Pith. sign in

REVIEW 2 major objections 5 minor 1 cited by

Is JPEG AI going to change image forensics?

T0 review · 2 major / 5 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 JPEG AI compression artifacts are mistaken for deepfake and splicing traces by state-of-the-art forensic detectors.

desk verdict Solid empirical vulnerability study of JPEG AI against forensic detectors; the main caveat is generalization from one reference implementation. read the letter →

arxiv 2412.03261 v2 pith:6W4H5YRJ submitted 2024-12-04 eess.IV cs.CV

Edoardo Daniele Cannas , Sara Mandelli , Nataša Popović , Ayman Alkhateeb , Alessandro Gnutti , Paolo Bestagini , Stefano Tubaro This is my paper · ORCID
classification eess.IVcs.CV
keywords JPEGAIneuralimagecompressiondeepfakedetectionsplicinglocalizationcounter-forensicsartifactsfalsepositivesforensics
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

JPEG AI is the first international image-compression standard built on neural networks, and its decoder leaves upsampling artifacts that resemble traces left by GANs and diffusion generators. The paper tries to show that these artifacts are strong enough to break two standard forensic tasks: telling real from deepfake images, and localizing spliced regions. Across a set of state-of-the-art detectors, pristine images compressed with JPEG AI were increasingly flagged as synthetic as compression tightened, and tampering masks became unreliable in both directions. The concern matters because JPEG AI is designed for broad adoption, so a widespread codec could turn ordinary compressed photos into false forensic alarms.

What carries the argument

The load-bearing object is the JPEG AI decoding pipeline, in particular the upsampling and synthesis transforms in its normative decoder that reconstruct the image from the compressed latent representation. The argument is that this transform imprints periodic spectral peaks and local pixel-interdependence patterns in the same family as the traces that deepfake detectors and splicing localizers are trained to recognize, so a detector reads compression artifacts as generation or manipulation artifacts. The Fourier-spectrum analysis of noise residuals is the diagnostic used to link JPEG AI outputs to synthetic-image artifacts.

What would settle it

Take the same pristine and spliced datasets, compress them with a second JPEG AI implementation produced by a different vendor, and rerun the same detectors at the strongest compression setting (0.12 bits per pixel). If the false-positive rate stays near the uncompressed baseline and the splicing localization area under the curve remains near its original value, the measured counter-forensic effect is specific to the reference implementation rather than the standard.

Watch

Extended reading notes

Core claim

The central claim is that JPEG AI is a counter-forensic operation: its output carries frequency-domain and interpolation artifacts that forensic detectors interpret as evidence of synthesis or tampering. At the strongest compression tested (0.12 bits per pixel), most deepfake detectors' false-positive rate jumped from near zero to 0.56–0.99 while balanced accuracy collapsed, and the effect softened but persisted at higher bit rates. For splicing localization, the pixel-level area under the curve dropped by roughly 0.15 and the F1 score fell by up to 0.39, with detectors both missing tampered pixels and flagging pristine ones. The paper also shows standard JPEG acts in the opposite direction, mostly hiding synthetic traces by raising false negatives rather than creating false positives, and that including JPEG AI images in training reduces but does not eliminate the effect.

Load-bearing premise

The experiments use a single official reference implementation of JPEG AI, and the paper assumes that its compression artifacts represent those of every compliant JPEG AI product.

Editorial extensions

If this is right

  • At the strongest setting tested, several deepfake detectors flag roughly 56–99% of JPEG AI-compressed pristine images as synthetic, meaning false alarms become the norm rather than the exception.
  • Splicing localization masks degrade in both directions at low bit rates: manipulated pixels are missed while pristine areas are flagged, so JPEG AI compression can invalidate tampering localization evidence.
  • The failure mode is opposite to standard JPEG: JPEG mostly moves deepfake scores toward the pristine side, raising false negatives, whereas JPEG AI moves pristine scores toward the deepfake side, raising false positives.
  • Retraining a detector with 20% JPEG AI-compressed samples improves balanced accuracy and false-positive rate at low bit rates, but performance still stays below 0.7 balanced accuracy, so full robustness would require a much larger augmentation campaign.
  • A strong second JPEG AI compression can erase traces left by a first compression, mirroring double-JPEG behavior and adding a new wrinkle for image-history forensic scenarios.

Reading between the lines

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

  • If JPEG AI adoption spreads, forensic pipelines may need a preprocessing or calibration step that distinguishes compression upsampling artifacts from generative upsampling artifacts, rather than treating every interpolation trace as manipulation evidence.
  • The measured failures suggest modern detectors rely heavily on generic spectral and upsampling cues, so a detector trained explicitly on paired JPEG AI-compressed and uncompressed versions of the same pristine images could provide a direct robustness benchmark.
  • The double-compression results hint that spatial inconsistencies in JPEG AI artifact strength might become a useful forensic trace for locating spliced regions, since a spliced area can carry compression history different from its surroundings.
  • Because the standard allows manufacturers to retrain parts of the encoder, the exact artifact spectrum may vary across compliant products; benchmarking the same detectors on images from several independent JPEG AI implementations would show how much of the effect is intrinsic to the standard.
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

2 major / 5 minor

Summary. The paper investigates whether JPEG AI compression causes counter-forensic effects on deepfake image detection and image splicing localization. The authors compress pristine and synthetic test images with the official JPEG AI Verification Model 7.0 at the High Operation Point, using six BPP values (0.12, 0.25, 0.5, 0.75, 1.0, 2.0), and evaluate nine deepfake detectors and three splicing localization detectors on the datasets used by the original authors. They compare the results with standard JPEG compression, report image quality metrics, retrain one detector after mixing JPEG AI samples into training, and present preliminary double-compression experiments. The headline findings are that, at low BPP, pristine JPEG AI images are increasingly classified as synthetic (for most detectors the FPR rises from 0.00–0.30 to 0.56–0.99 at BPP = 0.12), splicing localization degrades markedly (e.g., AUC for TruFor drops from about 0.91 to 0.75 and F1 from 0.66 to 0.37), and these effects become weaker but do not fully disappear at higher BPP values.

