Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Transforming Single Photon Camera Images to Color High Dynamic Range Images

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper shows that a two-stage pipeline of Pix2PixHD colorization followed by DeepHDR HDR reconstruction produces the best color HDR images from monochrome single-photon camera captures, with a single-stage variant nearly matching it.

desk verdict Useful synthetic benchmark, but oracle exposure selection and 8-bit tone-mapping make the claimed ranking unlikely to transfer to real SPC sensors. read the letter →

arxiv 2412.12942 v1 pith:4KXVSMQI submitted 2024-12-17 eess.IV

Sumit Sharma , Girish Rongali , Kaushik Mitra This is my paper · ORCID
classification eess.IV
keywords SinglePhotonAvalancheDiodeHighDynamicRangeImagingGenerativeAdversarialNetworksimage-to-imagetranslationimageHDRreconstructionSPADsimulatorHDR-VDP3colorization
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 asks whether the noisy, monochrome images produced by single-photon cameras—sensors that detect individual photons and therefore have an extremely wide dynamic range—can be turned into color high-dynamic-range (HDR) images using existing deep-learning tools. It answers with a two-stage recipe: first translate the monochrome single-photon-camera image into a color low-dynamic-range image, then reconstruct a color HDR image from that. Across twelve combinations of colorization and HDR networks, the paper reports that Pix2PixHD followed by DeepHDR gives the highest HDR-VDP3 quality scores, and that a one-stage Pix2PixHD variant trained to go directly from monochrome single-photon camera images to color HDR performs comparably. If correct, this provides a computational route to usable color HDR output from single-photon sensors before color-capable SPAD hardware is widely available.

What carries the argument

The load-bearing object is the passive-mode single-photon camera (SPC) model and the two-stage network stack built around it. In the simulator, photon arrivals per pixel follow Poisson statistics, and each detected photon imposes a dead time $\tau_d = 150\,\mathrm{ns}$, producing a soft-saturating response with mean $\mathbb{E}[N^{SPC}_T] = q_{SPAD}\phi T/(1+q_{SPAD}\phi\tau_d)$ and variance $q_{SPAD}\phi T/(1+q_{SPAD}\phi\tau_d)^3$; the exposure time is chosen from the ground-truth luminance, and the resulting monochrome frames are saved as 8-bit images. The pipeline then applies a GAN-based image-to-image translator (Pix2PixHD, Pix2Pix, or CycleGAN) for colorization and a single-image HDR network (DeepHDR, SIHDR, ExpandNet, or LHDR) for the HDR stage, with HDR-VDP3 as the evaluation metric.

What would settle it

Capture the same scenes with a real single-photon camera and a conventional color HDR camera, feed the real monochrome frames through the trained Pix2PixHD + DeepHDR pipeline, and compare the perceptual HDR quality scores against the simulated results; a large drop in score or a change in the method ranking would show that the simulator's representation is not faithful.

Watch

Extended reading notes

Core claim

Monochromatic single-photon camera images can be converted into color HDR images by a two-stage approach: an image-to-image translation network first colorizes the SPC image, and a single-image HDR network then expands it to HDR. On the indoor HDR dataset used, with SPC images generated by the paper's Poisson/dead-time simulator, the combination of Pix2PixHD for colorization and DeepHDR for HDR reconstruction achieves the highest HDR-VDP3 scores of all twelve two-stage combinations, at both 1024x512 and 2048x1024 resolutions, and averaging four SPC frames improves the results. A single-stage variant—Pix2PixHD modified to translate directly from monochrome SPC images to 32-bit-per-channel HDR images—produces comparable scores, although the paper reports that its colors deviate slightly from the ground truth.

Load-bearing premise

The entire ranking rests on the assumption that the simulated single-photon camera images—generated from ground-truth HDR photos with exposure time chosen from the known scene luminance and then stored as 8-bit monochrome—faithfully represent what a real single-photon camera would capture.

Editorial extensions

If this is right

  • A two-stage pipeline of Pix2PixHD plus DeepHDR gives a working computational path from monochrome single-photon camera frames to color HDR images without color-capable SPAD hardware.
  • Averaging four SPC frames instead of one raises both the colorization metrics (LPIPS, PSNR, SSIM) and the final HDR-VDP3 score across all tested method combinations.
  • Higher input resolution (2048x1024) improves HDR reconstruction quality for every combination tested, so resolution is a direct lever on output quality within this pipeline.
  • A single-stage Pix2PixHD variant trained directly from monochrome SPC to color HDR reaches scores close to the two-stage pipeline, making the conversion possible in one network when slight color deviations are acceptable.

