper/LWYZ5K5Q

@misc{pith2026250109845,
  author       = {Pith},
  title        = {Pith review of: Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LWYZ5K5Q}},
  note         = {Machine review of arXiv:2501.09845}
}

read the original abstract

This study presents the development of a spatially adaptive weighting strategy for Total Variation regularization, aimed at addressing under-determined linear inverse problems. The method leverages the rapid computation of an accurate approximation of the true image (or its gradient magnitude) through a neural network. Our approach operates without requiring prior knowledge of the noise intensity in the data and avoids the iterative recomputation of weights. Additionally, the paper includes a theoretical analysis of the proposed method, establishing its validity as a regularization approach. This framework integrates advanced neural network capabilities within a regularization context, thereby making the results of the networks interpretable. The results are promising as they enable high-quality reconstructions from limited-view tomographic measurements.

Figures

Figures reproduced from arXiv: 2501.09845 by the authors.

**Figure 1.** A plot of (w(x˜))i for different values of η, over (|Dx˜|)i , for p = 0.3. To provide a more precise specification on how to derive x˜ from y δ we consider Ψ : R m → R n as a Lipschitzcontinuous function that maps y δ to an approximate reconstruction x˜ = Ψ(y δ ) of x GT . In accordance with the nomenclature introduced in [25], we denote such mappings as reconstructors, and we name as Recm,n the set of such reconst… view at source ↗

**Figure 2.** Workflow of the considered scheme, where the reconstructor [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Results of the experiment on the synthetic image with higher noise ( [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: The reconstructor Ψ when it is constituted by two steps: a Filtered Back Projection and a neural network. θ ∗ ∈ arg min θ X ND j=1 ||x GT j − Ψθ(Ψ( ˜ y δ j ))||2 2 . (2.12) The choice of the second loss function is motivated by the idea of ”learning” the image gradient…

**Figure 5.** Figure 5: Images used as ground truth in the numerical experiments with some zoom-ins remarking regions of interest. [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

**Figure 6.** Figure 6: From left to right: crops of the solution images computed (in case [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]

**Figure 7.** Figure 7: Plots of the relative errors of the solutions computed for the Mayo test image depicted in Figure 5. In both [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]

**Figure 8.** Figure 8: Results on the Mayo real medical image for the [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Learnable Total Variation with Lambda Mapping for Low-Dose CT Denoising
cs.CV 2025-11 conditional novelty 5.0 of 10

A learnable per-pixel regularization map inside an unrolled total-variation solver improves low-dose CT denoising by ~3.7 dB PSNR over FBP+U-Net.

Reference graph

Works this paper leans on

46 extracted references · 45 canonical work pages · cited by 1 Pith paper

[1]

Rudin L I, Osher S and Fatemi E 1992 Physica D: nonlinear phenomena 60 259–268

1992
[2]

Sidky E Y and et al 2014 IEEE Journal of Translational Engineering in Health and Medicine

work page 2014
[3]

Piccolomini E L and Morotti E 2021 Journal of Imaging 7

work page 2021
[4]

2022 Physics in Medicine & Biology 67 205010

Friot L, Peyrin F, Maxim V et al. 2022 Physics in Medicine & Biology 67 205010

work page 2022
[5]

Chan R H, Kan K K, Nikolova M and Plemmons R J 2020 Journal of Mathematical Imaging and Vision 62 790–807

work page 2020
[6]

Pragliola M, Calatroni L, Lanza A and Sgallari F 2023 SIAM Review 65 601–685

work page 2023
[7]

Dong Y , Hinterm ¨uller M and Rincon-Camacho M M 2011 Journal of Mathematical Imaging and Vision 40 82–104

work page 2011
[8]

Hinterm ¨uller M, Rautenberg C N, Wu T and Langer A 2017 Journal of Mathematical Imaging and Vision 59 515–533

work page 2017

Show all 46 references

[9]

Bortolotti V , Brown R, Fantazzini P, Landi G and Zama F 2016Inverse Problems 33 015003
[10]

Cascarano P, Franchini G, Kobler E, Porta F and Sebastiani A 2023 Computational Optimization and Applica- tions 84 125–149

2023
[11]

Kan K, Fung S W and Ruthotto L 2021 SIAM Journal on Scientific Computing 43 S704–S726

2021
[12]

Grasmair M 2009 Locally adaptive total variation regularization International Conference on Scale Space and Variational Methods in Computer Vision(Springer) pp 331–342

2009
[13]

2010 Image and vision computing 28 298–306

Chen Q, Montesinos P, Sun Q S, Heng P A et al. 2010 Image and vision computing 28 298–306

2010
[14]

Bubba T A, Calatroni L, Catozzi A, Crisci S, Pock T, Pragliola M, Rautio S, Riccio D and Sebastiani A 2022 Bilevel learning of regularization models and their discretization for image deblurring and super-resolution IN- dAM Workshop: Advanced Techniques in Optimization for Mac...

