Pith. sign in

REVIEW 2 major objections 6 minor 61 references

Objective light-field metrics track human scores for pure coding artifacts but drop when view synthesis is added.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 02:05 UTC pith:5UENOIKW

load-bearing objection Solid JPEG Pleno-style package: usable public LF benchmark plus clear evidence that FR metrics drop under interpolation/3DGS, with one real scope limit (passive 2D presentation). the 2 major comments →

arxiv 2607.03494 v1 pith:5UENOIKW submitted 2026-07-03 cs.CV cs.MMeess.IV

Towards Standardized Light Field Quality Assessment: Hybrid Subjective Benchmarking and Objective Metric Evaluation

Saeed Mahmoudpour , Mylene C. Q. Farias , Gi-Mun Um , Myllena A. Prado , Ismael Seidel , Leonardo de Sousa Marques , Leonardo Andrade , Shengyang Zhao
show 1 more author
Carla L Pagliari
This is my paper
classification cs.CV cs.MMeess.IV
keywords light field quality assessmentsubjective evaluationobjective metricsview synthesisJPEG Plenohybrid pairwise comparison3D Gaussian Splattingview pooling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper builds a standardized workflow for light-field quality assessment that can serve coding and reconstruction pipelines. It pairs a hybrid human protocol—reference-anchored ratings refined by selective pairwise comparisons only in ambiguous regions—with a benchmark that includes both conventional compression and modern view-synthesis distortions (interpolation and 3D Gaussian Splatting). Across two independent observer cohorts the hybrid scores prove consistent, giving a fine-grained perceptual ground truth. When popular full-reference metrics are scored against that ground truth, several perform well on coding-only stimuli yet lose accuracy once reconstruction artifacts appear; how view-level scores are pooled also changes the ranking. The released dataset and protocol therefore expose a concrete limitation of today’s metrics and give future standardization a reproducible test bed.

Core claim

Several full-reference objective metrics achieve strong agreement with hybrid subjective scores on coding-only light-field stimuli, but their prediction accuracy consistently declines once interpolation or 3D-Gaussian-Splatting reconstruction distortions are included; view-pooling strategy is therefore a necessary design choice for future light-field metrics.

What carries the argument

Hybrid DSCS+PC protocol: Double Stimulus Comparison Scale ratings that anchor quality to a reference, followed by observer-specific pairwise comparisons only inside same-category bins, fused by Thurstone Case-V scaling into a continuous degradation scale.

Load-bearing premise

That showing light-field views as passive serpentine video on ordinary 2-D displays yields quality scores that still represent what people would judge under interactive free-viewpoint exploration.

What would settle it

Re-run the identical hybrid protocol on the same 144 stimuli with an interactive multi-view interface; if the new scores reorder reconstruction stimuli relative to coding-only ones, or reverse metric SRCC rankings, the claim that the present ground truth generalizes is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. This paper presents a standardization-oriented workflow for light-field (LF) quality assessment developed in the JPEG Pleno Part 7 context. It integrates (i) a 144-stimulus benchmark spanning coding-only (JPLM 4D-TM, x265) and sparse-to-dense reconstruction distortions (RIFE/SepConv++ interpolation and 3DGS), (ii) a hybrid DSCS+PC subjective protocol that anchors ratings to a reference and selectively refines ambiguous Slightly-worse/Worse bins via pairwise comparison, and (iii) systematic evaluation of classical, deep-feature, video, and immersive FR metrics under mean, Minkowski, and worst-view pooling. Subjective reliability is supported by two-lab screening (N=50), cross-cohort SRCC 0.943, and leave-one-observer-out log-likelihood gains for hybrid refinement. The central empirical result is that several metrics (e.g., ST-LPIPS, CVVDP, IW-SSIM) agree well with hybrid scores on coding-only stimuli but degrade when view-synthesis artifacts are included, with residual analysis showing systematic under-prediction of reconstruction degradation after coding-only logistic calibration; view-pooling strategy is metric-dependent and material for future LF metrics. The annotated dataset is released publicly.