Reading between the lines

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

  • A natural extension the paper leaves untested is applying the same two-stage recipe to real SPAD sensor data; the reported ranking may change because the simulator selects exposure time from the known ground-truth luminance, which a real camera cannot do.
  • The near-parity of the single-stage network suggests the colorization and HDR-expansion tasks share enough structure to be learned jointly; a lighter joint model with a color-aware loss might close the remaining gap.
  • Because the pipeline's input is tone-mapped to 8-bit monochrome, the paper effectively measures what is achievable from compressed SPC output; feeding higher-bit-depth representations to the first stage could recover more of the camera's native dynamic range.
  • The final HDR stage is judged only by HDR-VDP3; perceptual color-fidelity metrics or human ratings could rank the method combinations differently.
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

4 major / 6 minor

Summary. The manuscript evaluates computational pipelines for converting simulated monochromatic Single Photon Camera (SPC) images into color high dynamic range (HDR) images. It proposes a two-stage design that first colorizes the SPC image with an image-to-image translation network (Pix2PixHD, Pix2Pix, or CycleGAN) and then reconstructs HDR with a single-image HDR method (DeepHDR, SIHDR, ExpandNet, or LHDR), plus a single-stage variant based on a modified Pix2PixHD. Experiments on the Laval HDR dataset report that Pix2PixHD followed by DeepHDR yields the highest HDR-VDP3 scores, that frame averaging improves results, and that the single-stage method is comparable to the two-stage approach.

Significance. The paper is a systematic comparison of established components for a plausible future SPC processing pipeline. Its strengths include testing many combinations, two resolutions, single-frame versus averaged inputs, and both two-stage and single-stage routes, with HDR-VDP3 as an appropriate metric for HDR reconstruction quality. The authors also include a candid caveat in Section 3.3 that current SPAD prototypes are not yet competitive with CMOS sensors. However, the external validity of the central ranking rests on an idealized simulation protocol: per-image oracle exposure selection and 8-bit tone-mapping of SPC outputs. Because these choices alter the task being evaluated, the paper's claim to validate an SPC-to-color-HDR pipeline is stronger than the evidence supports. With a corrected evaluation protocol or substantially narrowed claims, the comparison could be a useful reference for the community.

major comments (4)
  1. [Section 3.2] The simulator selects the exposure time T based on the ground-truth luminance: the paper states, 'we select the exposure time based on the luminance of the ground truth HDR images.' This is oracle access to the test target. A real SPC camera uses a fixed or user-chosen exposure, not one hand-picked from the scene radiance to maximize visibility. This choice removes exactly the saturation and exposure-robustness problem that motivates SPC HDR imaging, so the reported ranking may reflect how well the methods exploit oracle-exposed inputs rather than how they would perform on genuine SPC captures. Please evaluate with held-out exposure settings, such as one fixed exposure per resolution or a sweep over exposures, and report how the ranking changes.
  2. [Sections 3.2 and 4] The SPC images are saved and fed to the pipeline as 8-bit tone-mapped monochromatic images. This discards the high-bit-depth photon-count information that constitutes the SPC's extended dynamic range, effectively turning the problem into monochrome-LDR-to-color-HDR rather than SPC-to-HDR. If the chosen architectures require 8-bit input, the claims should be narrowed accordingly, or the paper should include an experiment that preserves higher bit depth and, ideally, validate on real SPC sensor data.
  3. [Sections 4 and 5, Tables 1-3] The paper does not report the train/test split used for the Laval dataset, although the same dataset provides both the simulated SPC inputs and the ground-truth targets. Table 2 reports scores 'averaged over 262 images' but never states that those images were excluded from training. Without this information, the quantitative comparisons are potentially circular. Please specify the exact split, the number of training and test images, and ideally release code and the simulator settings so that the results can be reproduced.
  4. [Tables 2 and 3] The central ranking claim that Pix2PixHD + DeepHDR is 'the best possible HDR reconstruction' is based on mean HDR-VDP3 scores without any measure of variability or statistical significance. Since the differences between some competing combinations are relatively small (for example, 3.6388 versus 3.4545 for Pix2PixHD + DeepHDR versus Pix2Pix + DeepHDR at 1024x512), please report standard deviations, confidence intervals, or paired significance tests over the 262 test images to support the claim.
minor comments (6)
  1. [Section 3.2] There is a notation inconsistency: the text says 'the appropriate exposure time, denoted τd' but the equations use T for exposure time and τd for dead time.
  2. [Abstract and Section 1] The statement that SPAD sensors 'have surpassed' CMOS sensors is too strong given the authors' own caveat in Section 3.3 that current SPAD prototypes cannot compete directly with CMOS image sensors; please soften the wording.
  3. [Figure 3 caption] The caption says 'moving from Monochromatic HDR to Color HDR,' but the intended task is from monochromatic SPC to color HDR; please correct the wording.
  4. [Table 2] The row labels 'Low+High' are unclear; please replace them with explicit resolution names or add a note explaining the notation.
  5. [Section 5] The metric name is spelled inconsistently as 'HDR-VDP3' and 'HDR-VDP-3'; please use one form throughout.
  6. [Section 5 and conclusions] The single-stage results in Table 3 are described as 'comparable' to the two-stage results, but the Q-scores are consistently lower at 1024x512 and only close at 2048x1024; please qualify this statement more precisely.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the central claims are empirical benchmark results whose inputs are not equivalent to their outputs by construction.

