Pith. sign in

REVIEW 3 major objections 5 minor 46 references

ColorNet -- Estimating Colorfulness in Natural Images

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A fine-tuned convolutional network predicts perceived colorfulness more accurately than classical hand-crafted metrics, with Pearson correlation 0.937 on a 180-image combined subjective dataset.

desk verdict First learning-based colorfulness metric, and the dataset-alignment trick is genuinely useful; the weak point is the ground truth, built from 3 expert viewers and a linear fit over 12 images per source. read the letter →

arxiv 1908.08505 v1 pith:2EYYKAJL submitted 2019-08-22 cs.MM cs.GRcs.LGeess.IV

Emin Zerman , Aakanksha Rana , Aljosa Smolic This is my paper · ORCID
classification cs.MMcs.GRcs.LGeess.IV
keywords colourfulnessCNNcolormetricdeeplearningimagequalityassessmentsubjectivetestalignmentpsychometricscalingperception
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

This paper sets out to show that a convolutional neural network can estimate the perceived colorfulness of a natural image more accurately than hand-crafted color-statistics formulas. Because no large annotated colorfulness dataset existed, the authors combine two public subjective databases by running a small anchoring experiment on a 24-image common subset and linearly remapping both databases onto one rating scale, yielding 180 images with aligned human scores. They then fine-tune a pretrained CNN as a feature extractor and attach a small rating network to predict colorfulness. In 10-fold cross-validation, the VGG-based ColorNet variant reaches Pearson correlation 0.937 and Spearman correlation 0.921 with human ratings, outperforming the four classical metrics tested. If correct, this establishes deep learning as a viable route to colorfulness estimation, which matters for tone-mapping, aesthetic analysis, and image-quality assessment.

What carries the argument

The load-bearing component is the dataset-alignment procedure plus the two-stage CNN. For alignment, 12 images are selected from each of two public databases to cover the full colorfulness range, and three expert viewers compare them pairwise using an adaptive square design, scaled by Thurstone Case V psychometric scaling to a rating scale of 1 to 9. A linear mapping $y = ax + b$ is fitted between the anchor scores and each database's original scores, with $a_{\mathrm{EPFL}}=0.8748$, $b_{\mathrm{EPFL}}=1.4350$ and $a_{\mathrm{UCL}}=1.1388$, $b_{\mathrm{UCL}}=6.8759$, producing the 180-image Combined dataset. The model itself is the feature network followed by the rating network, trained end-to-end with the L1 loss, and the best configuration is ColorNet-VGG.

What would settle it

Conduct a fresh subjective experiment on a held-out set of natural images, or re-run the Anchor experiment with many more non-expert viewers and a rank-based alignment, then compare ColorNet-VGG's predictions against the newly collected scores and against the classical opponent-color formula; if ColorNet-VGG no longer shows a higher Pearson or Spearman correlation with the fresh ratings, the claimed advantage is an artefact of the alignment. Alternatively, train and test ColorNet-VGG separately on each original database without the linear alignment; if the advantage over classical metrics disappears, the claim depends on the combined labels.

Watch

Extended reading notes

Core claim

ColorNet is a two-stage model: a feature network taken from an ImageNet-pretrained CNN (VGG16, ResNet18, or MobileNet) with its classification head removed, followed by a rating network consisting of a dropout layer, a 10-unit fully connected layer, a ReLU, and a single output unit. It is trained end-to-end with an L1 loss on 180 images whose subjective colorfulness scores were aligned by a linear mapping through an anchoring experiment. The paper's central result is that ColorNet-VGG predicts these aligned human ratings with Pearson correlation 0.937 and Spearman correlation 0.921, compared with 0.841/0.884 for the classical opponent-color formula, 0.895/0.896 and 0.312/0.415 for the two logarithmic CQE versions, and 0.843/0.834 for the saturation-based model. The authors interpret this as evidence that learned high-level features capture colorfulness information that low-level color statistics miss, and as a first deep-learning step in color-perception modeling.