Significance. If the results hold, the paper provides a timely and falsifiable warning: the forensic community should include JPEG AI-compressed images in training and benchmarking, because the artifacts of this standard can be misinterpreted as deepfake or manipulation traces. The study has notable strengths: it covers nine public deepfake detectors and three splicing detectors across several standard datasets, it includes a standard-JPEG control condition that supports the claim of a distinct artifact signature, it reports perceptual quality metrics, and it ships code for replication. The tables are internally consistent (spot-checks confirm that BA, FPR, and FNR satisfy BA = (1 − FPR + 1 − FNR)/2), and the consistency of the direction across detectors makes the main effect credible. The central limitation is external validity: all experiments use a single JPEG AI reference implementation, so the quantitative generalization from this one codec, and its HOP configuration, to the JPEG AI standard as a whole is not yet established.

major comments (2)
  1. [Sec. 4.3 and Sec. 2]
  2. [Tables 1, 3, 4, 5 and Sec. 5.1–5.2] All quantitative comparisons are point estimates with no confidence intervals, significance tests, or information about image-level variability. The paper makes strong quantitative statements, such as “a drop of more than 0.15 and 0.13 points in AUC and BA” and “FPR risks to be high”, but the reader cannot assess whether the moderate changes at BPP = 1.0 (e.g., BA around 0.65–0.87 in Table 1) are stable across images or datasets. I recommend adding per-image score distributions or bootstrap confidence intervals for the headline metrics, or softening the quantitative language; the qualitative direction of the effect is nevertheless consistent and convincing.
minor comments (5)
  1. [Sec. 5.2] The sentence “we calculate these metrics for all the datasets considered and then average their values” is fine, but the reference to “Tab. 5 (second and third right column)” is confusing because the table has several columns separated by BPP/QF; please refer to the specific column blocks by the BPP value or by the metric name.
  2. [Abstract, Sec. 1, Fig. 1] The verb “prove” is too strong for an empirical measurement study; recommend “demonstrate” or “provide evidence” in the abstract and in the caption of Fig. 1.
  3. [Sec. 5.3] The double JPEG AI compression results are discussed only qualitatively in the main text and supported only by supplementary figures; adding one quantitative table or metric for the double-compression scenarios would make this preliminary analysis more useful.
  4. [Sec. 2, Fig. 2] The text refers to “blue modules” and “red modules” in Fig. 2, but in a black-and-white print or a non-color PDF these distinctions may be invisible; please add labels or textures.
  5. [References] Several references contain stray trailing numbers (e.g., [1] ends with “2, 4, 5, 3” and [32] with “1, 13”), likely artifacts of the page-layout process; these should be removed.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: this is an empirical detector-evaluation study anchored to external detectors, official codec software, and public datasets; the one self-citation is not load-bearing.

full rationale

This paper is a measurement paper, not a derivation. The central claim that JPEG AI-compressed pristine content is misclassified by forensic detectors is established by running nine independent deepfake detectors and three splicing-localization detectors on images produced by the official JPEG AI Reference Software (Verification Model 7.0, HOP) and comparing scores and localization masks before and after compression. No quantity is fitted to the reported FPR, AUC, or F1 outcomes and then relabeled as a prediction. The only appearance of the authors' own prior work is the retraining/fine-tuning experiment built on detector [63] in Section 5.1; that experiment is explicitly presented as a preliminary mitigation test, not as evidence for the counter-forensic effect, and the main Tables 1 and 5 stand without it. The claim in Section 2 that the normative decoder should produce similar artifacts across compliant JPEG AI implementations is a structural argument about the standard, not a circular reuse of the paper's own measured values. Generalizing from one Verification Model and one operation point is a real external-validity limitation, but it is a generalization-gap concern, not a circularity. Consequently, no circular step can be exhibited with the required quote-and-reduction standard. Score 1 reflects one minor self-citation that does not carry the argument.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claims rest on standard forensic measurement conventions rather than on fitted parameters. No new entities are introduced. The hand-chosen analysis choices, the AUC 0.75 exclusion cutoff and the 20% JPEG AI training mix, are disclosed design decisions that shape secondary numbers, not fitted values that manufacture the primary effect.

free parameters (2)
  • Baseline AUC exclusion cutoff = 0.75
    Datasets where a detector scores AUC below 0.75 are removed from the deepfake analysis before averaging (Sec. 5.1), so the aggregate numbers in Table 1 describe only the better-behaved subset. Disclosed, but hand-chosen and it changes the reported averages.
  • JPEG AI training augmentation fraction = 20%
    In the mitigation experiment (Sec. 5.1, 'Introducing JPEG AI into training'), 20% of the training and validation samples of detector [63] are compressed with JPEG AI. The authors themselves conclude this is insufficient ('a huge augmentation campaign should probably be considered'), so the choice bounds the mitigation claim.
assumptions (5)
  • domain assumption A 0 logit threshold is the operational decision boundary for deepfake detectors
    Sec. 3 defines higher scores as more likely fake and sets the boundary at 0 following the detectors' native design; the paper's FPR and FNR claims inherit this assumption.
  • domain assumption The normative modules of JPEG AI force similar artifacts across compliant implementations
    Sec. 2 states that the normative decoder means images decoded by different manufacturers 'will likely display similar characteristics and forensic artifacts', which is the bridge from one reference implementation to the standard.
  • domain assumption JPEG AI Verification Model 7.0 (HOP) is an adequate proxy for the final standard
    Sec. 4.3 compresses all test data with this version of the reference software; the six bitrates come from the JPEG AI Call for Evidence targets.
  • domain assumption Average Fourier spectra of noise residuals are a valid forensic artifact probe
    Sec. 2 follows the established residual-spectrum procedure of [32, 63] to motivate that JPEG AI leaves generation-like peaks.
  • domain assumption Unweighted averaging across datasets summarizes detector behavior fairly
    Tables 1, 3, and 5 report averages over datasets of different sizes and content, treating each dataset as one unit.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Is JPEG AI going to change image forensics?." pith.science (2026). https://pith.science/paper/6W4H5YRJ