full rationale

This is a benchmarking and evaluation paper, not a derivation of a physical model, so there is no chain of equations in which a predicted quantity reduces to an input by construction. The SPC simulator uses equations (1)-(2) taken from prior independent work (Liu et al. [9] and Ingle et al. [10]), and the paper does not claim to derive those equations. No self-citation chain is load-bearing: none of the cited prior results are authored by the present authors, and no uniqueness theorem is invoked to force the method choice. The main concern raised by the reader is the simulator's exposure-time selection based on ground-truth HDR luminance, and the 8-bit tone-mapped representation of SPC inputs. These choices may affect external validity and make the benchmark idealized, but they are not circular: the colorization and HDR-reconstruction networks never see the ground-truth HDR during inference, and the reported HDR-VDP2 scores are not statistically forced by the exposure-time choice. The method still has to learn color and HDR reconstruction from the monochromatic inputs. Similarly, saving SPC frames as 8-bit monochrome is a preprocessing decision that changes the task difficulty but does not make the output equivalent to the input. Absence of code and train/test splits is a reproducibility concern, not a circularity concern. Overall, the claims are self-contained as an empirical comparison on a synthetic benchmark, so the circularity score is 0.

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

The central claim rests on a simulator whose key inputs (per-image exposure time, pixel sensitivity, dead time) are either tuned to the ground-truth images or not fully specified, and on the assumption that 8-bit tone-mapped simulated SPC images plus pre-trained SIHDR models form a valid testbed. None of these are validated against real SPC sensor data.

free parameters (4)
  • per-image exposure time = not reported
    Chosen based on ground-truth luminance of each HDR image in Section 3.2, leaking target information into the simulation and making the reconstruction task easier.
  • SPC pixel sensitivity (q_SPAD) = not reported
    Adjusted during simulation to ensure all regions are visible (Section 3.2), with no fixed value given.
  • dead time = 150 ns
    The simulator uses a fixed dead time; the value only appears in Section 6 while Section 3.2 gives no number.
  • frame averaging count = 1 or 4
    Averaging factor chosen by hand in Section 3.3; results are reported for both values.
assumptions (4)
  • domain assumption SPAD detection follows a renewal process with Poisson arrivals and a dead-time period, giving mean and variance in Eq. (1)-(2)
    Equations are taken from cited SPC literature [9],[10]; the paper does not validate the model against real SPC hardware.
  • ad hoc to paper Simulated SPC images derived from Laval HDR images represent real SPC camera behavior closely enough to rank colorization and HDR reconstruction methods
    All evaluation uses synthetic inputs; no real SPAD data is included, so the transferability of the ranking is assumed.
  • ad hoc to paper Tone-mapping the simulated SPC image to 8-bit monochrome preserves enough information for color HDR reconstruction
    Section 3.2 saves all SPC images as 8-bit, discarding the photon-count range that gives SPC its HDR advantage; the input is effectively LDR.
  • ad hoc to paper Pre-trained single-image HDR models transfer from natural LDR images to colorized SPC-like images
    Stage-2 networks are used as pre-trained without fine-tuning on the colorized SPC domain (Section 3.3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transforming Single Photon Camera Images to Color High Dynamic Range Images." pith.science (2026). https://pith.science/paper/4KXVSMQI