Load-bearing premise

The linear transformation $y = ax + b$ that aligns each public database to the Anchor scores is fitted on only 12 selected images per database judged by 3 expert viewers, and the paper assumes this fit extends to all 180 images; if the anchor ratings are unreliable or the relationship is not linear, the combined ground-truth scores become miscalibrated and the reported correlations cannot be trusted.

Editorial extensions

If this is right

  • Colorfulness can be treated as a perceptual attribute learnable from a small, aligned dataset, rather than as a fixed statistical formula.
  • The anchor-alignment method offers a template for merging subjective datasets collected under different psychophysical protocols (rating versus pairwise comparison) into one training set for a deep metric.
  • Fine-tuned ImageNet features already encode enough color information to outperform specially designed opponent-color statistics, suggesting that other low-level perceptual attributes, such as contrast or naturalness, may also be learnable this way.
  • A dataset of only 180 images is sufficient for a first learning-based colorfulness model, providing a starting point for scaling to larger and more diverse color-perception data.

Reading between the lines

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

  • The reported advantage may be partly tied to the specific linear alignment and the very small anchor set; a rank-based or probabilistic alignment would likely change absolute scores but might preserve the relative ordering of models.
  • Because the 10-fold protocol leaves only 18 test images per fold, the performance difference between ColorNet-VGG and the classical formula should be re-examined with confidence intervals or significance tests, which the paper does not report.
  • The same two-stage transfer-learning recipe could be applied to other subjective perceptual attributes, such as brightness, naturalness, or contrast, using anchor-aligned databases, and might yield similar gains over hand-crafted metrics.
  • ColorNet-Mobile's lower performance suggests that for this perceptual regression task, depth and residual connections matter more than parameter efficiency; depth-wise separable convolutions may lose color-sensitive gradient information.
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

3 major / 5 minor

Summary. The paper proposes ColorNet, the first deep learning-based colorfulness estimation metric for natural images. Because no sufficiently large colorfulness dataset exists, the authors combine two existing subjective databases (EPFL and UCL) using a new 'Anchor' subjective test on 24 selected images rated by three expert viewers. They fit linear mappings y = ax + b between the Anchor scores and each source database's scores, yielding a 180-image combined dataset with aligned subjective scores. They then fine-tune three CNN feature extractors (VGG, ResNet, MobileNet) with a shared rating network and evaluate them using 10-fold cross-validation. Table 1 reports that ColorNet-VGG achieves Pearson correlation 0.937 and Spearman correlation 0.921 with the aligned subjective scores, outperforming Hasler, Panetta CQE1/CQE2, and Yendrikhovskij. The authors also provide qualitative examples of monotonic responses to color saturation and hue changes.

Significance. If the reported correlations are robust, this is a useful contribution: it introduces a learning-based alternative to classical colorfulness formulas, and the dataset-alignment procedure addresses a real practical bottleneck (the absence of a shared subjective scale across databases). The held-out 10-fold evaluation is an appropriate and more trustworthy design than fitting and testing on the same images, and the authors compare against four standard classical metrics. However, the central quantitative claim rests on an alignment fitted from only 24 images and three expert viewers, and the reported PCC/SROCC differences come without uncertainty quantification or significance testing. These issues must be addressed before the superiority claim over classical metrics can be considered established.