2022
[15]

Cuomo S, De Rosa M, Izzo S, Piccialli F and Pragliola M 2023 Applied Numerical Mathematics

2023
[16]

Kofler A, Altekr ¨uger F, Antarou Ba F, Kolbitsch C, Papoutsellis E, Schote D, Sirotenko C, Zimmermann F F and Papafitsoros K 2023 SIAM Journal on Imaging Sciences 16 2202–2246

2023
[17]

Pourya M, Neumayer S and Unser M 2024 Numerical Functional Analysis and Optimization 45 411–440

2024
[18]

Huang Y , Taubmann O, Huang X, Haase V , Lauritsch G and Maier A 2018IEEE Transactions on Radiation and Plasma Medical Sciences 2 307–314
[19]

Xi Y , Zhou P, Yu H, Zhang T, Zhang L, Qiao Z and Liu F 2023Medical Physics 50 5568–5584
[20]

Luo F, Li W, Tu W and Wu W 2018 IEEE Access 6 64225–64236

2018
[21]

Cand `es E J, Wakin M B and Boyd S P 2008 Journal of Fourier analysis and applications

2008
[22]

Daubechies I, DeV ore R, Fornasier M and G¨unt¨urk C S 2010Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences 63 1–38
[23]

Lazzaro D, Piccolomini E L and Zama F 2019 Inverse Problems 35 084002

2019
[24]

Scherzer O, Grasmair M, Grossauer H, Haltmeier M and Lenzen F 2009 Variational methods in imagingvol 167 (Springer) 19 Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging A PREPRINT

2009
[25]

Evangelista D, Nagy J, Morotti E and Piccolomini E L 2022 arXiv preprint arXiv:2211.13692

2022 arXiv
[26]

Chambolle A and Pock T 2011 Journal of mathematical imaging and vision 40 120–145

2011
[27]

Bauschke H H, Combettes P L, Bauschke H H and Combettes P L 2017Convex Analysis and Monotone Operator Theory in Hilbert Spaces (Springer)
[28]

Kak A C and Slaney M 2001 Principles of computerized tomographic imaging (SIAM)

2001
[29]

Wang Z, Bovik A C, Sheikh H R and Simoncelli E P 2004 IEEE transactions on image processing 13 600–612

2004
[30]

Morotti E, Evangelista D and Loli Piccolomini E 2021 Journal of Imaging 7 139

2021
[31]

Evangelista D, Morotti E, Piccolomini E L and Nagy J 2023Journal of Imaging9 ISSN 2313-433X URL https: //www.mdpi.com/2313-433X/9/7/133
[32]

Bertero M, Boccacci P and De Mol C 2021 Introduction to inverse problems in imaging(CRC press)

2021
[33]

Jørgensen J S, Kruschel C and Lorenz D A 2015 Inverse Problems in Science and Engineering 23 1283–1305

2015
[34]

Bianchi D, Evangelista D, Aleotti S, Donatelli M, Piccolomini E L and Li W 2023 arXiv preprint arXiv:2312.16936

2023 arXiv
[35]

McCollough C 2016 Medical physics 43 3759–3760

2016
[36]

Deng L, Mi D, He P, Feng P, Yu P, Chen M, Li Z, Wang J and Wei B 2015Bio-Medical Materials and Engineering 26 S1685–S1693
[37]

Wu W, Hu D, Cong W, Shan H, Wang S, Niu C, Yan P, Yu H, Vardhanabhuti V and Wang G 2022 Patterns 3 100474

2022
[38]

Wu W, Hu D, Cong W, Shan H, Wang S, Niu C, Yan P, Yu H, Vardhanabhuti V and Wang G 2022 Patterns 3 100475

2022
[39]

Loli Piccolomini E, Prato M, Scipione M and Sebastiani A 2023 Algorithms 16 270

2023
[40]

