pith. sign in

arxiv: 2605.00029 · v1 · submitted 2026-04-22 · 📡 eess.IV · cs.CV· physics.optics

Broadband Wide Field of View Imaging with Computational Mirrors

Vishwanath Saragadam , Niki Nezakati , Amit Roy-Chowdhury , Vivek Boominathan This is my paper

Pith reviewed 2026-05-09 22:29 UTC · model grok-4.3

classification 📡 eess.IV cs.CVphysics.optics
keywords computational imagingmirror opticsbroadband imagingVIS-SWIRpoint spread functionfocal stackoff-axis aberrations
0
0 comments X

The pith

A concave mirror plus 2-4 images and a new PSF model produces sharp all-in-focus VIS-SWIR images with one sensor.

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

The paper shows that traditional refractive lenses cannot focus the full visible-to-shortwave-infrared range at once, but a simple concave mirror can be made to work across that spectrum by taking a short focal stack and then computationally undoing the mirror's off-axis aberrations. The central mechanism is SeidelConv, a spatially varying PSF model derived from Seidel aberration theory that predicts how blur changes across the field of view. With this model the system reconstructs a single sharp image from only two to four captures, removing the need to refocus or use complex multi-element glass optics. If the approach holds, wide-field broadband imaging becomes feasible with inexpensive mirrors and existing VIS-SWIR sensors, revealing material properties that are invisible in any single spectral band.

Core claim

By capturing a minimal focal stack of 2-4 images and applying SeidelConv, a physics-inspired spatially-varying PSF model, the framework recovers a sharp, all-in-focus image across the complete VIS-SWIR spectrum using a single sensor and simple concave mirror.

What carries the argument

SeidelConv, a spatially-varying point-spread-function model that predicts and corrects off-axis aberrations of concave mirrors by incorporating Seidel aberration terms.

Load-bearing premise

That the SeidelConv model accurately captures the off-axis aberrations of the specific concave mirrors used and that a focal stack of only 2-4 images is always sufficient to recover high-resolution detail without artifacts across the full field of view and spectrum.

What would settle it

Acquire a focal stack with a different mirror curvature or a larger number of planes and check whether the SeidelConv reconstruction matches the sharpness and detail of a ground-truth all-in-focus image captured with a reference refractive system.

Figures

Figures reproduced from arXiv: 2605.00029 by Amit Roy-Chowdhury, Niki Nezakati, Vishwanath Saragadam, Vivek Boominathan.

