pith. sign in

arxiv: 2606.01638 · v1 · pith:X5YUA2PTnew · submitted 2026-06-01 · 💻 cs.CV

CanonCGT: Reference-Based Color Grading via Canonical Pivot Representation

Jinwon Ko , Keunsoo Ko , Chang-Su Kim This is my paper

Pith reviewed 2026-06-28 15:29 UTC · model grok-4.3

classification 💻 cs.CV
keywords reference-based color gradingcanonical pivottone mappingphotorealistic image processingstyle transferdual-phase training
0
0 comments X

The pith

CanonCGT maps images through a style-neutral canonical pivot for stable reference-based color grading.

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

The paper presents CanonCGT as a two-stage method that first converts an input image into a canonical pivot by stripping away its intrinsic tonal bias, then applies the tonal mood from a reference image. This structure targets the instability seen in prior reference-based grading techniques, where mappings either over-shift tones or fail to keep colors consistent. A dual-phase training process combines supervised learning on preset pairs with self-supervised refinement on unpaired photos. If the approach holds, graded outputs should retain scene structure and color harmony while matching the reference more reliably across varied inputs.

Core claim

CanonCGT is a two-stage framework built on a canonical pivot -- a style-neutral intermediate representation for stable color mapping. The first stage canonicalizes the input by removing intrinsic tonal bias, and the second color-grades it to match the reference style, trained via DP-CGT that mixes supervised preset learning with self-supervised refinement on unpaired photographs.

What carries the argument

The canonical pivot, a style-neutral intermediate representation that removes tonal bias from the input before applying reference style.

If this is right

  • Tone mappings remain stable across diverse datasets without over-shifting.
  • Color harmony and scene structure are preserved while matching reference mood.
  • Results surpass prior methods in both stability and visual fidelity.
  • Self-supervised refinement on unpaired photos extends applicability beyond paired data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The separation into canonicalization and grading stages could support editing pipelines that reuse the same pivot for multiple references.
  • Temporal consistency in video might follow if the canonical pivot is computed frame-by-frame with additional smoothing.
  • The dual-phase training pattern may transfer to other unpaired image-to-image tasks that require style neutrality.

Load-bearing premise

The canonical pivot acts as a style-neutral intermediate that produces stable color mappings without over-shifting or inconsistent retention.

What would settle it

A set of input-reference pairs where the canonical pivot still produces visible over-shifting or color inconsistency compared with direct mapping methods would falsify the stability claim.

Figures

Figures reproduced from arXiv: 2606.01638 by Chang-Su Kim, Jinwon Ko, Keunsoo Ko.

Figure 1
Figure 1. Figure 1: Our reference-based color grading results. Each input image is color-graded to match the tonal mood and lighting of its reference, yielding photorealistic results that preserve color harmony and scene structure. Abstract Reference-based color grading aims to reproduce the tonal mood and lighting of a reference while preserving color harmony and scene structure. Existing photorealistic and filter-based meth… view at source ↗
Figure 2
Figure 2. Figure 2: Color grading results of existing methods and the proposed CanonCGT. (a) Comparison with recent photorealistic style transfer [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the CanonCGT framework. The grade extractor derives grading vectors [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The architecture of the canonicalizer and grader. The network, referred to as the conditioned LUT generator, takes an input image [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The architecture of the FFN-FiLM block. modulation: α = W1g, β = W2g, (2) F (film) l = α ⊙ σ(F (sa) l W3) + β, (3) F (out) l = F (film) l W4 + F (sa) l , (4) where σ(·) denotes GELU [13]. The grading vector g ∈ R C generates scale and shift parameters α, β ∈ R C ′ , while W1, W2 ∈ R C ′×C , W3 ∈ R C×C ′ , and W4 ∈ R C ′×C are learnable projections. Unlike the original FiLM [35] that conditions on language … view at source ↗
Figure 6
Figure 6. Figure 6: Self-supervised learning phase: Two crops [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of color grading results. CanonCGT produces natural and tonally balanced outputs, closely matching reference [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Limitations of CanonCGT [PITH_FULL_IMAGE:figures/full_fig_p008_9.png] view at source ↗
read the original abstract