@misc{pith2026241212942,
  author       = {Pith},
  title        = {Pith review of: Transforming Single Photon Camera Images to Color High Dynamic Range Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KXVSMQI}},
  note         = {Machine review of arXiv:2412.12942}
}
read the original abstract

Traditional CMOS sensors suffer from restricted dynamic range and sub optimal performance under extreme lighting conditions. They are affected by electronic noise in low light conditions and pixel saturation while capturing high illumination. Recent High Dynamic Range (HDR) Imaging methods, often designed for CMOS Sensors, attempt to address these issues by fusing multiple exposures. However, they frequently introduce artifacts like ghosting and light flickering in dynamic scenarios and non-uniform signal-to-noise ratio in extreme dynamic range conditions. Recently, Single Photon Avalanche Diodes (SPADs), also known as Single Photon Camera (SPC) sensors, have surpassed CMOS sensors due to their capability to capture individual photons with high timing precision. Unlike traditional digital cameras that first convert light energy into analog electrical currents and then digitize them, SPAD sensors perform direct photon detection, making them less susceptible to extreme illumination conditions. Their distinctive non-linear response curve aids in capturing photons across both low-light and high-illumination environments, making them particularly effective for High Dynamic Range Imaging. Despite their advantages, images from SPAD Sensors are often noisy and visually unappealing. To address these challenges, we evaluate state-of-the-art architectures for converting monochromatic SPAD images into Color HDR images at various resolutions. Our evaluation involves both qualitative and quantitative assessments of these architectures, focusing on their effectiveness in each stage of the conversion process.

Figures

Figures reproduced from arXiv: 2412.12942 by the authors.

Figure 1
Figure 1. SPC simulation pipeline: Different averaged SPC frames at different res [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. This study examines various methods for converting a monochromatic [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Best possible combination among the different state-of-the-art architec [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: This figure presents a qualitative analysis of 1-frame SPC to color trans [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: This figure presents qualitative analysis of four-frame averaged SPC for [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Qualitative analysis of 1-frame SPC for color HDR translation at a res [PITH_FULL_IMAGE:figures/full_fig_p012_6.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

  1. SPC to 3D: Novel View Synthesis from Binary SPC via I2I translation

    eess.IV 2025-06 conditional novelty 3.0 of 10

    A two-stage pipeline of Pix2PixHD image translation followed by NeRF or 3DGS produces color novel views from simulated binary single-photon camera images.

Reference graph

Works this paper leans on

47 extracted references · 43 canonical work pages · cited by 1 Pith paper

  1. [1]

    High dynamic range imaging

    Reinhard, Erik. "High dynamic range imaging." In Computer Vision: A Reference Guide, pp. 1-6. Cham: Springer International Publishing, 2020

    work page 2020

  2. [2]

    Recovering high dynamic range radiance maps from photographs

    Debevec, Paul E., and Jitendra Malik. "Recovering high dynamic range radiance maps from photographs." In Seminal Graphics Papers: Pushing the Boundaries, Volume 2, pp. 643-652. 2023

    work page 2023

  3. [3]

    Beingundigital’with digital cameras

    Mann, S., and R. Picard. "Beingundigital’with digital cameras." MIT Media Lab Perceptual 1, no. 2 (1994): 3

    work page 1994

  4. [4]

    The stateoftheartinHDRdeghosting:Asurveyandevaluation

    Tursun, Okan Tarhan, Ahmet Oğuz Akyüz, Aykut Erdem, and Erkut Erdem. "The stateoftheartinHDRdeghosting:Asurveyandevaluation."InComputerGraphics Forum, vol. 34, no. 2, pp. 683-707. 2015

    work page 2015

  5. [5]

    LED flicker: Root cause, impact and measurement for automotive imaging applications

    Deegan, Brian. "LED flicker: Root cause, impact and measurement for automotive imaging applications." Electronic Imaging 30 (2018): 1-6

    work page 2018

  6. [6]

    A 0.68 e-rms random-noise 121dB dynamic-range sub-pixel architecture CMOS image sensor with LED flicker mitigation

    Iida, S., Y. Sakano, T. Asatsuma, M. Takami, I. Yoshiba, N. Ohba, H. Mizuno et al. "A 0.68 e-rms random-noise 121dB dynamic-range sub-pixel architecture CMOS image sensor with LED flicker mitigation." In 2018 IEEE International Electron Devices Meeting (IEDM), pp. 10-2. IEEE, 2018

    work page 2018

  7. [7]