Figure 1
Figure 1. Figure 1: Computational mirrors. (a) Traditional refractive (glass) optics require com￾plex, multi-element assemblies that are inherently bulky and prone to chromatic aber￾ration across the VIS-SWIR spectrum. In contrast, reflective optics are achromatic but suffer from severe off-axis geometric aberrations. (b) Our approach captures a sparse focal stack using simple mirror optics and employs a multi-image deconvolu… view at source ↗
Figure 2
Figure 2. Figure 2: Challenges and opportunities with mirrors. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Computational mirrors hardware. (a) Mirror optics enable imaging with￾out any chromatic aberration but suffer from severe field curvature. To tackle this, we leverage a computational reconstruction approach by measuring a sparse focal stack that spans the whole Petzval curved field, and then perform a multi-image deconvo￾lution with our proposed SeidelConv. (b) shows our optical setup with a broadband VIS-… view at source ↗
Figure 4
Figure 4. Figure 4: SeidelConv architecture. (a) shows SeidelConv architecture described in eq. (2) for learning the forward operator for a single wavelength range. The input image is warped by an affine (inspired by Seidel aberrations), and each warped image is blurred and weighed in a per-pixel manner and summed to get the output. (b) shows three images captured with our optical setup for the green wavelengths, and the deco… view at source ↗
Figure 5
Figure 5. Figure 5: Accuracy of PSF calibration. The first column shows the calibration setup, where we displayed an image with sparse and random dots on an OLED monitor. The next three columns shows a random set of dots for three focus settings, (top row) measured by the optical system and (bottom row) output by SeidelConv. The model accurate captures the highly varying PSF with an error of 4 × 10−3 , enabling accurate decon… view at source ↗
Figure 6
Figure 6. Figure 6: Comparison against other PSF models. The figure above shows deconvo￾lution results with various PSF models including SeidelConv, CoordGate [24] that does not include warping, and two variants of patch-wise convolution with two kernel sizes. The table reports PSNR and SSIM evaluated on-axis (green box) and off-axis (red and blue box). All baselines perform similarly on axis, where the PSFs are compact. Howe… view at source ↗
Figure 7
Figure 7. Figure 7: High-resolution VIS-SWIR multispectral imaging. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Advantages of Mirrors. We compare visible and broadband (400 - 1700nm) images for a (first column) SLR objective, (second column) proposed computational mirrors, and (third column) a plano-convex lens, all at 100mm focal length. The scene consisted of a printed USAF test target illuminated by broadband light. SLR objectives enable very sharp images in visible wavelengths (where they are optimized for), but… view at source ↗
Figure 9
Figure 9. Figure 9: Effect of number of images in focal stack. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Estimating best focus properties. We estimated best focus distance and the range of focal sweep for optimal results. (a) We capture a dense focal stack of a random binary target at 20µ m steps over a 2 mm range, and then use a contrast operator to estimate the best focus distance (b, c). (d) We also estimate a “best focal surface to estimate the range of focal sweep. In our case, 400µ m was optimal for 50… view at source ↗
Figure 11
Figure 11. Figure 11: Monitor-to-sensor calibration. We estimate a homography transformation between monitor and sensor. (a) We display a cross-hair combined with a 25×25 ARuCo tile which has several unique spatial features. The captured image in (b) is then used to perform feature matching, as shown in (c). (d) shows the warped monitor image overlaid over the measured image in (b), showing a near-perfect match in the center o… view at source ↗
Figure 12
Figure 12. Figure 12: Offset and vignetting. Figure (a) shows the image sensor’s dark frame, with fixed pattern noise, which we subtract from every measurement. (b) shows vignetting when the objective is closest to the sensor and (c) shows when the objective is farthest. While similar in relative intensity distribution, we see a change in intensity value, as shown in (d). We collected a one-time vignetting stack and used the a… view at source ↗
Figure 13
Figure 13. Figure 13: Multispectral capture setup. We used red, green, blue absorptive filters, and NIR and SWIR longpass filters. To obtain NIR bandpass images, we illuminated the scene with a compact fluorescent light, which had intensity only till 800 nm. A.4 Multispectral capture [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Effect of focal stack size. Full-image PSNR as a function of the number of focal stack images Nfocus for (a) the f /1 50 mm and (b) the f /2 100 mm mirror objec￾tives. In both cases, performance improves sharply from a single image and saturates at Nfocus = 3 (dashed), beyond which the gains become marginal. 100 mm objectives. We collected red, green, and blue images by displaying in￾dividual channels of … view at source ↗
Figure 15
Figure 15. Figure 15: Computational mirrors dataset. To enable further research on compu￾tational mirrors, we will be releasing a 100-image dataset captured from the DIV2K validation dataset with our optical setup. The figure above shows ground truth, 2 sam￾ple images in the dense focal stack, and reconstruction with SeidelConv. varying warp field, and kernel size. All experiments are conducted on the f /1 50 mm mirror objecti… view at source ↗
Figure 16
Figure 16. Figure 16: Advantages of mirrors and SeidelConv. (a) we compare mirror and planoconvex lens for broadband imaging. Mirrors are inherently broadband, implying that a model trained on green wavelengths can be used for any spectral range. There￾fore, mirrors produce sharp results whether in green wavelengths, or full VIS-SWIR wavelengths. In contrast, planoconvex lenses, when trained on green wavelengths, en￾able sharp… view at source ↗
Figure 17
Figure 17. Figure 17: Effect of SeidelConv hyperparameters. (a) Full-image PSNR vs. number of affine terms Nterms (with kernel size = 11): performance improves rapidly for small values and stabilizes beyond Nterms ≈ 17. We adopt Nterms = 31 (dashed) as a conser￾vative choice within the saturated regime. (b) Full-image PSNR vs. kernel size (with Nterms = 31): performance is stable for small to moderate kernel sizes but degrades… view at source ↗
Figure 18
Figure 18. Figure 18: Prior and SeidelConv. (a) We compare the effect of prior on reconstruction. We observe that PnP enables better reconstruction than a simple total variation (TV) prior, informing the choice in the main paper. (b) We further evaluate the importance of SeidelConv. We used a TV-based reconstruction along with various PSF models. SeidelConv outperforms all other approaches even with a simple TV-based prior. sm… view at source ↗
Figure 19
Figure 19. Figure 19: Effect of PnP solver hyperparameters. (a) Full-image PSNR vs. regu￾larization weight λ (σmin = 10−3 ): reconstruction quality degrades monotonically as λ increases, reflecting excessive penalization of the data term. Performance converges for small λ, and we adopt λ = 10−5 (dashed) as a conservative choice in this stable regime. (b) Full-image PSNR vs. minimum noise level σmin (λ = 10−5 ): performance is … view at source ↗
Figure 20
Figure 20. Figure 20: Wide field-of-view imaging. Here we show that we achieve an MTF30 (contrast ratio of 30%) for a line pair of 0.3 or higher across the whole sensor, empha￾sizing the full usable field of view after computational reconstruction with a 3-image focal stack. References 1. José M Bioucas-Dias, Antonio Plaza, Gustavo Camps-Valls, Paul Scheunders, Nasser Nasrabadi, and Jocelyn Chanussot. Hyperspectral remote sens… view at source ↗
read the original abstract