@misc{pith2026241203261,
  author       = {Pith},
  title        = {Pith review of: Is JPEG AI going to change image forensics?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6W4H5YRJ}},
  note         = {Machine review of arXiv:2412.03261}
}
read the original abstract

In this paper, we investigate the counter-forensic effects of the new JPEG AI standard based on neural image compression, focusing on two critical areas: deepfake image detection and image splicing localization. Neural image compression leverages advanced neural network algorithms to achieve higher compression rates while maintaining image quality. However, it introduces artifacts that closely resemble those generated by image synthesis techniques and image splicing pipelines, complicating the work of researchers when discriminating pristine from manipulated content. We comprehensively analyze JPEG AI's counter-forensic effects through extensive experiments on several state-of-the-art detectors and datasets. Our results demonstrate a reduction in the performance of leading forensic detectors when analyzing content processed through JPEG AI. By exposing the vulnerabilities of the available forensic tools, we aim to raise the urgent need for multimedia forensics researchers to include JPEG AI images in their experimental setups and develop robust forensic techniques to distinguish between neural compression artifacts and actual manipulations.

Figures

Figures reproduced from arXiv: 2412.03261 by the authors.

Figure 1
Figure 1. Main findings of our experimental campaign on the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. JPEG AI framework. pressed and transmitted to the receiver, allowing the de￾coder to either reconstruct the image or carry out additional operations directly in the compressed domain exploiting the latent tensor [17]. The entire JPEG AI codec is trained end￾to-end by minimizing a rate-distortion loss function. By ad￾justing the weight of the rate term, different models can be trained to achieve variable-rate compres… view at source ↗
Figure 3
Figure 3. Example of JPEG AI artifacts left at different BPP values in the Fourier and spatial domains. Left block: FFHQ dataset; right [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Scores distribution for [42] over COCO pristine samples (top row) and their synthetic counterpart samples (bottom row) compressed at different BPP values. Impact of compression ratio. We begin our analysis by illustrating the effect of JPEG AI on the score distribution…
Figure 5
Figure 5. Figure 5: Scores distribution for [34]-A over LSUN pristine sam￾ples (top row) and their synthetic counterpart (bottom row) com￾pressed with JPEG at different QF values. JPEG compression shifts the scores distribution of deepfake images below the 0, effectively “canceling” out f…
Figure 6
Figure 6. Figure 6: Output of TruFor [43] and MMFusion [85] for the same DSO-1 input under different JPEG AI compression settings, to￾gether with the JPEG QF = 65 setting. Other examples available in Sec. 10 of the supplementary material. ization metrics illustrated in Sec. 4 to quantify …
Figure 7
Figure 7. Figure 7: Scores distribution over the different datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Scores distribution over the ImageNet, COCO, and FFHQ datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Scores distribution over the LSUN, LAION, and RAISE datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]
Figure 10
Figure 10. Figure 10: Scores distribution over the Imagenet, COCO, and CelebA datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p021_10.png]
Figure 11
Figure 11. Figure 11: Scores distribution over the LSUN and LAION datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Scores distribution over the Imagenet, COCO, and CelebA datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Scores distribution over the LSUN and LAION datasets of the [ [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 15
Figure 15. Figure 15: Output of the considered detectors for the same DS01 [PITH_FULL_IMAGE:figures/full_fig_p027_15.png]
Figure 16
Figure 16. Figure 16: Output of the considered detectors for the same DS01 [PITH_FULL_IMAGE:figures/full_fig_p028_16.png]
Figure 19
Figure 19. Figure 19: Output of MMFusion for DSO-1 double JPEG AI com [PITH_FULL_IMAGE:figures/full_fig_p029_19.png]
Figure 20
Figure 20. Figure 20: Output of MMFusion for DSO-1 double JPEG AI com [PITH_FULL_IMAGE:figures/full_fig_p030_20.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. Unmasking Synthetic Realities in Generative AI: A Comprehensive Review of Adversarially Robust Deepfake Detection Systems

    cs.CR 2025-07 conditional novelty 3.0 of 10

    A systematic review of deepfake detection finds a pervasive lack of adversarial robustness evaluation across all modalities and calls for resilient, modality-agnostic detectors.

Reference graph

Works this paper leans on

115 extracted references · 71 canonical work pages · cited by 1 Pith paper

  1. [1]

    Jpeg call for evidence on learning-based image coding solu- tions, 2020. © 2020. All rights reserved. 2, 4, 5, 3

    2020

  2. [2]

    Dall·e mini, 2021. 2

    2021

  3. [3]

    https:// ds.jpeg.org/whitepapers/jpeg- ai- white- paper.pdf, 2021

    White paper on jpeg ai scope and framework. https:// ds.jpeg.org/whitepapers/jpeg- ai- white- paper.pdf, 2021. Accessed: 2024-11-11. 1, 5

    2021

  4. [4]

    https : / / ds

    Jpeg trust: an international standard for establishing trust in digital media. https : / / ds . jpeg . org / whitepapers / jpeg - trust - whitepaper . pdf,

  5. [5]

    https : / / contentauthenticity

    Content authenticity initiative. https : / / contentauthenticity . org/, 2024. Accessed: 2024-11-11. 8