    140 db dynamic range sub-electron noise floor image sen- sor

    Velichko, Sergey, Scott Johnson, Dan Pates, Chris Silsby, Cornelis Hoekstra, Ray Mentzer, and Jeff Beck. "140 db dynamic range sub-electron noise floor image sen- sor." Proceedings of the IISW 1 (2017)

    work page 2017

  8. [8]

    Passive inter-photon imaging

    Ingle, Atul, Trevor Seets, Mauro Buttafava, Shantanu Gupta, Alberto Tosi, Mohit Gupta, and Andreas Velten. "Passive inter-photon imaging." In Proceedings of the TSPCHDR 13 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8585-

    work page

Show all 47 references
  1. [9]

    Single-photon camera guided extreme dynamic range imaging

    Liu, Yuhao, Felipe Gutierrez-Barragan, Atul Ingle, Mohit Gupta, and Andreas Vel- ten. "Single-photon camera guided extreme dynamic range imaging." In Proceed- ings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 1575-1585. 2022

    2022

  2. [10]

    High flux passive imaging with single-photon sensors

    Ingle, Atul, Andreas Velten, and Mohit Gupta. "High flux passive imaging with single-photon sensors." In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 6760-6769. 2019

    2019

  3. [11]

    Gradient domain high dynamic range compression

    Fattal, Raanan, Dani Lischinski, and Michael Werman. "Gradient domain high dynamic range compression." In Seminal Graphics Papers: Pushing the Boundaries, Volume 2, pp. 671-678. 2023

    2023

  4. [12]

    Sub-pixel architecture of cmos image sensor achieving over 120 db dynamic range with less motion artifact characteristics

    Asatsuma, T., Y. Sakano, S. Iida, M. Takami, I. Yoshiba, N. Ohba, H. Mizuno et al. "Sub-pixel architecture of cmos image sensor achieving over 120 db dynamic range with less motion artifact characteristics." In Proceedings of the 2019 International Image Sensor Workshop, vol. 1. 2019

    2019

  5. [13]

    A wide dynamic range CMOS image sensor with multiple exposure-time signal outputs and 12-bit column-parallel cyclic A/D converters

    Mase, Mitsuhito, Shoji Kawahito, Masaaki Sasaki, Yasuo Wakamori, and Masanori Furuta. "A wide dynamic range CMOS image sensor with multiple exposure-time signal outputs and 12-bit column-parallel cyclic A/D converters." IEEE Journal of Solid-State Circuits 40, no. 12 (2005): 2787-2795

    2005

  6. [14]

    Comparative analysis of SNR for im- age sensors with enhanced dynamic range

    Yang, David XD, and Abbas El Gamal. "Comparative analysis of SNR for im- age sensors with enhanced dynamic range." In Sensors, cameras, and systems for scientific/industrial applications, vol. 3649, pp. 197-211. SPIE, 1999

    1999

  7. [15]

    Generative adversarial nets

    Goodfellow, Ian, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. "Generative adversarial nets." Advances in neural information processing systems 27 (2014)

    2014

  8. [16]

    Image-to-image translation with conditional adversarial networks

    Isola, Phillip, Jun-Yan Zhu, Tinghui Zhou, and Alexei A. Efros. "Image-to-image translation with conditional adversarial networks." In Proceedings of the IEEE con- ference on computer vision and pattern recognition, pp. 1125-1134. 2017

    2017

  9. [17]

    Perceptual losses for real-time style transfer and super-resolution

    Johnson, Justin, Alexandre Alahi, and Li Fei-Fei. "Perceptual losses for real-time style transfer and super-resolution." In Computer Vision–ECCV 2016: 14th Euro- pean Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part II 14, pp. 694-711. Springer In...