Traditional glass-based optics are typically optimized for narrow spectral bands, such as the visible (400-700nm) or shortwave infrared (1000-1800nm). While the emergence of VIS-SWIR sensors (400-1700nm) offers transformative potential, refractive optics struggle to focus this entire range simultaneously. Mirrors represent a promising achromatic alternative; however, they are often sidelined by field curvature, and off-axis aberrations. This paper introduces Computational Mirrors, a framework that enables high-resolution, wide-field-of-view imaging across the complete VIS-SWIR spectrum using a single sensor. Our method is built on the observation that distinct regions of the field of view reach focus at varying distances from the mirror. By capturing a minimal focal stack (2-4 images), we utilize a computational backend to recover a sharp, all-in-focus image. A key contribution of this work is SeidelConv, a novel, physics-inspired, spatially-varying point spread function (PSF) model designed to accurately characterize and correct the off-axis aberrations inherent in simple concave mirrors. We demonstrate the efficacy of our approach using a first-of-its-kind 50mm F/1 optical system equipped with a VIS-SWIR sensor. Our system produces sharp images across RGB, NIR, and SWIR wavelengths without requiring refocusing, revealing material details invisible within individual spectral bands. We further validate the scalability of our approach with a 100mm F/2 system optimized for long-range imaging.

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 / 1 minor

Summary. The manuscript introduces Computational Mirrors, a framework for high-resolution broadband imaging across the full VIS-SWIR spectrum (400-1700 nm) using a single sensor and simple concave mirrors. The approach captures a minimal focal stack of 2-4 images exploiting the observation that different field regions focus at varying distances, then applies SeidelConv—a novel physics-inspired spatially-varying PSF model derived from Seidel aberrations—to computationally recover a sharp, all-in-focus image. Efficacy is demonstrated on a 50 mm F/1 system and a 100 mm F/2 long-range system, claiming sharp results in RGB, NIR, and SWIR without refocusing.