Significance. If the reported protocol and findings hold, the work supplies a reproducible, multi-lab subjective target and evaluation pipeline that is directly usable for JPEG Pleno objective-metric CfP activity and for codec/reconstruction benchmarking beyond traditional coding-only tests. Strengths that should be credited include: public release of the subjectively annotated set; explicit multi-cohort consistency checks (Table II, Fig. 5); leave-one-observer-out validation of selective PC refinement (Table III); residual transfer analysis across distortion families (Fig. 6); and a controlled pooling study (Table VI). These elements go beyond a typical dataset paper and give the community a concrete baseline for metric design under emerging sparse-coding + synthesis pipelines.

major comments (2)
  1. [§IV.A, §VII] §IV.A and §VII: All subjective scores are obtained from passive serpentine pseudo-video on calibrated 2D displays, with PC restricted to Slightly-worse/Worse bins and capped at 200 trials. The design implications for future LF metrics—especially the claim that worst-view/Minkowski pooling should be considered because observers are driven by localized poor views—rest on which angular inconsistencies dominate under this presentation. The manuscript should state more explicitly as a scope limitation that rankings and optimal pooling may shift under interactive free-viewpoint navigation, and should avoid language that treats the pooling conclusions as presentation-invariant without additional evidence.
  2. [§V.B–C, Table VI] §V.B–C and Table VI: Pooling hyperparameters (Worst-X% with X∈{5,10,20,30}; Minkowski p∈{3,5,7,9}) and the four-parameter logistic mapping are free choices fitted per metric/subset. The paper reports that non-mean pooling helps several deep metrics and can hurt classical fidelity metrics, but does not show sensitivity of the ranking of top metrics (ST-LPIPS, CVVDP, IW-SSIM) to these choices or to the DSCS SRCC≥0.7 / PC z<−3 screening thresholds. A short sensitivity check (or fixed-protocol statement that rankings are stable under reasonable alternatives) is needed so that the “pooling matters for future design” claim is not tied to a single hyperparameter setting.
minor comments (6)
  1. [§III.D] §III.D: The choice of RIFE vs SepConv++ is described as “based on visual suitability and baseline,” but no quantitative criterion or per-scene decision table is given. A brief appendix listing which method was used for each scene (and why) would aid reproducibility of the 40 interpolation stimuli.
  2. [Fig. 4] Fig. 4: Scene-wise scores with CIs are informative, but the figure is dense; separating coding-only from reconstruction tracks into two panels (or using consistent marker styles per distortion family) would improve readability.
  3. [Table IV] Table IV: Report whether SRCC/PLCC differences among top metrics are statistically significant (e.g., Steiger’s test or bootstrap CIs). Even a short note would strengthen the ranking statements.
  4. [§IV.C, Eqs. (1)–(2)] §II.A / Eq. (1)–(2): Thurstone Case V is standard; a one-sentence note that the half-vote treatment of DSCS ties is the conventional mid-rank convention (and that PC overwrites ties when available) would help readers less familiar with hybrid scaling.
  5. [Front matter / references] Minor typography: “Myllena A. Prado” / author list spacing; “Myl `ene” accent rendering; “arXiv:2607.03494v1” date line is fine. Ensure consistent “light field” vs “light-field” hyphenation and that all codec toolkit URLs remain accessible at publication.
  6. [Fig. 6] §VI.C.3 / Fig. 6: Residual distributions are central; adding median residual and IQR numerically in the caption or a small table would make the under-prediction claim easier to cite without reading the figure.

Circularity Check

0 steps flagged

No circularity: empirical benchmark of new subjective scores against external objective metrics; self-citations supply only background.

full rationale

The paper's load-bearing claims are empirical measurements on a newly generated 144-stimulus set with 50 screened observers: hybrid DSCS+PC scores (Thurstone Case V reconstruction of observer-specific ratings plus selective PC), cross-cohort SRCC 0.943 and leave-one-out log-likelihood gains (Tables II–III), metric SRCC/PLCC drops from coding-only to full set (Table IV), residual bias after coding-only logistic calibration (Fig. 6), and pooling-strategy effects (Table VI). These quantities are obtained from human judgments and off-the-shelf FR metrics; none is defined in terms of the quantity it is said to predict, none is a parameter fitted to the target and then re-labeled a prediction, and no uniqueness theorem or ansatz is imported from prior author work to force the result. Self-citations ([5], [7], [24]) describe the hybrid protocol's prior analytical motivation and earlier LF-QA surveys; the present scores, residual analysis, and metric tables are independent new data. Standard IQA logistic mapping for PLCC/RMSE is fitted per metric solely for reporting and does not enter the ranking claims. The derivation chain is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 1 invented entities