    2016

  10. [18]

    Semantic image synthesis via adversarial learning

    Dong, Hao, Simiao Yu, Chao Wu, and Yike Guo. "Semantic image synthesis via adversarial learning." In Proceedings of the IEEE international conference on com- puter vision, pp. 5706-5714. 2017

    2017

  11. [19]

    Generative attribute controller with conditional filtered generative adversarial networks

    Kaneko, Takuhiro, Kaoru Hiramatsu, and Kunio Kashino. "Generative attribute controller with conditional filtered generative adversarial networks." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 6089-6098. 2017

    2017

  12. [20]

    Learning to generate images of outdoor scenes from attributes and semantic layouts

    Karacan, Levent, Zeynep Akata, Aykut Erdem, and Erkut Erdem. "Learning to generate images of outdoor scenes from attributes and semantic layouts." arXiv preprint arXiv:1612.00215 (2016)

    2016 arXiv

  13. [21]

    Unsupervised pixel-level domain adaptation with generative ad- versarial networks

    Bousmalis, Konstantinos, Nathan Silberman, David Dohan, Dumitru Erhan, and Dilip Krishnan. "Unsupervised pixel-level domain adaptation with generative ad- versarial networks." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3722-3731. 2017

    2017

  14. [22]

    Unsupervised image-to-image translation networks

    Liu, Ming-Yu, Thomas Breuel, and Jan Kautz. "Unsupervised image-to-image translation networks." Advances in neural information processing systems 30 (2017). 14 Sumit Sharma, Girish Rongali, and Kaushik Mitra

    2017

  15. [23]

    Learning from simulated and unsupervised images through adversarial training

    Shrivastava, Ashish, Tomas Pfister, Oncel Tuzel, Joshua Susskind, Wenda Wang, and Russell Webb. "Learning from simulated and unsupervised images through adversarial training." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2107-2116. 2017

    2017

  16. [24]

    Unsupervised cross-domain image generation

    Taigman, Yaniv, Adam Polyak, and Lior Wolf. "Unsupervised cross-domain image generation." arXiv preprint arXiv:1611.02200 (2016)

    2016 arXiv

  17. [25]

    Photographic image synthesis with cascaded refinement networks

    Chen, Qifeng, and Vladlen Koltun. "Photographic image synthesis with cascaded refinement networks." In Proceedings of the IEEE international conference on com- puter vision, pp. 1511-1520. 2017

    2017

  18. [26]

    Generating images with perceptual simi- larity metrics based on deep networks

    Dosovitskiy, Alexey, and Thomas Brox. "Generating images with perceptual simi- larity metrics based on deep networks." Advances in neural information processing systems 29 (2016)

    2016

  19. [27]

    Image style transfer using convolutional neural networks

    Gatys, Leon A., Alexander S. Ecker, and Matthias Bethge. "Image style transfer using convolutional neural networks." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2414-2423. 2016

    2016

  20. [28]

    High-resolution image synthesis and semantic manipulation with condi- tional gans

    Wang,Ting-Chun,Ming-YuLiu, Jun-YanZhu,AndrewTao, Jan Kautz,andBryan Catanzaro. "High-resolution image synthesis and semantic manipulation with condi- tional gans." In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 8798-8807. 2018

    2018

  21. [29]

    Exposure fusion: A simple and practical alternative to high dynamic range photography

    Mertens, Tom, Jan Kautz, and Frank Van Reeth. "Exposure fusion: A simple and practical alternative to high dynamic range photography." In Computer graphics forum, vol. 28, no. 1, pp. 161-171. Oxford, UK: Blackwell Publishing Ltd, 2009

    2009

  22. [30]

    HDR image reconstruction from a single exposure using deep CNNs

    Gabriel, Eilertsen. "HDR image reconstruction from a single exposure using deep CNNs." ACM Trans. Graph. 36 (2017): 178-1

    2017

  23. [31]

    Image correction via deep reciprocating HDR transformation

    Yang,Xin,KeXu,YibingSong,QiangZhang,XiaopengWei,andRynsonWHLau. "Image correction via deep reciprocating HDR transformation." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1798-1807. 2018

    2018

  24. [32]

    Hybrid loss for learning single-image-based HDR reconstruction

    Moriwaki, Kenta, Ryota Yoshihashi, Rei Kawakami, Shaodi You, and Takeshi Nae- mura. "Hybrid loss for learning single-image-based HDR reconstruction." arXiv preprint arXiv:1812.07134 (2018)

    2018 arXiv

  25. [33]

    Single-image HDR reconstruction by learn- ing to reverse the camera pipeline

    Liu, Yu-Lun, Wei-Sheng Lai, Yu-Sheng Chen, Yi-Lung Kao, Ming-Hsuan Yang, Yung-Yu Chuang, and Jia-Bin Huang. "Single-image HDR reconstruction by learn- ing to reverse the camera pipeline." In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp....