Significance. If the central claims hold after quantitative validation, the work offers a practical route to achromatic wide-FOV imaging by combining the inherent achromaticity of mirrors with targeted computational correction of field curvature and off-axis aberrations. This could reduce reliance on complex multi-element refractive optics for VIS-SWIR sensors, lowering cost and complexity for applications in remote sensing, material inspection, and surveillance. The physical demonstrations on two distinct mirror systems provide a concrete existence proof for the minimal-stack strategy.

major comments (2)
  1. [SeidelConv model description] The SeidelConv model is described as a third-order Seidel approximation for the spatially-varying PSF of concave mirrors; however, for the 50 mm F/1 system, higher-order terms (oblique spherical aberration, higher-order coma) grow rapidly off-axis, and the manuscript provides no measured-vs-predicted PSF residual quantification or validation that the model residuals permit artifact-free inversion from only 2-4 images.
  2. [Experimental results / demonstrations] The experimental section reports successful demonstrations on two physical systems but supplies no quantitative metrics (PSNR, MTF, edge sharpness, or spectral-band comparisons), no error analysis accounting for sensor noise or model mismatch, and no baseline comparisons (e.g., single-image deconvolution or traditional focal-stack fusion). This absence prevents verification that the 2-4-image recovery meets the claimed high-resolution performance across the full FOV and spectrum.
minor comments (1)
  1. [Abstract] The abstract states that the method 'reveals material details invisible within individual spectral bands' without specifying the materials, wavelengths, or quantitative contrast improvement; a brief example or figure reference would strengthen the claim.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review of our manuscript on Computational Mirrors. The comments highlight important areas for strengthening the presentation of the SeidelConv model and the experimental validation. We address each major comment below and will incorporate the suggested improvements in the revised manuscript.

read point-by-point responses

  1. Referee: [SeidelConv model description] The SeidelConv model is described as a third-order Seidel approximation for the spatially-varying PSF of concave mirrors; however, for the 50 mm F/1 system, higher-order terms (oblique spherical aberration, higher-order coma) grow rapidly off-axis, and the manuscript provides no measured-vs-predicted PSF residual quantification or validation that the model residuals permit artifact-free inversion from only 2-4 images.

    Authors: We appreciate the referee's emphasis on rigorous model validation. The SeidelConv formulation deliberately employs the third-order Seidel aberrations to provide a computationally tractable, physics-based spatially-varying PSF that matches the minimal focal-stack acquisition strategy. Although higher-order aberrations are indeed more pronounced in the fast 50 mm F/1 system, our physical experiments indicate that the approximation remains sufficiently accurate for stable inversion. To directly address the concern, the revised manuscript will include measured PSF data acquired at multiple field angles, side-by-side comparisons with SeidelConv predictions, residual error maps, and an analysis demonstrating that the observed residuals support artifact-free recovery from 2-4 images. This addition will quantify the model's practical limits. revision: yes

  2. Referee: [Experimental results / demonstrations] The experimental section reports successful demonstrations on two physical systems but supplies no quantitative metrics (PSNR, MTF, edge sharpness, or spectral-band comparisons), no error analysis accounting for sensor noise or model mismatch, and no baseline comparisons (e.g., single-image deconvolution or traditional focal-stack fusion). This absence prevents verification that the 2-4-image recovery meets the claimed high-resolution performance across the full FOV and spectrum.

    Authors: We agree that quantitative metrics and baseline comparisons are necessary to substantiate the performance claims. The original manuscript relied primarily on visual results, which we recognize is insufficient for objective evaluation. In the revision we will add PSNR, SSIM, and MTF measurements across the field of view and spectral bands (RGB, NIR, SWIR), edge-sharpness quantification, and an error analysis that incorporates sensor noise and model mismatch effects. We will also include direct comparisons against single-image deconvolution and conventional focal-stack fusion methods for both the 50 mm F/1 and 100 mm F/2 systems. These quantitative results will be presented to verify the high-resolution, broadband imaging performance. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on external Seidel theory and new modeling