major comments (3)
  1. [Section 3, 'Combining the Datasets', Eq. (1)] The linear alignment y = ax + b is fitted on only 12 selected images per database, with Anchor scores obtained from just three expert viewers. The paper states that the relationship is linear with 'very high correlation scores' but does not report the correlations, residuals, or confidence intervals for a and b. Because the aligned scores are used both as training labels and as the ground truth for the quantitative comparison in Table 1, any miscalibration of Eq. (1) propagates directly into the claimed advantage of ColorNet-VGG. Please report alignment fit diagnostics and perform a leave-one-anchor-out or bootstrap refit of Eq. (1) to show that the PCC/SROCC rankings in Table 1 are stable under realistic variation in the anchor mapping.
  2. [Section 5, 'Quantitative Evaluation' and Table 1] The best ColorNet variant is selected by comparing VGG-, ResNet-, and MobileNet-based features on the same 10-fold splits, yet no error bars, confidence intervals, or significance tests accompany the reported PCC and SROCC values. With only 18 test images per fold, the difference between ColorNet-VGG (0.937) and CQE1 (0.895) could lie within sampling noise, especially since model selection is performed on the same evaluation data. Please provide per-fold results, bootstrap confidence intervals, or a paired significance test (e.g., Wilcoxon on per-fold correlations or Fisher z-transformation) to support the claim that ColorNet-VGG outperforms the classical methods.
  3. [Section 3, 'Selection of Images from Each Dataset'] The selected 12 images per database are described as representative because they 'cover the whole quality scale' of the source scores, but no evidence is given that they are representative under the Anchor experiment's subjective judgments. The validation in Fig. 1(e) plots Combined versus Anchor scores using the same 24 images that were used to estimate the alignment, so it is an in-sample check and cannot certify extrapolation to the remaining 156 images. Please provide an external validation (e.g., a small hold-out set of newly rated images) or explicitly discuss this limitation and its possible effect on the combined labels.
minor comments (5)
  1. [Title and header] The running header reads 'ESTIMA TING' with an internal space; this appears to be a typesetting artifact and should be corrected.
  2. [Eq. (2)] The loss function sums over k but uses the index i for the image and rating; please make the index notation consistent, e.g., sum over i of |y_i - phi^j_r(phi^j_f(X_i))|.
  3. [Section 5, 'Training and Implementation Details'] The description of the 10-fold protocol says one piece is used for validation in each iteration, which implies training uses only 8 of 10 pieces; please clarify how the validation piece is chosen and how the final averaged results relate to the validation split.
  4. [Abstract and Introduction] The phrase 'the first deep learning-based colorfulness estimation metric' should be qualified as 'to the best of our knowledge' to be precise about novelty.
  5. [Section 5, 'Qualitative Evaluation'] Figure 3 shows only a few hand-picked examples without quantitative agreement among observers; please either state that this is illustrative only or add a small multi-rater study to support the qualitative claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: ColorNet is trained and evaluated on held-out subjective colorfulness labels, and the anchor alignment is data calibration, not a self-referential prediction.

full rationale

The paper's derivation chain is an ordinary supervised-learning pipeline: (1) two existing subjective colorfulness databases are collected; (2) a third 'Anchor' experiment on 24 selected images produces scores on a common scale; (3) Eq. (1), Qhat_DB = a_DB * Q_DB + b_DB, linearly maps each source database's scores onto that scale; (4) the 180 mapped scores become training/validation/test labels; and (5) ColorNet is trained end-to-end to minimize L1 loss between its predicted rating and those labels, then evaluated in 10-fold cross-validation against held-out labels. The CNN's output is not defined in terms of the labels, and the alignment parameters a and b are not outputs of the proposed metric; they are auxiliary calibration constants used to construct ground truth. Even if the 12-image, three-viewer anchor fit is noisy or the linearity assumption is imperfect, that is a threat to label validity and result robustness, not a circular reduction: the Table 1 correlations of ColorNet are computed on test folds that were not used to train that fold's model, and the classical metrics are evaluated against the same labels without having been fitted to them. The selection of the best among three architectures on the same dataset is a model-selection risk, not a constructional circularity. The paper does not invoke a self-citation chain to force its choice, and its central claim is empirically testable against held-out human opinion scores. No step in the derivation is equivalent to its own input by definition.