This is an empirical standardization/benchmark paper, not a first-principles derivation. Load-bearing premises are domain conventions (ITU subjective methods, Thurstone Case V, FR metric suite, mean/worst/Minkowski pooling) plus experimental design choices (scene set, sparse grids, bitrate points, PC bin restriction, passive serpentine viewing). Free parameters are thresholds and caps chosen for screening and session length rather than physics constants. No new physical entities are postulated.

free parameters (6)
  • DSCS reliability SRCC threshold = 0.7
    Observers retained only if leave-one-out SRCC ≥ 0.7 against DMOS (§IV.B); threshold is a conventional but free design choice that affects who enters the final scale.
  • PC outlier z-threshold = z < -3
    PC-stage observers excluded if standardized log-likelihood z < −3 (§IV.B).
  • Selective PC trial cap and bin restriction = 200 trials; bins −1 and −2 only
    PC limited to Slightly-worse and Worse bins and capped at 200 comparisons per observer (~82% of eligible pairs retained) to keep sessions 20–25 min (§IV).
  • Four-parameter logistic mapping β1…β4 per metric/subset = fitted per metric and subset
    Fitted before PLCC/RMSE (§V.C, Eq. 5); standard IQA practice but free parameters of the reported accuracy numbers.
  • Pooling hyperparameters X and p = X in {5,10,20,30}; p in {3,5,7,9}
    Worst-X% with X∈{5,10,20,30} and Minkowski p∈{3,5,7,9} explored in Table VI; choices affect which metrics look best.
  • Bitrate operating points and sparse angular grids = design-selected per scene/codec
    Five coding rates per codec/scene, selected sparse grids (Table I), and per-scene choice of RIFE vs SepConv++ shape the distortion distribution the claims rest on.
axioms (5)
  • domain assumption Thurstone Case V model: preference probability is Φ(qi−qj) with independent unit-variance Gaussian noise; reference fixed at q_ref=0.
    Used to fuse DSCS-derived and PC counts into continuous scores (Eqs. 1–2, §IV.C).
  • domain assumption Passive serpentine pseudo-video on calibrated 2D displays is an adequate proxy for light-field perceptual quality under ITU-R BT.500 viewing conditions.
    Core presentation method in §IV.A; common in LF QA but not equivalent to interactive free navigation.
  • domain assumption Full-reference metrics aggregated over views (mean or alternative pooling) are the appropriate primary anchors for fidelity-oriented LF coding evaluation.
    Stated in §V.A–B; NR metrics de-emphasized as less suitable anchors.
  • ad hoc to paper Equal DSCS ratings can be converted to half-vote ties, then overwritten by selective PC when available, without destroying reference-anchored meaning.
    Hybrid matrix construction in §IV.C; central to claiming finer local resolution.
  • domain assumption Standard FR image/video/multiview metrics (PSNR, SSIM family, LPIPS, VMAF, CVVDP, IV-PSNR/SSIM, etc.) form a representative baseline set for current practice.
    Anchor set in §V.A; conclusions about ‘current metrics’ depend on this coverage.
invented entities (1)
  • Hybrid DSCS+PC continuous degradation scale for the 144-stimulus LF benchmark independent evidence
    purpose: Provide fine-grained, reference-anchored subjective ground truth for objective metric evaluation and JPEG Pleno standardization.
    Constructed in this paper from DSCS ratings plus selective PC via Thurstone scaling; not a physical entity but a new measurement object the claims depend on.

pith-pipeline@v1.1.0-grok45 · 24545 in / 3729 out tokens · 27477 ms · 2026-07-12T02:05:23.256240+00:00 · methodology