full rationale

The paper's core chain—capturing a 2-4 image focal stack from a concave mirror, modeling off-axis aberrations via the new SeidelConv PSF (derived from classical third-order Seidel terms), and recovering an all-in-focus image—does not reduce any output to a fitted parameter or self-defined quantity by construction. No equations equate the final sharp image to inputs via tautology, no uniqueness theorem is imported from self-citations, and no ansatz is smuggled through prior author work. The approach is self-contained against external physical benchmarks (Seidel aberrations) and experimental validation on 50mm F/1 and 100mm F/2 systems.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the accuracy of the SeidelConv model for real mirror aberrations and on the assumption that focus variation across the field is the dominant and correctable effect; no free parameters or new physical entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Distinct regions of the field of view reach focus at varying distances from a concave mirror
    Stated directly in the abstract as the foundation of the focal-stack approach.
invented entities (1)
  • SeidelConv no independent evidence
    purpose: Spatially-varying PSF model to characterize and correct off-axis aberrations in concave mirrors
    New model introduced in the paper; no independent evidence provided in abstract.

pith-pipeline@v0.9.0 · 5581 in / 1434 out tokens · 20587 ms · 2026-05-09T22:29:06.369996+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

  1. [1]

    Hyperspectral remote sensing data analysis and future challenges.IEEE Geoscience and Remote Sensing Magazine, 1(2):6–36, 2013

    José M Bioucas-Dias, Antonio Plaza, Gustavo Camps-Valls, Paul Scheunders, Nasser Nasrabadi, and Jocelyn Chanussot. Hyperspectral remote sensing data analysis and future challenges.IEEE Geoscience and Remote Sensing Magazine, 1(2):6–36, 2013. 1

    work page 2013

  2. [2]

    Deep learning classifiers for hyperspectral imaging: A review.J

    Mercedes Eugenia Paoletti, Juan Mario Haut, Javier Plaza, and Antonio Plaza. Deep learning classifiers for hyperspectral imaging: A review.J. Photogrammetry and Remote Sensing, 158:279–317, 2019. 1

    work page 2019

  3. [3]

    Hyperspectral imaging and its applications: A review.Heliyon, 10(12), 2024

    Anuja Bhargava, Ashish Sachdeva, Kulbhushan Sharma, Mohammed H Alsharif, Peerapong Uthansakul, and Monthippa Uthansakul. Hyperspectral imaging and its applications: A review.Heliyon, 10(12), 2024. 1 computational mirrors 25

    work page 2024

  4. [4]

    Study on shortwave infrared long-distance imag- ing performance based on multiband imaging experiments.Optical Engineering, 52(4):045008–045008, 2013

    Lang Junwei, Wang Yueming, Xiao Xizhong, Zhuang Xiaoqiong, Wang Sheng- wei, Liu Jun, and Wang Jianyu. Study on shortwave infrared long-distance imag- ing performance based on multiband imaging experiments.Optical Engineering, 52(4):045008–045008, 2013. 1

    work page 2013

  5. [5]

    Comparative long-time visible and shortwave infrared night illumination measurements.Appl

    Jürgen Krieg and Uwe Adomeit. Comparative long-time visible and shortwave infrared night illumination measurements.Appl. optics, 58(36):9876–9882, 2019. 1

    work page 2019

  6. [6]

    Academic Press, 2012

    Dietrich Korsch.Reflective optics. Academic Press, 2012. 2

    work page 2012

  7. [7]

    Reflecting microscopes with spherical mirrors.J

    KP Norris, WE Seeds, and MHF Wilkins. Reflecting microscopes with spherical mirrors.J. Optical Society of America, 41(2):111–119, 1951. 2, 5

    work page 1951

  8. [8]

    The james webb space telescope mission.Astronomical Society of the Pacific, 135(1048):068001, 2023

    Jonathan P Gardner, John C Mather, Randy Abbott, James S Abell, Mark Aber- nathy, Faith E Abney, John G Abraham, Roberto Abraham, Yasin M Abul-Huda, Scott Acton, et al. The james webb space telescope mission.Astronomical Society of the Pacific, 135(1048):068001, 2023. 2

    work page 2023

  9. [9]

    academic press,

    Rudolf Kingslake and R Barry Johnson.Lens design fundamentals. academic press,

    work page

  10. [10]