Assumptions & free parameters 6 free parameters · 7 assumptions · 0 invented entities

The central claim depends on the validity of the combined subjective ground truth, which is built from fitted alignment parameters and a three-viewer anchor experiment, and on the transferability of ImageNet features to colorfulness perception. No new theoretical entities are introduced.

free parameters (6)
  • a_EPFL = 0.8748
    Slope of the linear mapping from EPFL source scores to the Anchor scale, fitted by least squares on 12 selected EPFL images.
  • b_EPFL = 1.4350
    Intercept of the linear mapping from EPFL source scores to the Anchor scale, fitted on 12 selected EPFL images.
  • a_UCL = 1.1388
    Slope of the linear mapping from UCL source scores to the Anchor scale, fitted on 12 selected UCL images.
  • b_UCL = 6.8759
    Intercept of the linear mapping from UCL source scores to the Anchor scale, fitted on 12 selected UCL images.
  • ColorNet-VGG learned weights = millions of parameters fitted to training labels
    The CNN weights are optimized on the training folds of the 180-image combined dataset; they are the model itself rather than a small set of derivation constants.
  • Training hyperparameters (learning rates, dropout, batch size, crop size) = 1e-4, 1e-3, 0.75, 4, 512
    These hyperparameters are chosen by hand and affect the final model; no sensitivity analysis is provided.
assumptions (7)
  • domain assumption The subjective colorfulness scores in the EPFL and UCL databases are valid interval-scale measurements of human colorfulness perception.
    All training and evaluation assumes these scores are the ground truth; any error in the source scores propagates through the combined dataset.
  • domain assumption The anchor scores from the 3 expert viewers are reliable enough to estimate the linear mapping between the two databases.
    Section 3 reports only three expert viewers in the anchor experiment; an unusually small panel for psychometric scaling.
  • domain assumption The relationship between anchor scores and source database scores is linear over the full quality range.
    Eq. (1) applies a linear transform to all images; if the true relationship is nonlinear outside the 24 selected images, the combined ground truth is miscalibrated.
  • domain assumption ImageNet-pretrained weights provide a useful starting point for colorfulness estimation.
    All three ColorNet variants are initialized with ImageNet classification weights, assuming transferability to a low-level perceptual color attribute.
  • domain assumption Colorfulness can be predicted from 512x512 crops of images, and global image colorfulness is captured by that spatial resolution.
    The model is trained and tested on 512x512 crops of 600x600 images, so the metric cannot see the full original image at once.
  • ad hoc to paper The 12 selected images per database are representative of the full colorfulness range.
    The selection was made manually to cover the scale; this can inflate the apparent correlation and linearity of the anchor fit.
  • standard math Thurstone Case V scaling produces an interval-scale score from pairwise comparisons.
    Used to derive UCL and anchor scores; standard psychometric assumption but not checked here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ColorNet -- Estimating Colorfulness in Natural Images." pith.science (2026). https://pith.science/paper/2EYYKAJL

@misc{pith2026190808505,
  author       = {Pith},
  title        = {Pith review of: ColorNet -- Estimating Colorfulness in Natural Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2EYYKAJL}},
  note         = {Machine review of arXiv:1908.08505}
}
read the original abstract

Measuring the colorfulness of a natural or virtual scene is critical for many applications in image processing field ranging from capturing to display. In this paper, we propose the first deep learning-based colorfulness estimation metric. For this purpose, we develop a color rating model which simultaneously learns to extracts the pertinent characteristic color features and the mapping from feature space to the ideal colorfulness scores for a variety of natural colored images. Additionally, we propose to overcome the lack of adequate annotated dataset problem by combining/aligning two publicly available colorfulness databases using the results of a new subjective test which employs a common subset of both databases. Using the obtained subjectively annotated dataset with 180 colored images, we finally demonstrate the efficacy of our proposed model over the traditional methods, both quantitatively and qualitatively.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 44 canonical work pages

  1. [1]

    It affects hu- man behavior and decision processes both in nature and in society, as color conveys pivotal information about the surroundings

    INTRODUCTION Color is a crucial factor in human visual perception. It affects hu- man behavior and decision processes both in nature and in society, as color conveys pivotal information about the surroundings. Thus, its accurate acquisition and display are necessary for multimedia, entertainment, and image processing systems. Within the imaging pipeline, ...

    work page

  2. [2]

    RELA TED WORK Several studies have attempted to understand and estimate the color- fulness in visual content. Color appearance models (CAMs) utilize some form of chroma and colorfulness estimation to estimate the local visual perception and to reproduce the colors considering the lighting conditions [6, 20, 21]. However, such estimations are valid mostly ...