0 comments
read the original abstract

Benchmarking immersive media coding solutions, especially in the standardization context, requires reliable and reproducible subjective quality assessment (QA) procedures, along with objective quality metrics that remain accurate across different distortion types. This paper presents a standardized workflow for light field QA, developed and deployed in the context of JPEG Pleno standardization activities, which integrates benchmark generation, a hybrid subjective evaluation, and objective metric analysis into a common workflow. The benchmark is designed to encompass not only traditional coding-only artifacts but also distortions that arise in processing pipelines in which light field encoding is accompanied with view synthesis and reconstruction techniques. A hybrid subjective method is proposed enabling fine-grained assessment by combining reference-anchored quality rating with targeted pairwise refinement in perceptually ambiguous regions. The reliability of subjective scores is verified using statistical consistency analyses between observers of two cohorts. Finally, a large set of objective metrics is systematically evaluated in terms of global prediction accuracy, local agreement in ambiguous quality regions, and robustness across distortion families. The results show that several metrics achieve strong agreement for coding-only stimuli, but their performance consistently drops when view synthesis distortions are included. The analysis further highlights the importance of view-pooling strategy in the design of future light field quality metrics. The work provides a reproducible and standardization-ready framework for fine-grained light field QA, while identifying key limitations of current objective metrics under emerging coding pipelines. The subjectively annotated dataset is publicly available at https://plenodb.jpeg.org/lfqa/objectivecfp.

Figures

Figures reproduced from arXiv: 2607.03494 by Carla L Pagliari, Gi-Mun Um, Ismael Seidel, Leonardo Andrade, Leonardo de Sousa Marques, Mylene C. Q. Farias, Myllena A. Prado, Saeed Mahmoudpour, Shengyang Zhao.

Figure 1
Figure 1. Figure 1: Source LF scenes used in the benchmark. (a) Representative views. (b) Spatial information (SI) and temporal/view [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the hybrid DSCS+PC subjective assessment protocol. In Stage I, each distorted stimulus is rated relative [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of subjective scores of 144 stimuli. Scores [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Scene-wise subjective quality scores with confidence intervals. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Cross-laboratory consistency of final subjective scores. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Cross-distortion residual distribution. For each metric, [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗

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

61 extracted references · 6 linked inside Pith

  1. [1]

    Light fields and computational imaging,

    M. Levoy, “Light fields and computational imaging,”Computer, vol. 39, no. 8, pp. 46–55, 2006

    2006

  2. [2]

    The Plenoptic Function and the Elements of Early Vision,

    E. H. Adelson and J. R. Bergen, “The Plenoptic Function and the Elements of Early Vision,”M. Landy and J. A. Movshon, (eds)Compu- tational Models of Visual Processing, 1991

    1991

  3. [3]

    Light field rendering,

    M. Levoy and P. Hanrahan, “Light field rendering,” inProceedings of the 23rd Annual Conference on Computer Graphics and Interactive Techniques. New York, NY , USA: ACM, 1996, pp. 31–42

    1996

  4. [4]

    Information technology — Plenoptic image coding system (JPEG Pleno) — Part 2: light field coding,

    “Information technology — Plenoptic image coding system (JPEG Pleno) — Part 2: light field coding,” ISO/IEC 21794-2:2021, Apr

    2021

  5. [5]

    Available: https://www.iso.org/standard/74532.html

    [Online]. Available: https://www.iso.org/standard/74532.html

  6. [6]

    Bench- marking subjective quality assessment methodologies for light field compression,

    S. Mahmoudpour, M. A. Prado, S. Zhao, C. L. Pagliari, J. Prazeres, M. C. Farias, A. M. Pinheiro, A. Munteanu, and P. Schelkens, “Bench- marking subjective quality assessment methodologies for light field compression,”IEEE Transactions on Broadcasting, 2026

    2026

  7. [7]

    A survey on visual quality assessment methods for light fields,

    S. Alamgeer and M. C. Farias, “A survey on visual quality assessment methods for light fields,”Signal Processing: Image Communication, vol. 110, p. 116873, 2023

    2023

  8. [8]

    On the performance of objective quality metrics for lightfields,

    S. Mahmoudpour and P. Schelkens, “On the performance of objective quality metrics for lightfields,”Signal Processing: Image Communica- tion, vol. 93, p. 116179, 2021