    Spatial transformer networks.Adv

    Max Jaderberg, Karen Simonyan, Andrew Zisserman, et al. Spatial transformer networks.Adv. Neural Info. Processing Systems, 28, 2015. 3

    work page 2015

  11. [11]

    Learning lens blur fields.IEEE Trans

    Esther YH Lin, Zhecheng Wang, Rebecca Lin, Daniel Miau, Florian Kainz, Jiawen Chen, Xuaner Zhang, David B Lindell, and Kiriakos N Kutulakos. Learning lens blur fields.IEEE Trans. Pattern Analysis and Machine Intelligence, 47(9):7383– 7395, 2025. 3, 9

    work page 2025

  12. [12]

    Plug-and-playimagerestorationwithdeepdenoiserprior.IEEE Trans

    Kai Zhang, Yawei Li, Wangmeng Zuo, Lei Zhang, Luc Van Gool, and Radu Timo- fte. Plug-and-playimagerestorationwithdeepdenoiserprior.IEEE Trans. Pattern Analysis and Machine Intelligence, 44(10):6360–6376, 2021. 3, 8, 10, 21, 22

    work page 2021

  13. [13]

    Cambridge University Press, 7th edition, 1999

    Max Born and Emil Wolf.Principles of Optics: Electromagnetic Theory of Prop- agation, Interference and Diffraction of Light. Cambridge University Press, 7th edition, 1999. 4, 7

    work page 1999

  14. [14]

    Anastigmatic three-mirror telescope.Appl

    Dietrich Korsch. Anastigmatic three-mirror telescope.Appl. optics, 16(8):2074– 2077, 1977. 4

    work page 2074

  15. [15]

    All-spherical catadioptric telescope design for wide-field imaging.Appl

    Mehdi Bahrami and Alexander V Goncharov. All-spherical catadioptric telescope design for wide-field imaging.Appl. optics, 49(30):5705–5712, 2010. 5

    work page 2010

  16. [16]

    Adaptation of the schupmann medial telescope to a large scale astronomical optical system.Appl

    John J Villa. Adaptation of the schupmann medial telescope to a large scale astronomical optical system.Appl. Optics, 11(8):1814–1821, 1972. 5

    work page 1972

  17. [17]

    Extended depth of field catadioptric imag- ing using focal sweep

    Ryunosuke Yokoya and Shree K Nayar. Extended depth of field catadioptric imag- ing using focal sweep. InIEEE Comp. Vision and Pattern Recognition (CVPR),

    work page

  18. [18]

    Field curvature correction using focal sweep.IEEE Trans

    Shigehiko Matsunaga and Shree K Nayar. Field curvature correction using focal sweep.IEEE Trans. Computational Imaging, 1(4):259–269, 2015. 5, 6, 11, 15

    work page 2015

  19. [19]

    Fast multicolor 3d imaging using aberration-corrected multifocus microscopy.Nature methods, 10(1):60–63, 2013

    Sara Abrahamsson, Jiji Chen, Bassam Hajj, Sjoerd Stallinga, Alexander Y Katsov, Jan Wisniewski, Gaku Mizuguchi, Pierre Soule, Florian Mueller, Claire Dugast Darzacq, et al. Fast multicolor 3d imaging using aberration-corrected multifocus microscopy.Nature methods, 10(1):60–63, 2013. 5

    work page 2013

  20. [20]

    Modeling and removing spatially-varying optical blur

    Eric Kee, Sylvain Paris, Simon Chen, and Jue Wang. Modeling and removing spatially-varying optical blur. InIEEE Intl. Conf. Computational Photography (ICCP). IEEE, 2011. 7

    work page 2011

  21. [21]

    EigenCWD: A spatially varying deconvolution algorithm for single metalens imag- ing.Optics Express, 33(13):28481–28492, 2025