    2024

  6. [6]

    https://openai.com/index/dall- e- 2/, 2024

    Dall·e 2. https://openai.com/index/dall- e- 2/, 2024. Accessed: 2024-11-13. 2

    2024

  7. [7]

    https://openai.com/index/dall- e- 3/, 2024

    Dall·e 3. https://openai.com/index/dall- e- 3/, 2024. Accessed: 2024-11-13. 2

    2024

  8. [8]

    https : / / www

    Adobe firefly. https : / / www . adobe . com / it / products/firefly.html, 2024. Accessed: 2024-11-

    2024

Show all 115 references
  1. [9]

    Jpeg ai reaches draft international standard stage, 2024. ©

    2024

  2. [10]

    Jpeg ai reference software, 2024. © 2024. All rights re- served. 2, 4, 3

    2024

  3. [11]

    All rights reserved. 2

  4. [12]

    https://huggingface.co/ CompVis / stable - diffusion - v1 - 3, 2024

    Stable diffusion v1-3. https://huggingface.co/ CompVis / stable - diffusion - v1 - 3, 2024. Ac- cessed: 2024-11-13. 2

    2024

  5. [13]

    https://www.midjourney.com/home,

    Midjourney. https://www.midjourney.com/home,

  6. [14]

    Accessed: 2024-11-13. 2

    2024

  7. [15]

    https : / / huggingface

    Stable diffusion xl. https : / / huggingface . co / stabilityai/stable- diffusion- xl- base- 0. 9, 2024. Accessed: 2024-11-13. 2

    2024

  8. [16]

    https://huggingface.co/ CompVis / stable - diffusion - v1 - 4, 2024

    Stable diffusion v1-4. https://huggingface.co/ CompVis / stable - diffusion - v1 - 4, 2024. Ac- cessed: 2024-11-13. 2

    2024

  9. [17]

    https://huggingface.co/ spaces/stabilityai/stable-diffusion , 2024

    Stable diffusion 2.1. https://huggingface.co/ spaces/stabilityai/stable-diffusion , 2024. Accessed: 2024-11-13. 2

    2024

  10. [18]

    Simoncelli

    Johannes Ball ´e, Valero Laparra, and Eero P. Simoncelli. End-to-end optimized image compression. In International Conference on Learning Representations, 2017. 1

    2017

  11. [19]

    Jpeg ai becomes an international standard, 2025. © 2025. All rights reserved. 1, 2

    2025

  12. [20]

    The JPEG AI standard: Providing efficient human and machine visual data consumption

    Joao Ascenso, Elena Alshina, and Touradj Ebrahimi. The JPEG AI standard: Providing efficient human and machine visual data consumption. IEEE Multimedia, 30(1):100–111,

  13. [21]

    Stamm, and Benedetta Tondi

    Mauro Barni, Matthew C. Stamm, and Benedetta Tondi. Adversarial multimedia forensics: Overview and challenges ahead. In 2018 26th European Signal Processing Conference (EUSIPCO), 2018. 8

    2018

  14. [22]

    Variational image compres- sion with a scale hyperprior

    Johannes Ball ´e, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compres- sion with a scale hyperprior. In International Conference on Learning Representations, 2018. 1

    2018

  15. [23]

    Aligned and non-aligned double jpeg detection using convolutional neural networks

    Mauro Barni, Luca Bondi, Nicol `o Bonettini, Paolo Bestagini, Andrea Costanzo, Marco Maggini, Benedetta Tondi, and Stefano Tubaro. Aligned and non-aligned double jpeg detection using convolutional neural networks. Jour- nal of Visual Communication and Image Representation, 49:...

    2017

  16. [24]

    Ai-based com- pression: A new unintended counter attack on jpeg-related image forensic detectors? In 2022 IEEE International Con- ference on Image Processing (ICIP), 2022

    Alexandre Berthet and Jean-Luc Dugelay. Ai-based com- pression: A new unintended counter attack on jpeg-related image forensic detectors? In 2022 IEEE International Con- ference on Image Processing (ICIP), 2022. 2

    2022

  17. [25]

    Frequency-domain analysis of traces for the detection of ai-based compression

    Sandra Bergmann, Denise Moussa, Fabian Brand, Andr ´e Kaup, and Christian Riess. Frequency-domain analysis of traces for the detection of ai-based compression. In 2023 11th International Workshop on Biometrics and Forensics (IWBF), 2023. 1, 2

    2023

  18. [26]

    Forensic analysis of ai- compression traces in spatial and frequency domain

    Sandra Bergmann, Denise Moussa, Fabian Brand, Andr ´e Kaup, and Christian Riess. Forensic analysis of ai- compression traces in spatial and frequency domain. Pattern Recognition Letters, 180:41–47, 2024. 1, 2

    2024

  19. [27]

    Evading deepfake-image detectors with white- and black-box attacks

    Nicholas Carlini and Hany Farid. Evading deepfake-image detectors with white- and black-box attacks. In 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pages 2804–2813, 2020. 8

    2020

  20. [28]

    Training strategies and data augmentations in cnn-based deepfake video detection

    Luca Bondi, Edoardo Daniele Cannas, Paolo Bestagini, and Stefano Tubaro. Training strategies and data augmentations in cnn-based deepfake video detection. 2020. 8

    2020

  21. [29]

    Large scale GAN training for high fidelity natural image synthe- sis

    Andrew Brock, Jeff Donahue, and Karen Simonyan. Large scale GAN training for high fidelity natural image synthe- sis. In International Conference on Learning Representa- tions, 2019. 2

    2019

  22. [30]

    Variable rate deep image compression with a conditional autoencoder

    Yochai Choi, Mostafa El-Khamy, and Jungwon Lee. Variable rate deep image compression with a conditional autoencoder. In 2019 IEEE/CVF International Conference on Computer Vision (ICCV), pages 3146–3154, 2019. 1