    2021

  9. [9]

    Quality assessment of compression solutions for ICIP 2017 grand challenge on light field image coding,

    I. Viola and T. Ebrahimi, “Quality assessment of compression solutions for ICIP 2017 grand challenge on light field image coding,” in2018 IEEE International Conference on Multimedia & Expo Workshops (ICMEW). IEEE, 2018, pp. 1–6

    2017

  10. [10]

    Datasets for the quality assessment of light field imaging: Comparison and future directions,

    E. Shafiee and M. G. Martini, “Datasets for the quality assessment of light field imaging: Comparison and future directions,”IEEE Access, vol. 11, pp. 15 014–15 029, 2023

    2023

  11. [11]

    Methodologies for the subjective assessment of the quality of television images,

    ITU-R, “Methodologies for the subjective assessment of the quality of television images,” International Telecommunication Union, Geneva, CH, Recommendation ITU-R BT.500-15, 2023. [Online]. Available: https://www.itu.int/rec/R-REC-BT.500-15-202305-I/en

    2023

  12. [12]

    Subjective video quality assessment methods for multimedia applications,

    ITU-T, “Subjective video quality assessment methods for multimedia applications,” International Telecommunication Union, Geneva, CH, Recommendation ITU-T P.910, 10 2023. [Online]. Available: https: //www.itu.int/rec/T-REC-P.910-202310-I/en

    2023

  13. [13]

    Subjective quality assessment: a study on the grading scales: illustrations for stereoscopic and 2d video content,

    B. Rania, “Subjective quality assessment: a study on the grading scales: illustrations for stereoscopic and 2d video content,” Ph.D. dissertation, Institut National des T ´el´ecommunications, 2018

    2018

  14. [14]

    Active sampling for pairwise comparisons via approximate message passing and information gain maximization,

    A. Mikhailiuk, C. Wilmot, M. Perez-Ortiz, D. Yue, and R. K. Mantiuk, “Active sampling for pairwise comparisons via approximate message passing and information gain maximization,” in2020 25th International Conference on Pattern Recognition (ICPR). IEEE, 2021, pp. 2559– 2566

    2021

  15. [15]

    Estimation from pairwise comparisons: Sharp minimax bounds with topology dependence,

    N. B. Shah, S. Balakrishnan, J. Bradley, A. Parekh, K. Ramchandran, and M. J. Wainwright, “Estimation from pairwise comparisons: Sharp minimax bounds with topology dependence,”Journal of Machine Learn- ing Research, vol. 17, no. 58, pp. 1–47, 2016. VOL. XX, NO. X, JULY 2026 13

    2016

  16. [16]

    Comparison of subjective methodologies for local perception of distortion in videos and impact on objective metrics resolving power,

    A. Pastor, P. Le Callet, V . Baroncini, Z. Li, and C. Bampis, “Comparison of subjective methodologies for local perception of distortion in videos and impact on objective metrics resolving power,” inElectronic Imaging, vol. 36, no. 11. Society for Imaging Science and Technology, 2024, pp. 1–6

    2024

  17. [17]

    H. A. David,The Method of Paired Comparisons. London, 1963, vol. 12

    1963

  18. [18]

    From pairwise comparisons and rating to a unified quality scale,

    M. Perez-Ortiz, R. K. Mantiuk, P. Knuchel, and P. Szwirowski, “From pairwise comparisons and rating to a unified quality scale,”IEEE Transactions on Image Processing, vol. 29, pp. 1139–1151, 2019

    2019

  19. [19]

    ”discriminability- experimental cost

    A. Pastor, Z. Li, C. Bampis, and P. Le Callet, “”discriminability- experimental cost” tradeoff in subjective video quality assessment of codec: Dcr with evp rating scale versus acr-hr,”arXiv preprint arXiv:2309.06227, 2023

    Pith/arXiv arXiv 2023

  20. [20]

    Measurement of visual impairment scales for digital video,

    A. B. Watson and L. Kreslake, “Measurement of visual impairment scales for digital video,” inHuman Vision and Electronic Imaging VI, vol. 4299. SPIE, 2001, pp. 79–89