Evangelista D, Morotti E and Piccolomini E L 2023 Computerized Medical Imaging and Graphics 103 102156

2023
[41]

Ongie G, Jalal A, Metzler C A, Baraniuk R G, Dimakis A G and Willett R 2020 IEEE Journal on Selected Areas in Information Theory 1 39–56

2020
[42]

Fondo per il Programma Nazionale di Ricerca e Progetti di Rilevante Interesse Nazionale (PRIN)

Beck A 2017 First-order methods in optimization (SIAM) Acknowledgement E. Loli Piccolomini, D. Evangelista and E. Morotti are supported by the “Fondo per il Programma Nazionale di Ricerca e Progetti di Rilevante Interesse Nazionale (PRIN)” 2022 project “STILE: Sustainable Tomo...

2017
[43]

xk ∈ ker(K)c, which implies that ||Kx k − yδ||2 2 ≥ f1(||xk||2), where f1(||xk||2) → ∞as k → ∞,
[44]

Since JΨ,δ(xk) ≥ min{||Kxk − yδ||2 2, RΨ,δ(xk)}, it implies that: JΨ,δ(xk) ≥ min{f1(||xk||2), f2(||xk||2)} → ∞, k → ∞, (A.1) concluding the proof

xk ∈ ker(W Ψ,δD)c, which implies that RΨ,δ(xk) ≥ f2(||xk||2), where f2(||xk||2) → ∞as k → ∞, 20 Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging A PREPRINT as both ||Kx − yδ||2 2 and RΨ,δ(x) are coercive on ker(K)c and ker(W Ψ,δD...
[45]

By convexity of JΨ,δ(x) and of X , necessarily JΨ,δ(x1) = JΨ,δ(x2) = JΨ,δ( x1+x2 2 ) = J ∗ Ψ,δ

(A.2) Now, let x1, x2 ∈ M. By convexity of JΨ,δ(x) and of X , necessarily JΨ,δ(x1) = JΨ,δ(x2) = JΨ,δ( x1+x2 2 ) = J ∗ Ψ,δ. Therefore: J ∗ Ψ,δ ≤ 1 2 (J ∗ Ψ,δ + J ∗ Ψ,δ) − 1 8 ||Kx 1 − Kx 2||2 2 ⇐ ⇒ J∗ Ψ,δ ≤ J∗ Ψ,δ − 1 8 ||Kx1 − Kx2||2 2 ⇐ ⇒ ||Kx1 − Kx 2||2 2 ≤ 0 ⇐ ⇒Kx 1 = Kx 2,...
[46]

(A.5) 21 Adaptive Weighted Total Variation boosted by learning techniques in few-view tomographic imaging A PREPRINT Similarly, considering ||w(Ψ(yδ1 ) ⊙ |Dx| ||1 = ||w(Ψ(yδ2 ) ⊙ |Dx| + (w(Ψ(yδ1 )) − w(Ψ(yδ2 ))) ⊙ |Dx| ||1 with p = 1: ||w(Ψ(yδ1 ) ⊙ |Dx| ||1 ≤ ||w(Ψ(yδ2 ) ⊙ |Dx...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Rudin L I, Osher S and Fatemi E 1992 Physica D: nonlinear phenomena 60 259–268

1992

[2] [2]

Sidky E Y and et al 2014 IEEE Journal of Translational Engineering in Health and Medicine

work page 2014

[3] [3]

Piccolomini E L and Morotti E 2021 Journal of Imaging 7

work page 2021

[4] [4]

2022 Physics in Medicine & Biology 67 205010

Friot L, Peyrin F, Maxim V et al. 2022 Physics in Medicine & Biology 67 205010

work page 2022

[5] [5]

Chan R H, Kan K K, Nikolova M and Plemmons R J 2020 Journal of Mathematical Imaging and Vision 62 790–807

work page 2020

[6] [6]

Pragliola M, Calatroni L, Lanza A and Sgallari F 2023 SIAM Review 65 601–685

work page 2023

[7] [7]

Dong Y , Hinterm ¨uller M and Rincon-Camacho M M 2011 Journal of Mathematical Imaging and Vision 40 82–104

work page 2011

[8] [8]

Hinterm ¨uller M, Rautenberg C N, Wu T and Langer A 2017 Journal of Mathematical Imaging and Vision 59 515–533

work page 2017