    2019

  23. [31]

    Faria, H ´elio Pedrini, Ricardo da S

    Tiago Carvalho, F ´abio A. Faria, H ´elio Pedrini, Ricardo da S. Torres, and Anderson Rocha. Illuminant-based trans- formed spaces for image forensics. IEEE Transactions on Information Forensics and Security, 2016. 4

    2016

  24. [32]

    Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation

    Yunjey Choi, Minje Choi, Munyoung Kim, Jung-Woo Ha, Sunghun Kim, and Jaegul Choo. Stargan: Unified genera- tive adversarial networks for multi-domain image-to-image translation. In Proceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition, 2018. 2

    2018

  25. [33]

    Splicebuster: A new blind image splicing detector

    Davide Cozzolino, Giovanni Poggi, and Luisa Verdoliva. Splicebuster: A new blind image splicing detector. In 2015 IEEE International Workshop on Information Forensics and Security (WIFS), pages 1–6. IEEE, 2015. 8

    2015

  26. [34]

    Intriguing properties of syn- thetic images: From generative adversarial networks to dif- fusion models

    Riccardo Corvi, Davide Cozzolino, Giovanni Poggi, Koki Nagano, and Luisa Verdoliva. Intriguing properties of syn- thetic images: From generative adversarial networks to dif- fusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CV...

    2023

  27. [35]

    On the detection of synthetic images generated by diffusion mod- els

    Riccardo Corvi, Davide Cozzolino, Giada Zingarini, Gio- vanni Poggi, Koki Nagano, and Luisa Verdoliva. On the detection of synthetic images generated by diffusion mod- els. In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1–5, 2023. 2...

    2023

  28. [36]

    Exposing digital image forgeries by illumination color classification

    Tiago Jos ´e de Carvalho, Christian Riess, Elli Angelopoulou, H´elio Pedrini, and Anderson de Rezende Rocha. Exposing digital image forgeries by illumination color classification. IEEE Transactions on Information Forensics and Security ,

  29. [37]

    Raising the Bar of AI-generated Image Detection with CLIP

    Davide Cozzolino, Giovanni Poggi, Riccardo Corvi, Matthias Nießner, and Luisa Verdoliva. Raising the Bar of AI-generated Image Detection with CLIP. In IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2024. 3, 5, 6, 1, 2, 7, 8, 13

    2024

  30. [38]

    Raise – a raw images dataset for digi- tal image forensics

    Duc-Tien Dang-Nguyen, Cecilia Pasquini, Valentina Conot- ter, and Giulia Boato. Raise – a raw images dataset for digi- tal image forensics. In Proceedings of the ACM Multimedia Systems Conference, 2015. 4

    2015

  31. [39]

    Casia image tam- pering detection evaluation database

    Jing Dong, Wei Wang, and Tieniu Tan. Casia image tam- pering detection evaluation database. In 2013 IEEE China Summit and International Conference on Signal and Infor- mation Processing, pages 422–426, 2013. 4

    2013

  32. [40]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248–255, 2009. 4

    2009

  33. [41]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alex Nichol. Diffusion models beat gans on image synthesis. In Proceedings of the 35th Inter- national Conference on Neural Information Processing Sys- tems, 2021. 2

    2021

  34. [42]

    Are GAN generated images easy to detect? a critical analysis of the state-of-the- art

    Diego Gragnaniello, Davide Cozzolino, Francesco Marra, Giovanni Poggi, and Luisa Verdoliva. Are GAN generated images easy to detect? a critical analysis of the state-of-the- art. In IEEE International Conference on Multimedia and Expo (ICME), pages 1–6. IEEE, 2021. 1, 3, 5, 6,...

    2021

  35. [43]

    Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributions

    Ricard Durall, Margret Keuper, and Janis Keuper. Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributions. In IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2020. 1

    2020

  36. [44]

    Taming transformers for high-resolution image synthesis

    Patrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image synthesis. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12873–12883, 2021. 2

    2021

  37. [45]

    A taxonomy of miscompres- sions: Preparing image forensics for neural compression,

    Nora Hofer and Rainer B ¨ohme. A taxonomy of miscompres- sions: Preparing image forensics for neural compression,

  38. [46]

    Trufor: Leveraging all-round clues for trustworthy image forgery detection and localiza- tion

    Fabrizio Guillaro, Davide Cozzolino, Avneesh Sud, Nicholas Dufour, and Luisa Verdoliva. Trufor: Leveraging all-round clues for trustworthy image forgery detection and localiza- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR),...

    2023

  39. [47]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In 2016 IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR) , pages 770– 778, 2016. 1

    2016

  40. [48]

    Adversarial deepfakes: Evaluating vulnerability of deepfake detectors to adversar- ial examples

    Shehzeen Hussain, Paarth Neekhara, Malhar Jere, Farinaz Koushanfar, and Julian McAuley. Adversarial deepfakes: Evaluating vulnerability of deepfake detectors to adversar- ial examples. In Proceedings of the IEEE/CVF Winter Con- ference on Applications of Computer Vision (WACV)...