    2001

  21. [21]

    Subjective image quality assessment with boosted triplet comparisons,

    H. Men, H. Lin, M. Jenadeleh, and D. Saupe, “Subjective image quality assessment with boosted triplet comparisons,”IEEE Access, vol. 9, pp. 138 939–138 975, 2021

    2021

  22. [22]

    Strategy for boosting pair comparison and improving quality assessment accu- racy,

    S. Ling, J. Li, A. F. Perrin, Z. Li, L. Krasula, and P. L. Callet, “Strategy for boosting pair comparison and improving quality assessment accu- racy,”arXiv preprint arXiv:2010.00370, 2020

    Pith/arXiv arXiv 2010

  23. [23]

    A new framework for interactive quality assessment with application to light field coding,

    I. Viola and T. Ebrahimi, “A new framework for interactive quality assessment with application to light field coding,” inApplications of Digital Image Processing XL, vol. 10396. SPIE, 2017, pp. 282–298

    2017

  24. [24]

    Subjective evaluation of low distortion coded light fields with view synthesis,

    D. Saraiva, J. Prazeres, M. Pereira, and A. M. Pinheiro, “Subjective evaluation of low distortion coded light fields with view synthesis,”arXiv preprint arXiv:2509.14761, 2025

    arXiv 2025

  25. [25]

    A hybrid subjective quality assessment framework for light field coding,

    S. Mahmoudpour, M. C. Q. Farias, and S. Zhao, “A hybrid subjective quality assessment framework for light field coding,” inProceedings of the 18th International Conference on Quality of Multimedia Experience (QoMEX), 2026

    2026

  26. [26]

    Perceptual video quality assessment: A survey,

    X. Min, H. Duan, W. Sun, Y . Zhu, and G. Zhai, “Perceptual video quality assessment: A survey,”Science China Information Sciences, vol. 67, no. 11, p. 211301, 2024

    2024

  27. [27]

    No- reference image and video quality assessment: a classification and review of recent approaches,

    M. Shahid, A. Rossholm, B. L ¨ovstr¨om, and H.-J. Zepernick, “No- reference image and video quality assessment: a classification and review of recent approaches,”EURASIP Journal on image and Video Processing, vol. 2014, no. 1, p. 40, 2014

    2014

  28. [28]

    Video quality assessment: A comprehensive survey,

    Q. Zheng, Y . Fan, L. Huang, T. Zhu, J. Liu, Z. Hao, S. Xing, C.-J. Chen, X. Min, A. C. Boviket al., “Video quality assessment: A comprehensive survey,”arXiv preprint arXiv:2412.04508, 2024

    Pith/arXiv arXiv 2024

  29. [29]

    Perceptual evaluation of light field image,

    L. Shi, S. Zhao, W. Zhou, and Z. Chen, “Perceptual evaluation of light field image,” inIEEE International Conference on Image Processing (ICIP), 2018

    2018

  30. [30]

    Re- liability of the most common objective metrics for light field quality assessment,

    H. Amirpour, A. M. G. Pinheiro, M. Pereira, and C. Guillemot, “Re- liability of the most common objective metrics for light field quality assessment,” inICASSP, 2019

    2019

  31. [31]

    Towards a quality metric for dense light fields,

    V . Kiran Adhikarla, M. Vinkler, D. Sumin, R. K. Mantiuk, K. Myszkowski, H.-P. Seidel, and P. Didyk, “Towards a quality metric for dense light fields,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR), July 2017

    2017

  32. [32]

    A light field image quality assessment model based on symmetry and depth features,

    Y . Tian, H. Zeng, J. Hou, J. Chen, and J. Zhu, “A light field image quality assessment model based on symmetry and depth features,”IEEE Transactions on Computational Imaging, vol. 6, pp. 526–538, 2020

    2020

  33. [33]

    Light field image quality assessment via the light field coherence,

    Y . Tian, H. Zeng, J. Hou, and J. Chen, “Light field image quality assessment via the light field coherence,”IEEE Transactions on Image Processing, vol. 29, pp. 7945–7956, 2020

    2020

  34. [34]

    Full reference light field image quality evaluation based on angular-spatial characteristic,