    2020

  26. [34]

    Single image HDR reconstruction using a CNN with masked features and perceptual loss

    Santos, Marcel Santana, Tsang Ing Ren, and Nima Khademi Kalantari. "Single image HDR reconstruction using a CNN with masked features and perceptual loss." arXiv preprint arXiv:2005.07335 (2020)

    2020 arXiv

  27. [35]

    Single-image hdr reconstruction by multi-exposure generation

    Le, Phuoc-Hieu, Quynh Le, Rang Nguyen, and Binh-Son Hua. "Single-image hdr reconstruction by multi-exposure generation." In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pp. 4063-4072. 2023

    2023

  28. [36]

    Photon-starved scene inference using single photon cameras

    Goyal, Bhavya, and Mohit Gupta. "Photon-starved scene inference using single photon cameras." In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 2512-2521. 2021

    2021

  29. [37]

    Beyond the pixel: a photometrically calibrated hdr dataset for luminance and color prediction

    Bolduc, Christophe, Justine Giroux, Marc Hébert, Claude Demers, and Jean- François Lalonde. "Beyond the pixel: a photometrically calibrated hdr dataset for luminance and color prediction." In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 8071-8081. 2023

    2023

  30. [38]

    Probability and random processes

    Grimmett, Geoffrey, and David Stirzaker. Probability and random processes. Ox- ford university press, 2020. TSPCHDR 15

    2020

  31. [39]

    Qs-attn: Query-selected attention for contrastive learning in i2i translation

    Hu, Xueqi, Xinyue Zhou, Qiusheng Huang, Zhengyi Shi, Li Sun, and Qingli Li. "Qs-attn: Query-selected attention for contrastive learning in i2i translation." In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recog- nition, pp. 18291-18300. 2022

    2022

  32. [40]

    Seeing Photons in Color

    Ma, Sizhuo, Varun Sundar, Paul Mos, Claudio Bruschini, Edoardo Charbon, and Mohit Gupta. "Seeing Photons in Color." ACM Transactions on Graphics (TOG) 42, no. 4 (2023): 1-16

    2023

  33. [41]

    Canon Has Successfully Developed an Advanced SPAD Sensor

    “Canon Has Successfully Developed an Advanced SPAD Sensor.” Canon Global. Last modified 2023. https://global.canon/en/technology/spad-sensor-2023

    2023

  34. [42]

    Lhdr: Hdr reconstruction for legacy content using a lightweight dnn

    Guo, Cheng, and Xiuhua Jiang. "Lhdr: Hdr reconstruction for legacy content using a lightweight dnn." In Proceedings of the Asian Conference on Computer Vision, pp. 3155-3171. 2022

    2022

  35. [43]

    HDR-VDP-3: A multi- metric for predicting image differences, quality and contrast distortions in high dynamic range and regular content

    Mantiuk, Rafal K., Dounia Hammou, and Param Hanji. "HDR-VDP-3: A multi- metric for predicting image differences, quality and contrast distortions in high dynamic range and regular content." arXiv preprint arXiv:2304.13625 (2023)

    2023 arXiv

  36. [44]

    Jonathan, Fatih Bakir, Göksu Güvendiren, and Dr

    Hatchett, Dr. Jonathan, Fatih Bakir, Göksu Güvendiren, and Dr. Josh McNamee. OpenHDR Viewer. Accessed July 7, 2024. https://viewer.openhdr.org/

    2024

  37. [45]

    Expandnet: A deep convolutional neural network for high dynamic range expansion from low dynamic range content

    Marnerides, Demetris, Thomas Bashford-Rogers, Jonathan Hatchett, and Kurt Debattista. "Expandnet: A deep convolutional neural network for high dynamic range expansion from low dynamic range content." In Computer Graphics Forum, vol. 37, no. 2, pp. 37-49. 2018

    2018

  38. [46]

    Unpaired image- to-image translation using cycle-consistent adversarial networks

    Zhu, Jun-Yan, Taesung Park, Phillip Isola, and Alexei A. Efros. "Unpaired image- to-image translation using cycle-consistent adversarial networks." In Proceedings of the IEEE international conference on computer vision, pp. 2223-2232. 2017

    2017

  39. [47]

    Learning to predict indoor illumination from a single image

    Gardner, Marc-André, Kalyan Sunkavalli, Ersin Yumer, Xiaohui Shen, Emiliano Gambaretto, Christian Gagné, and Jean-François Lalonde. "Learning to predict indoor illumination from a single image." arXiv preprint arXiv:1704.00090 (2017)

    2017 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.