    2021

  41. [49]

    High- fidelity image compression with score-based generative models

    Emiel Hoogeboom, Eirikur Agustsson, Fabian Mentzer, Luca Versari, George Toderici, and Lucas Theis. High- fidelity image compression with score-based generative models. arXiv preprint arXiv:2305.18231, 2023. 1

    2023 arXiv

  42. [50]

    Hsu and S.-F

    Y .-F. Hsu and S.-F. Chang. Detecting image splicing using geometry invariants and camera characteristics consistency. In International Conference on Multimedia and Expo, 2006. 4

    2006

  43. [51]

    Jpeg ai use cases and require- ments

    ISO/IEC JTC 1/SC 29/WG 1. Jpeg ai use cases and require- ments. Technical Report N100724, International Organiza- tion for Standardization and International Electrotechnical Commission, 2023. JPEG AI use case and requirements. 1

    2023

  44. [52]

    Coding of still pictures

    ISO/IEC JTC 1/SC 29/WG 1. Coding of still pictures. Tech- nical Report N83058, International Organization for Stan- dardization and International Electrotechnical Commission,

  45. [53]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 2

    2019

  46. [54]

    Jpeg ai overview slides

    ISO/IEC JTC 1/SC 29/WG 1. Jpeg ai overview slides. Tech- nical Report N100658, International Organization for Stan- dardization and International Electrotechnical Commission,

  47. [55]

    Analyzing and improv- ing the image quality of StyleGAN

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improv- ing the image quality of StyleGAN. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020. 2

    2020

  48. [56]

    An im- age is worth 16x16 words: Transformers for image recogni- tion at scale

    Alexander Kolesnikov, Alexey Dosovitskiy, Dirk Weis- senborn, Georg Heigold, Jakob Uszkoreit, Lucas Beyer, Matthias Minderer, Mostafa Dehghani, Neil Houlsby, Syl- vain Gelly, Thomas Unterthiner, and Xiaohua Zhai. An im- age is worth 16x16 words: Transformers for image recogni-...

    2021

  49. [57]

    Progressive growing of gans for improved quality, stability, and variation

    Tero Karras, Timo Aila, Samuli Laine, and Jaakko Lehtinen. Progressive growing of gans for improved quality, stability, and variation. In Proceedings of the International Confer- ence on Learning Representations, 2018. 1, 2

    2018

  50. [58]

    Aaron, I

    Zhi Li, A. Aaron, I. Katsavounidis, A. Moorthy, and M. Manohara. Toward a practical perceptual video quality met- ric. https://netflixtechblog.com/toward- 10 a - practical - perceptual - video - quality - metric-653f208b9652. Accessed: 2024-11-12. 3

    2024

  51. [59]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. In 2019 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 4396–4405, 2019. 2, 4

    2019

  52. [60]

    Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll ´ar, and C

    Tsung-Yi Lin, Michael Maire, Serge J. Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll ´ar, and C. Lawrence Zitnick. Microsoft coco: Common objects in context. In European Conference on Computer Vision, 2014. 4

    2014

  53. [61]

    Deep learning face attributes in the wild

    Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep learning face attributes in the wild. In Proceedings of International Conference on Computer Vision (ICCV), 2015. 4

    2015

  54. [62]

    Simoncelli

    Valero Laparra, Johannes Ball ´e, Anthony Berardino, and Eero P. Simoncelli. Perceptual image quality assessment us- ing a normalized laplacian pyramid. In S&T Symposium on Electronic Imaging: Conf. on Human Vision and Electronic Imaging, 2016. 2

    2016

  55. [63]

    When synthetic traces hide real content: Analysis of stable diffu- sion image laundering

    Sara Mandelli, Paolo Bestagini, and Stefano Tubaro. When synthetic traces hide real content: Analysis of stable diffu- sion image laundering. In Proceedings of the IEEE Inter- national Workshop on Information Forensics and Security (WIFS), 2024. 1, 2, 3, 5, 6, 7, 8, 13

    2024

  56. [64]

    Neural image compression: Generaliza- tion, robustness, and spectral biases

    Kelsey Lieberman, James Diffenderfer, Charles Godfrey, and Bhavya Kailkhura. Neural image compression: Generaliza- tion, robustness, and spectral biases. In Advances in Neural Information Processing Systems, 2023. 1

    2023

  57. [65]

    Channel-wise autoregres- sive entropy models for learned image compression

    David Minnen and Saurabh Singh. Channel-wise autoregres- sive entropy models for learned image compression. InIEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP). IEEE, 2020

    2020

  58. [66]

    Toderici

    David Minnen, Johannes Ball ´e, and George D. Toderici. Joint autoregressive and hierarchical priors for learned im- age compression. Advances in Neural Information Process- ing Systems, 31, 2018. 1

    2018

  59. [67]

    Training cnns in presence of jpeg compression: Multimedia forensics vs computer vision

    Sara Mandelli, Nicolo Bonettini, Paolo Bestagini, and Ste- fano Tubaro. Training cnns in presence of jpeg compression: Multimedia forensics vs computer vision. In 2020 IEEE In- ternational Workshop on Information Forensics and Security (WIFS), pages 1–6. IEEE, 2020. 6, 8

    2020

  60. [68]

    Synthetic realities and artificial intelligence-generated con- tents

    Daniel Moreira, S ´ebastien Marcel, and Anderson Rocha. Synthetic realities and artificial intelligence-generated con- tents. IEEE Security & Privacy, 22(3):7–10, 2024. 8

    2024

  61. [69]

    High-fidelity generative image compression

    Fabian Mentzer, Eirikur Agustsson, Michael Tschannen, Radu Timofte, and Luc Van Gool. High-fidelity generative image compression. arXiv preprint arXiv:2006.09965, 2020. 1

    2006 arXiv

  62. [70]

    Towards uni- versal fake image detectors that generalize across genera- tive models

    Utkarsh Ojha, Yuheng Li, and Yong Jae Lee. Towards uni- versal fake image detectors that generalize across genera- tive models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 24480–24489, 2023. 3, 5, 6, 1, 2, 9, 10, 13

    2023

  63. [71]

    Deepfakes and cheap fakes

    Britt Paris and Joan Donovan. Deepfakes and cheap fakes. Data & Society, 2019. 3

    2019

  64. [72]

    Singh, and Ranjan K

    Deepak Mishra, Sharad K. Singh, and Ranjan K. Singh. Deep architectures for image compression: A critical review. Signal Processing, 191:108346, 2022. 2

    2022

  65. [73]

    A. Piva. An overview on image forensics. ISRN Signal Pro- cessing, 2013. 3

    2013

  66. [74]

    GLIDE: Towards photorealis- tic image generation and editing with text-guided diffusion models

    Alexander Quinn Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob Mcgrew, Ilya Sutskever, and Mark Chen. GLIDE: Towards photorealis- tic image generation and editing with text-guided diffusion models. In Proceedings of the 39th International Conferenc...