    C. Meng, P. An, X. Huang, C. Yang, and D. Liu, “Full reference light field image quality evaluation based on angular-spatial characteristic,” IEEE Signal Processing Letters, vol. 27, pp. 525–529, 2020

    2020

  35. [35]

    Light field image quality as- sessment using contourlet transform,

    H. Huang, H. Zeng, J. Hou, and J. Chen, “Light field image quality as- sessment using contourlet transform,” inIEEE International Conference on Multimedia and Expo Workshops (ICMEW), 2021

    2021

  36. [36]

    A metric for light field reconstruction, compression, and display quality evaluation,

    X. Min, J. Zhou, G. Zhai, P. L. Calletet al., “A metric for light field reconstruction, compression, and display quality evaluation,”IEEE Transactions on Image Processing, vol. 29, pp. 8188–8201, 2020

    2020

  37. [37]

    Light field image quality assessment using natural scene statistics and texture degradation,

    J. Ma, X. Zhang, C. Jin, P. An, and G. Xu, “Light field image quality assessment using natural scene statistics and texture degradation,”IEEE Transactions on Circuits and Systems for Video Technology, 2023

    2023

  38. [38]

    Eddmf: An efficient deep discrepancy measuring framework for full-reference light field image quality assessment,

    Z. Zhang, S. Tian, W. Zou, L. Morin, and L. Zhang, “Eddmf: An efficient deep discrepancy measuring framework for full-reference light field image quality assessment,”IEEE Transactions on Image Processing, vol. 32, pp. 6426–6440, 2023

    2023

  39. [39]

    Multi-dimension attention network for full-reference light field image quality assessment,

    Y . Zhang, J. Jiang, D. Liu, X. Zhouet al., “Multi-dimension attention network for full-reference light field image quality assessment,”IEEE Transactions on Image Processing, 2025

    2025

  40. [40]

    3d gaussian splatting for real-time radiance field rendering

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, G. Drettakiset al., “3d gaussian splatting for real-time radiance field rendering.”ACM Trans. Graph., vol. 42, no. 4, pp. 139–1, 2023

    2023

  41. [41]

    https://gitlab.com/wg1/ jpeg-pleno-refsw

    JPEG Pleno Reference Software. https://gitlab.com/wg1/ jpeg-pleno-refsw. Accessed: 2025-08-27

    2025

  42. [42]

    https://x265.org

    x265 HEVC Encoder / H.265 Video Codec. https://x265.org. Accessed: 2025-08-27

    2025

  43. [43]

    JPEG Pleno light field coding common test conditions v3.4,

    F. Pereira, C. Pagliari, E. A. B. da Silva, I. Tabus, H. Amirpour, M. Bernardo, and A. Pinheiro, “JPEG Pleno light field coding common test conditions v3.4,” ISO/IEC JTC 1/SC29/WG1, Covilh ˜a, Portugal, Document ISO/IEC JTC 1/SC29/WG1 N100556, July 2023. [Online]. Available: https: //ds.jpeg.org/documents/jpegpleno/wg1n100556-100-PCQ-Common Test Condition...

    2023

  44. [44]

    Seidel, L

    I. Seidel, L. de Sousa Marques, A. F. S. Fernandes et al.(2026) JPEG Pleno light field coding toolkit (LFC toolkit). BSD 3-Clause License. [Online]. Available: https://gitlab.com/eclufsc/light-field-coding/jpeg-pleno-ctc-tools

    2026

  45. [45]

    Real-time intermediate flow estimation for video frame interpolation,

    Z. Huang, T. Zhang, W. Heng, B. Shi, and S. Zhou, “Real-time intermediate flow estimation for video frame interpolation,” inEuropean conference on computer vision. Springer, 2022, pp. 624–642

    2022

  46. [46]

    Revisiting adaptive convolutions for video frame interpolation,

    S. Niklaus, L. Mai, and O. Wang, “Revisiting adaptive convolutions for video frame interpolation,” inIEEE Winter Conference on Applications of Computer Vision, 2021

    2021

  47. [47]

    A practical guide and soft- ware for analysing pairwise comparison experiments,

    M. Perez-Ortiz and R. K. Mantiuk, “A practical guide and soft- ware for analysing pairwise comparison experiments,”arXiv preprint arXiv:1712.03686, 2017