Reference-based color grading aims to reproduce the tonal mood and lighting of a reference while preserving color harmony and scene structure. Existing photorealistic and filter-based methods often produce unstable tone mappings -- over-shifting or inconsistently retaining colors -- leading to unnatural results. We propose CanonCGT, a two-stage framework built on a canonical pivot -- a style-neutral intermediate representation for stable color mapping. The first stage canonicalizes the input by removing intrinsic tonal bias, and the second color-grades it to match the reference style. A dual-phase training scheme, DP-CGT, combines supervised preset learning with self-supervised refinement on unpaired photographs. CanonCGT delivers photorealistic and tonally consistent results across diverse datasets, surpassing state-of-the-art methods in stability and visual fidelity. Our codes are available at \href{https://github.com/Jinwon-Ko/CanonCGT}{https://github.com/Jinwon-Ko/CanonCGT}

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces CanonCGT, a two-stage reference-based color grading framework that relies on a canonical pivot as a style-neutral intermediate representation. The first stage canonicalizes the input image by removing its intrinsic tonal bias; the second stage maps the canonicalized result to the tonal style of a reference image. Training uses a dual-phase scheme (DP-CGT) that combines supervised preset learning with self-supervised refinement on unpaired data. The authors claim that the method produces photorealistic, tonally consistent outputs that surpass prior photorealistic and filter-based approaches in stability and visual fidelity across diverse datasets, with code released at the cited GitHub repository.

Significance. If the central claim holds, the work would offer a concrete architectural solution to the documented instability problems (over-shifting, inconsistent color retention) that affect existing reference-based grading pipelines. The explicit separation into canonicalization and grading stages, together with the dual-phase training protocol and public code release, would constitute a reproducible contribution that could be directly tested and extended by the community.

major comments (2)
  1. [§3.2] §3.2 (Canonical Pivot Construction): the manuscript must supply the precise mathematical definition of the canonical pivot (including any learned parameters or loss terms that enforce style neutrality). Without an explicit equation or algorithmic listing, it is impossible to verify whether the pivot is truly parameter-free or whether its construction inadvertently encodes reference-specific statistics that would undermine the stability claim.
  2. [§4.2, Table 2] §4.2 and Table 2 (Quantitative Evaluation): the reported superiority over SOTA methods is stated in the abstract but the specific metrics, baselines, and statistical significance tests are not visible in the provided abstract; the full manuscript must include per-dataset PSNR/SSIM/LPIPS tables with error bars and a clear statement of the number of reference–input pairs used, because the central claim of “surpassing state-of-the-art in stability” rests on these numbers.
minor comments (2)
  1. The abstract mentions “diverse datasets” but does not name them; the experiments section should list the exact datasets and splits used for both supervised and self-supervised phases.
  2. Notation for the two stages (canonicalization network vs. grading network) should be introduced once and used consistently; currently the abstract uses only descriptive phrases.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation for major revision. We address each major comment below and will update the manuscript to improve clarity and completeness.

read point-by-point responses

  1. Referee: [§3.2] §3.2 (Canonical Pivot Construction): the manuscript must supply the precise mathematical definition of the canonical pivot (including any learned parameters or loss terms that enforce style neutrality). Without an explicit equation or algorithmic listing, it is impossible to verify whether the pivot is truly parameter-free or whether its construction inadvertently encodes reference-specific statistics that would undermine the stability claim.