    Joel Yeo, N Duane Loh, Ramon Paniagua-Dominguez, and Arseniy I Kuznetsov. EigenCWD: A spatially varying deconvolution algorithm for single metalens imag- ing.Optics Express, 33(13):28481–28492, 2025. 7, 8

    work page 2025

  22. [22]

    Experimental test of an analytical model of aberration in an oil-immersion objective lens used in three-dimensional light microscopy.J

    Sarah Frisken Gibson and Frederick Lanni. Experimental test of an analytical model of aberration in an oil-immersion objective lens used in three-dimensional light microscopy.J. Optical Society of America A, 8(10):1601–1613, 1991. 7 26 V. Saragadam et al

    work page 1991

  23. [23]

    Simulating anisoplanatic turbulence by sampling intermodal and spatially correlated zernike coefficients.Optical Engi- neering, 59(8):083101–083101, 2020

    Nicholas Chimitt and Stanley H Chan. Simulating anisoplanatic turbulence by sampling intermodal and spatially correlated zernike coefficients.Optical Engi- neering, 59(8):083101–083101, 2020. 7

    work page 2020

  24. [24]

    CoordGate: Efficiently comput- ing spatially-varying convolutions in convolutional neural networks.arXiv preprint arXiv:2401.04680, 2024

    Sunny Howard, Peter Norreys, and Andreas Döpp. CoordGate: Efficiently comput- ing spatially-varying convolutions in convolutional neural networks.arXiv preprint arXiv:2401.04680, 2024. 8, 11

    work page arXiv 2024

  25. [25]

    Hullin, Bjorn Labitzke, Wolfgang Hei- drich, and Andreas Kolb

    Felix Heide, Mushfiqur Rouf, Matthias B. Hullin, Bjorn Labitzke, Wolfgang Hei- drich, and Andreas Kolb. High-quality computational imaging through simple lenses.ACM Trans. Graphics, 32(5), 2013. 11

    work page 2013

  26. [26]

    Accessed: 2022-11-09

    Kodak lossless true color image suite.http://r0k.us/graphics/kodak/, 1999. Accessed: 2022-11-09. 11, 16

    work page 1999

  27. [27]

    Ring deconvolution microscopy: exploiting symmetry for efficient spatially varying aberration correction.Nature methods, 22(6):1311–1320, 2025

    Amit Kohli, Anastasios N Angelopoulos, David McAllister, Esther Whang, Sixian You, Kyrollos Yanny, Federico M Gasparoli, Bo-Jui Chang, Reto Fiolka, and Laura Waller. Ring deconvolution microscopy: exploiting symmetry for efficient spatially varying aberration correction.Nature methods, 22(6):1311–1320, 2025. 16

    work page 2025

  28. [28]

    Ntire 2017 challenge on single image super- resolution: Dataset and study

    Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super- resolution: Dataset and study. InConf. Computer Vision and Pattern Recognition (CVPR) Workshops, 2017. 19

    work page 2017

  29. [29]

    Ntire 2017 challenge on single image super-resolution: Methods and results

    Radu Timofte, Eirikur Agustsson, Luc Van Gool, Ming-Hsuan Yang, Lei Zhang, Bee Lim, et al. Ntire 2017 challenge on single image super-resolution: Methods and results. InConf. Computer Vision and Pattern Recognition (CVPR) Workshops,

    work page 2017

  30. [30]

    Deep image prior

    Dmitry Ulyanov, Andrea Vedaldi, and Victor Lempitsky. Deep image prior. In IEEE Comp. Vision and Pattern Recognition (CVPR), 2018. 21, 22

    work page 2018

  31. [31]

    Wire: Wavelet implicit neural repre- sentations

    Vishwanath Saragadam, Daniel LeJeune, Jasper Tan, Guha Balakrishnan, Ashok Veeraraghavan, and Richard G Baraniuk. Wire: Wavelet implicit neural repre- sentations. InIEEE Comp. Vision and Pattern Recognition (CVPR), 2023. 21, 22

    work page 2023