    Pith/arXiv arXiv 2017

  48. [48]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,”IEEE transactions on image processing, vol. 13, no. 4, pp. 600–612, 2004

    2004

  49. [49]

    Multiscale structural similarity for image quality assessment,

    Z. Wang, E. P. Simoncelli, and A. C. Bovik, “Multiscale structural similarity for image quality assessment,” inThe Thrity-Seventh Asilomar Conference on Signals, Systems Computers, 2003, vol. 2, Nov 2003, pp. 1398–1402 V ol.2

    2003

  50. [50]

    Information content weighting for perceptual image quality assessment,

    Z. Wang and Q. Li, “Information content weighting for perceptual image quality assessment,”IEEE Transactions on Image Processing, vol. 20, no. 5, pp. 1185–1198, 2011

    2011

  51. [51]

    Image information and visual quality,

    H. R. Sheikh and A. C. Bovik, “Image information and visual quality,” IEEE Transactions on Image Processing, vol. 15, no. 2, pp. 430–444, Feb 2006

    2006

  52. [52]

    FSIM: A feature similarity index for image quality assessment,

    L. Zhang, L. Zhang, X. Mou, and D. Zhang, “FSIM: A feature similarity index for image quality assessment,”IEEE Transactions on Image Processing, vol. 20, no. 8, pp. 2378–2386, 2011

    2011

  53. [53]

    Gradient magnitude similarity deviation: A highly efficient perceptual image quality index,

    W. Xue, L. Zhang, X. Mou, and A. C. Bovik, “Gradient magnitude similarity deviation: A highly efficient perceptual image quality index,” IEEE Transactions on Image Processing, vol. 23, no. 2, pp. 684–695, Feb 2014

    2014

  54. [54]

    Most apparent distortion: Full-reference image quality assessment and the role of strategy,

    E. Larson and D. Chandler, “Most apparent distortion: Full-reference image quality assessment and the role of strategy,”J. Electronic Imaging, vol. 19, p. 011006, 01 2010

    2010

  55. [55]

    Perceptually optimized image rendering,

    V . Laparra, A. Berardino, J. Ball ´e, and E. P. Simoncelli, “Perceptually optimized image rendering,”Journal of the Optical Society of America A, vol. 34, no. 9, pp. 1511–1525, 2017

    2017

  56. [56]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isola, A. A. Efros, E. Shechtman, and O. Wang, “The unreasonable effectiveness of deep features as a perceptual metric,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2018, pp. 586–595

    2018

  57. [57]

    Image quality assess- ment: Unifying structure and texture similarity,

    K. Ding, K. Ma, S. Wang, and E. P. Simoncelli, “Image quality assess- ment: Unifying structure and texture similarity,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 5, pp. 2567– 2581, 2022

    2022

  58. [58]

    Shift-tolerant perceptual similarity metric,

    A. Ghildyal and F. Liu, “Shift-tolerant perceptual similarity metric,” in European Conference on Computer Vision, 2022

    2022

  59. [59]

    Deepdc: Deep distance correlation as a perceptual image quality evaluator,

    H. Zhu, B. Chen, L. Zhu, S. Wang, and W. Lin, “Deepdc: Deep distance correlation as a perceptual image quality evaluator,” CoRR, vol. abs/2211.04927v2, 2023. [Online]. Available: https: //arxiv.org/pdf/2211.04927v2.pdf

    Pith/arXiv arXiv 2023

  60. [60]

    VMAF: The Video Multi-Method Assessment Fusion tool,

    Netflix Inc., “VMAF: The Video Multi-Method Assessment Fusion tool,” GitHub repository, 2016, https://github.com/Netflix/vmaf

    2016

  61. [61]

    Colorvideovdp: A visual difference predictor for image, video and display distortions,

    R. K. Mantiuk, P. Hanji, M. Ashraf, Y . Asano, and A. Chapiro, “Colorvideovdp: A visual difference predictor for image, video and display distortions,”arXiv preprint arXiv:2401.11485, 2024

    Pith/arXiv arXiv 2024