    work page Pith review arXiv 1908

  3. [3]

    SUBJECTIVE DA TA In this study, we use the subjective colorfulness scores collected from participants for two different colorfulness databases: EPFL Dataset [5] with 84 images and UCL Dataset1 [19] with 96 images2. 1University College London Colourfulness Dataset - http:// reality.cs.ucl.ac.uk/projects/image-colourfulness/ image-colourfulness.html 2Althou...

    work page

  4. [4]

    PROPOSED COLOR RA TING MODEL The ColorNet architecture is illustrated in Fig. 2. The proposed model has two major building blocks, 1) the feature network Φf and 2) the rating network Φr. We base our feature network Φf on the state-of-the-art deep learning models, namely, VGG [22], ResNet [23] and MobileNet [24]. For all these models, we specifi- cally remo...

    work page

  5. [5]

    In our paper, we removed the three fully connected layers of the VGG16 architecture and simply fed the resulting feature vector into the proposed rating network

    ColorNet-VGG has a 13 convolutional layered feature net- work Φvgg f adopted from the VGG16 [22] architecture, with small 3× 3 convolutions, resulting in a feature vector of di- mension 512×7×7. In our paper, we removed the three fully connected layers of the VGG16 architecture and simply fed the resulting feature vector into the proposed rating network. ...

    work page

  6. [6]

    Similar to VGG [22], the ResNet architecture has small3× 3 convolutions, however with an additional concept of residual learning applied to every few stacked layers

    ColorNet-ResNet consists of an 18 layer deep residual feature network Φresnet f adopted from the ResNet architecture [23]. Similar to VGG [22], the ResNet architecture has small3× 3 convolutions, however with an additional concept of residual learning applied to every few stacked layers. In our paper, we removed the last fully connected layer of 1000 chan...

    work page

  7. [7]

    MobileNet has been a widely adopted network for many mobile and embedded applications

    ColorNet-Mobile consists of a 28 convolution layers fea- ture network Φmobile f adopted from the MobileNet architec- ture [24], which comprises of both depth-wise and point-wise convolutions. MobileNet has been a widely adopted network for many mobile and embedded applications. In this paper, we removed its last fully connected and soft-max layer to obtai...

    work page

  8. [8]

    For training, the input image size is fixed at600× 600, and random crops of size 512× 512 are applied to the image

    RESULTS Training and Implementation Details: We split the dataset of 180 images into training, validation and test sets in an 80%, 10% and 10% setting. For training, the input image size is fixed at600× 600, and random crops of size 512× 512 are applied to the image. Ad- ditional data augmentation techniques such as rotation and flipping are applied to scal...

    work page

Show all 46 references
  1. [9]

    CONCLUSION In this study, we propose a CNN based model for the estimation of colorfulness ratings. To prepare a well-annotated colored image dataset, we combine two colorfulness databases with subjective user scores, using the results of an anchor subjective experiment with a ...

  2. [10]

    Automated colour grading using colour distribution transfer,

    F. Piti ´e, A. C. Kokaram, and R. Dahyot, “Automated colour grading using colour distribution transfer,” Comput. Vis. Image Underst. , vol. 107, no. 1-2, pp. 123–137, Jul. 2007

    2007

  3. [11]

    Display adaptive tone map- ping,

    R. Mantiuk, S. Daly, and L. Kerofsky, “Display adaptive tone map- ping,” ACM Transactions on Graphics, Aug 2008

    2008

  4. [12]

    A gamut-mapping framework for color- accurate reproduction of HDR images,

    E. Sikudov ´a, T. Pouli, A. Artusi, A. O. Aky ¨uz, F. Banterle, Z. M. Ma- zlumoglu, and E. Reinhard, “A gamut-mapping framework for color- accurate reproduction of HDR images,” IEEE Computer Graphics and Applications, vol. 36, no. 4, pp. 78–90, 2016

    2016

  5. [13]