    2022

  67. [75]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In Proceedings of the 38th International Conference on Machine Learning , 2021. 2

    2021

  68. [76]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. 1, 2

    2022

  69. [77]

    Gaugan: semantic image synthesis with spatially adap- tive normalization

    Taesung Park, Ming-Yu Liu, Ting-Chun Wang, and Jun-Yan Zhu. Gaugan: semantic image synthesis with spatially adap- tive normalization. ACM SIGGRAPH 2019 Real-Time Live!,

    2019

  70. [78]

    Laion-400m: Open dataset of clip-filtered 400 million image-text pairs,

    Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk, Clayton Mullis, Aarush Katta, Theo Coombes, Jenia Jitsev, and Aran Komatsuzaki. Laion-400m: Open dataset of clip-filtered 400 million image-text pairs,

  71. [79]

    Avatar fingerprinting for autho- rized use of synthetic talking-head videos

    Ekta Prashnani, Koki Nagano, Shalini De Mello, David Lue- bke, and Orazio Gallo. Avatar fingerprinting for autho- rized use of synthetic talking-head videos. arXiv preprint arXiv:2305.03713, 2023. 8

    2023 arXiv

  72. [80]

    M. C. Stamm, Min Wu, and K. J. R. Liu. Information foren- sics: An overview of the first decade. IEEE Access, 1:167– 200, 2013. 1

    2013

  73. [81]

    Rethinking the up-sampling op- erations in cnn-based generative network for generalizable deepfake detection

    Chuangchuang Tan, Yao Zhao, Shikui Wei, Guanghua Gu, Ping Liu, and Yunchao Wei. Rethinking the up-sampling op- erations in cnn-based generative network for generalizable deepfake detection. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (...

    2024

  74. [82]

    Exploring the adver- sarial robustness of clip for ai-generated image detection

    Vincenzo De Rosa, Fabrizio Guillaro, Giovanni Poggi, Da- vide Cozzolino, and Luisa Verdoliva. Exploring the adver- sarial robustness of clip for ai-generated image detection. In Proceedings of the IEEE International Workshop on Infor- mation Forensics and Security (WIFS), 2024. 8

    2024

  75. [83]

    Lossy image compression with compressive autoen- coders

    Lucas Theis, Wenzhe Shi, Andrew Cunningham, and Ferenc Husz´ar. Lossy image compression with compressive autoen- coders. 2017. 1

    2017

  76. [84]

    Sheikh and Alan C

    H.R. Sheikh and Alan C. Bovik. Image information and vi- sual quality. In IEEE International Conference on Acoustics, Speech, and Signal Processing, 2004. 2

    2004

  77. [85]

    Exploring multi-modal fusion for image manipulation detection and lo- calization

    Konstantinos Triaridis and Vasileios Mezaris. Exploring multi-modal fusion for image manipulation detection and lo- calization. In Proc. 30th Int. Conf. on MultiMedia Modeling (MMM 2024), 2024. 4, 7, 8 11

    2024

  78. [86]

    Jpeg ai image compression visual artifacts: Detection methods and dataset, 2024

    Daria Tsereh, Mark Mirgaleev, Ivan Molodetskikh, Roman Kazantsev, and Dmitriy Vatolin. Jpeg ai image compression visual artifacts: Detection methods and dataset, 2024. 1, 2

    2024

  79. [87]

    Mingxing Tan and Quoc V . Le. Efficientnet: Rethinking model scaling for convolutional neural networks. In Inter- national Conference on Machine Learning, (ICML) 2019 , pages 6105–6114. PMLR, 2019. 2

    2019

  80. [88]

    CNN-generated images are sur- prisingly easy to spot

    Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, and Alexei A Efros. CNN-generated images are sur- prisingly easy to spot... for now. In IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8695– 8704, 2020. 1, 3, 5, 6, 8, 2, 4, 10, 12

    2020

  81. [89]

    Toderici, Damien Vincent, Nick Johnston, Sung Jin Hwang, David Minnen, Joel Shor, and Michele Covell

    George D. Toderici, Damien Vincent, Nick Johnston, Sung Jin Hwang, David Minnen, Joel Shor, and Michele Covell. Full resolution image compression with recurrent neural networks. IEEE Conference on Computer Vision and Pattern Recognition, pages 5306–5314, 2017. 1

    2017

  82. [90]

    Simoncelli, and Alan C

    Zhou Wang, Eero P. Simoncelli, and Alan C. Bovik. Multi- scale structural similarity for image quality assessment. In 37th IEEE Asilomar Conference on Signals, Systems and Computers, 2003. 2

    2003

  83. [91]

    Coverage – a novel database for copy-move forgery detection

    Bihan Wen, Ye Zhu, Ramanathan Subramanian, Tian-Tsong Ng, Xuanjing Shen, and Stefan Winkler. Coverage – a novel database for copy-move forgery detection. In IEEE Interna- tional Conference on Image processing (ICIP) , pages 161– 165, 2016. 4

    2016

  84. [92]

    Media forensics and deepfakes: an overview

    Luisa Verdoliva. Media forensics and deepfakes: an overview. IEEE journal of selected topics in signal process- ing, 14(5):910–932, 2020. 3

    2020

  85. [93]

    Feng Yang, Lucas Herranz, Yong Cheng, and M. G. Moze- rov. Slimmable compressive autoencoders for practical neu- ral image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4998–5007, 2021. 1

    2021

  86. [94]