    Color correction for tone mapping,

    R. Mantiuk, R. Mantiuk, A. Tomaszewska, and W. Heidrich, “Color correction for tone mapping,” Computer Graphics F orum , vol. 28, no. 2, pp. 193–202, 2009

    2009

  6. [14]

    Measuring colorfulness in natural im- ages,

    D. Hasler and S. E. S ¨usstrunk, “Measuring colorfulness in natural im- ages,” inSPIE Electronic Imaging, Human Vision and Electronic Imag- ing VIII, 2003, pp. 87–96

    2003

  7. [15]

    M. D. Fairchild, Color Appearance Models. John Wiley & Sons, 2013

    2013

  8. [16]

    Calibrated image appearance reproduction,

    E. Reinhard, T. Pouli, T. Kunkel, B. Long, A. Ballestad, and G. Damberg, “Calibrated image appearance reproduction,”ACM Trans- actions on Graphics (TOG), vol. 31, no. 6, 2012

    2012

  9. [17]

    Learning-based tone mapping operator for efficient image matching,

    A. Rana, G. Valenzise, and F. Dufaux, “Learning-based tone mapping operator for efficient image matching,”IEEE Transactions on Multime- dia, vol. 21, no. 1, pp. 256–268, 2019

    2019

  10. [18]

    Learning-based Adaptive Tone Mapping for Keypoint De- tection,

    ——, “Learning-based Adaptive Tone Mapping for Keypoint De- tection,” in IEEE International Conference on Multimedia & Expo (ICME’2017), Hong Kong, China, Jul. 2017

    2017

  11. [19]

    Learning-Based Tone Mapping Operator for Image Matching,

    ——, “Learning-Based Tone Mapping Operator for Image Matching,” in IEEE International Conference on Image Processing (ICIP’2017) . Beijing, China: IEEE, 2017

    2017

  12. [20]

    Studying aesthetics in photo- graphic images using a computational approach,

    R. Datta, D. Joshi, J. Li, and J. Z. Wang, “Studying aesthetics in photo- graphic images using a computational approach,” in European Confer- ence on Computer Vision. Springer, 2006, pp. 288–301

    2006

  13. [21]

    Automated aesthetic analy- sis of photographic images,

    T. O. Aydın, A. Smolic, and M. Gross, “Automated aesthetic analy- sis of photographic images,” IEEE Transactions on Visualization and Computer Graphics, vol. 21, no. 1, pp. 31–42, 2015

    2015

  14. [22]

    Improving color reproduction ac- curacy on cameras,

    H. Can Karaimer and M. S. Brown, “Improving color reproduction ac- curacy on cameras,” in The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2018

    2018

  15. [23]

    Analysis of public image and video databases for qual- ity assessment,

    S. Winkler, “Analysis of public image and video databases for qual- ity assessment,” IEEE Journal of Selected Topics in Signal Processing , vol. 6, no. 6, pp. 616–625, 2012

    2012

  16. [24]

    Selecting scenes for 2D and 3D subjective video quality tests,

    M. H. Pinson, M. Barkowsky, and P. Le Callet, “Selecting scenes for 2D and 3D subjective video quality tests,”EURASIP Journal on Image and Video Processing, vol. 2013, no. 1, 2013

    2013

  17. [25]

    Optimizing color reproduction of natural images,

    S. Yendrikhovskij, F. J. Blommaert, and H. de Ridder, “Optimizing color reproduction of natural images,” in Color and Imaging Confer- ence. Society for Imaging Science and Technology, 1998, pp. 140– 145

    1998

  18. [26]

    No reference color image contrast and quality measures,

    K. Panetta, C. Gao, and S. Agaian, “No reference color image contrast and quality measures,” IEEE Transactions on Consumer Electronics , vol. 59, no. 3, pp. 643–651, 2013

    2013

  19. [27]

    Towards generating ambisonics using audio-visual cue for virtual reality,

    A. Rana, C. Ozcinar, and A. Smolic, “Towards generating ambisonics using audio-visual cue for virtual reality,” in 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2019

    2019

  20. [28]