    Deepfake detection: A comprehensive sur- vey from the reliability perspective

    Tianyi Wang, Xin Liao, Kam Pui Chow, Xiaodong Lin, and Yinglong Wang. Deepfake detection: A comprehensive sur- vey from the reliability perspective. ACM Comput. Surv. ,

  87. [95]

    LSUN: Construction of a large-scale image dataset using deep learning with humans in the loop

    Fisher Yu, Ari Seff, Yinda Zhang, Shuran Song, Thomas Funkhouser, and Jianxiong Xiao. LSUN: Construction of a large-scale image dataset using deep learning with humans in the loop. Technical Report 1506.03365, arXiv preprint,

    arXiv

  88. [96]

    Large-scale evaluation of splicing localization algorithms for web images

    Markos Zampoglou, Symeon Papadopoulos, and Yiannis Kompatsiaris. Large-scale evaluation of splicing localization algorithms for web images. Multimedia Tools and Applica- tions, 76(4):4801–4834, 2017. 8

    2017

  89. [97]

    Robust image forgery detection against transmission over online social networks

    Haiwei Wu, Jiantao Zhou, Jinyu Tian, Jun Liu, and Yu Qiao. Robust image forgery detection against transmission over online social networks. IEEE Transactions on Information Forensics and Security, 17:443–456, 2022. 4, 7, 8

    2022

  90. [98]

    Unified multivariate gaussian mixture for efficient neural image compression

    Xiaoheng Zhu, Jingwen Song, Liyuan Gao, Fanglong Zheng, and Heng Tao Shen. Unified multivariate gaussian mixture for efficient neural image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 17612–17621, 2022. 1 12 I...

    2022

  91. [99]

    Lossy image compression with conditional diffusion models

    Ruihan Yang and Stephan Mandt. Lossy image compression with conditional diffusion models. In Advances in Neural Information Processing Systems, 2023. 1

    2023

  92. [102]

    Fsim: a feature similarity index for image quality assess- ment

    Lin Zhang, Lei Zhang, Xuanqin Mou, and David Zhang. Fsim: a feature similarity index for image quality assess- ment. IEEE Transactions on Image Processing, 20(8):2378– 2386, 2011. 2

    2011

  93. [104]

    Neural image compression and JPEG AI specification Neural image compression basics. In Neural Image Com- pression (NIC), an input image is transformed into a com- pact file or bitstream for storage or transmission, which is later used to reconstruct the original image. The bas...

    2017

  94. [105]

    fakeness

    Experimental setup additional details Experiments reproducibility. All code and additional fig- ures and tables for all detectors is available at the follow- ing URL https://github.com/polimi- ispl/ jpeg-ai-antifor. Deepfake image detectors. We provide some more infor- mation ...

  95. [106]

    Deepfake image detection additional results Preliminary analysis on synthetic detection. Tab. 7 re- ports the results for the simple case of synthetic image de- tection without JPEG AI image compression of the test sam- ples. We report the complete numbers for all detectors on...

  96. [107]

    − 0.89 0 .96 − 0.93 − −

  97. [108]

    − 0.87 0 .96 − 0.89 − −

  98. [109]

    0.98 0 .99 1 .00 0 .99 − 0.96 − [34]-A 0.95 0 .97 0 .97 − 0.96 0 .96 0 .80 [34]-B 0.79 0 .83 0 .89 − 0.88 0 .74 0 .86

  99. [110]

    0.94 0 .91 1 .00 1 .00 − 0.99 −

  100. [111]

    shift” of the scores distribution towards the right, i.e., towards “deepfake

    0.86 0 .87 0 .71 1 .00 1 .00 0 .98 0 .83 • Feature Similarity (FSIM) [97]: a metric based on phase congruency and gradient magnitude to assess quality. The value range for FS is between 0 and 1, with 1 indicating the highest quality; • Multi-Scale SIMilarity index (MS-SSIM) [9...

  101. [112]

    14 to Fig

    Image splicing localization additional re- sults Fig. 14 to Fig. 17 report some examples of output maps ˜M for the TruFor and MMFusion detectors. In all plots, the detectors process the same DSO1 input image compressed with different JPEG AI BPP. As we reported in Sec. 5 of th...

  102. [113]

    Therefore, in the following, we briefly analyze some of the implications of double JPEG AI compression

    Double JPEG AI compression The study field relative to multiple JPEG compression is lively [73]. Therefore, in the following, we briefly analyze some of the implications of double JPEG AI compression. Please notice that this analysis is not exhaustive, and we plan to explore t...

  103. [114]

    detector extracted from pristine images compressed in the scenarios illustrated in Sec. 5.3, namely i) single JPEG AI compression, ii) double JPEG AI compression where the original-format images have been first compressed with the lowest BPP considered, i.e., BPP = 0.12, and t...

  104. [115]

    D-JPEGAIBPP1st = 0.12 Input sample ˜Msingle ˜Mdouble D-JPEGAIBPP1st = 0.25 D-JPEGAIBPP1st = 0.5 D-JPEGAIBPP1st = 0.75 D-JPEGAIBPP1st = 1.0 D-JPEGAIBPP1st = 2.0 Figure 19

    detector considering different JPEG AI compression stages. D-JPEGAIBPP1st = 0.12 Input sample ˜Msingle ˜Mdouble D-JPEGAIBPP1st = 0.25 D-JPEGAIBPP1st = 0.5 D-JPEGAIBPP1st = 0.75 D-JPEGAIBPP1st = 1.0 D-JPEGAIBPP1st = 2.0 Figure 19. Output of MMFusion for DSO-1 double JPEG AI com...

  105. [2019]

    JPEG AI investigations start. 1

  106. [2023]

    JPEG AI overview. 2, 1

  107. [2024]

    Accessed: 2024-11-11. 8

    2024

Pith tools

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