    A study of image colourful- ness,

    C. Amati, N. J. Mitra, and T. Weyrich, “A study of image colourful- ness,” in Proceedings of the Workshop on Computational Aesthetics . ACM, 2014, pp. 23–31

    2014

  21. [29]

    R. W. G. Hunt, The Reproduction of Colour, 5th ed. Wiley, 1995

    1995

  22. [30]

    Revision of the chroma and hue scales of a nonlinear color-appearance model,

    Y . Nayatani, “Revision of the chroma and hue scales of a nonlinear color-appearance model,”Color Research & Application, vol. 20, no. 3, pp. 143–155, 1995

    1995

  23. [31]

    Very deep convolutional networks for large-scale image recognition,

    K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in International Conference on Learn- ing Representations, 2015

    2015

  24. [32]

    Deep residual learning for im- age recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for im- age recognition,” IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016

    2016

  25. [33]

    MobileNets: Efficient convo- lutional neural networks for mobile vision applications

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “MobileNets: Efficient convo- lutional neural networks for mobile vision applications.” CoRR, 2017

    2017

  26. [34]

    Objective evalua- tion of naturalness, contrast, and colorfulness of tone-mapped images,

    L. Krasula, K. Fliegel, P. Le Callet, and M. Kl ´ıma, “Objective evalua- tion of naturalness, contrast, and colorfulness of tone-mapped images,” in SPIE Optical Engineering + Applications, Applications of Digital Image Processing XXXVII. International Society for Optics and Ph...

    2014

  27. [35]

    Evaluation of HDR tone mapping methods using essential perceptual attributes,

    M. ˇCad´ık, M. Wimmer, L. Neumann, and A. Artusi, “Evaluation of HDR tone mapping methods using essential perceptual attributes,” Computers & Graphics, vol. 32, no. 3, pp. 330–349, 2008

    2008

  28. [36]

    NIMA: Neural image assessment,

    H. Talebi and P. Milanfar, “NIMA: Neural image assessment,” IEEE Transactions on Image Processing, vol. 27, pp. 3998–4011, 2018

    2018

  29. [37]

    Convolutional neural net- works for no-reference image quality assessment,

    L. Kang, P. Ye, Y . Li, and D. Doermann, “Convolutional neural net- works for no-reference image quality assessment,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR) , June 2014

    2014

  30. [38]

    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,” stat.AP - arXiv:1712.03686, 2017

    2017 arXiv

  31. [39]

    P. G. Engeldrum, Psychometric Scaling: A Toolkit for Imaging Systems Development. Imcotek press, 2000

    2000

  32. [40]

    Aligning subjective tests using a low cost common set,

    Y . Pitrey, U. Engelke, M. Barkowsky, R. P ´epion, and P. Le Callet, “Aligning subjective tests using a low cost common set,” in Euro ITV, 2011

    2011

  33. [41]

    Boosting paired comparison methodology in measuring visual discomfort of 3DTV: performances of three different designs,

    J. Li, M. Barkowsky, and P. Le Callet, “Boosting paired comparison methodology in measuring visual discomfort of 3DTV: performances of three different designs,” in SPIE Electronic Imaging, Stereoscopic Displays and Applications XXIV. International Society for Optics and Photon...

    2013

  34. [42]

    Subjective assessment methods for 3D video quality,

    ITU-R, “Subjective assessment methods for 3D video quality,” ITU-R Recommendation P.915, Mar 2016

    2016

  35. [43]

    What’s new in Psychtoolbox-3,

    M. Kleiner, D. Brainard, D. Pelli, A. Ingling, R. Murray, C. Broussard et al. , “What’s new in Psychtoolbox-3,” Perception, vol. 36, no. 14, 2007

    2007

  36. [44]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” CoRR, 2014

    2014

  37. [45]

    Automatic differen- tiation in PyTorch,

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, “Automatic differen- tiation in PyTorch,” inNIPS-W, 2017

    2017

  38. [46]

    Ima- geNet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Ima- geNet: A large-scale hierarchical image database,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2009

    2009

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.