    Authors: We agree that an explicit mathematical definition is required for reproducibility and to substantiate the stability claims. Section 3.2 describes the canonical pivot conceptually as a style-neutral intermediate representation obtained via the first-stage canonicalization network, but we acknowledge that the precise equations, including any parameters and the loss terms enforcing neutrality (e.g., the style-invariance loss), were not presented in equation form. In the revised manuscript we will insert the full mathematical formulation of the pivot construction together with the relevant loss terms. revision: yes

  2. Referee: [§4.2, Table 2] §4.2 and Table 2 (Quantitative Evaluation): the reported superiority over SOTA methods is stated in the abstract but the specific metrics, baselines, and statistical significance tests are not visible in the provided abstract; the full manuscript must include per-dataset PSNR/SSIM/LPIPS tables with error bars and a clear statement of the number of reference–input pairs used, because the central claim of “surpassing state-of-the-art in stability” rests on these numbers.

    Authors: The full manuscript already contains Table 2 in §4.2 reporting PSNR, SSIM and LPIPS on multiple datasets against the listed baselines. To strengthen the presentation we will augment the table with per-dataset error bars (standard deviation across runs), explicitly state the number of reference–input pairs evaluated for each metric, and add a brief note on statistical significance where appropriate. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The provided abstract and description contain no equations, derivations, or explicit parameter-fitting steps that could reduce a claimed result to its inputs by construction. The central claim is a two-stage framework using a canonical pivot for color grading, presented at the level of architectural description without any visible self-definitional loops, fitted-input predictions, or load-bearing self-citations that would require external verification. No load-bearing mathematical steps are supplied that could be inspected for equivalence to inputs, making the derivation self-contained at the level of available text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The central claim rests on the existence and learnability of a style-neutral canonical pivot representation and the effectiveness of the dual-phase training scheme; no free parameters, axioms, or additional invented entities beyond the pivot are described in the abstract.

invented entities (1)
  • canonical pivot no independent evidence
    purpose: style-neutral intermediate representation for stable color mapping
    Introduced as the core building block of the two-stage framework that separates canonicalization from style application.

pith-pipeline@v0.9.1-grok · 5692 in / 1140 out tokens · 27064 ms · 2026-06-28T15:29:16.145555+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

55 extracted references · 3 canonical work pages · 2 internal anchors

  1. [1]

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

    Eirikur Agustsson and Radu Timofte. Ntire 2017 challenge on single image super-resolution: Dataset and study. InProc. IEEE CVPR Workshops, pages 126–135, 2017. 6

    2017

  2. [2]

    Ultrafast photorealistic style transfer via neural architecture search

    Jie An, Haoyi Xiong, Jun Huan, and Jiebo Luo. Ultrafast photorealistic style transfer via neural architecture search. In Proc. AAAI, 2020. 2, 6, 13, 14

    2020

  3. [3]

    Food-101–mining discriminative components with random forests

    Lukas Bossard, Matthieu Guillaumin, and Luc Van Gool. Food-101–mining discriminative components with random forests. InProc. ECCV, 2014. 6

    2014

  4. [4]

    Learning photographic global tonal adjustment with a database of input/output image pairs

    Vladimir Bychkovsky, Sylvain Paris, Eric Chan, and Fr ´edo Durand. Learning photographic global tonal adjustment with a database of input/output image pairs. InProc. IEEE CVPR,

  5. [5]

    Stylebank: An explicit representation for neural image style transfer

    Dongdong Chen, Lu Yuan, Jing Liao, Nenghai Yu, and Gang Hua. Stylebank: An explicit representation for neural image style transfer. InProc. IEEE CVPR, 2017. 2

    2017

  6. [6]

    Nlut: Neural-based 3d lookup tables for video photorealistic style transfer.arXiv preprint arXiv:2303.09170, 2023

    Yaosen Chen, Han Yang, Yuexin Yang, Yuegen Liu, Wei Wang, Xuming Wen, and Chaoping Xie. Nlut: Neural-based 3d lookup tables for video photorealistic style transfer.arXiv preprint arXiv:2303.09170, 2023. 3

    work page arXiv 2023

  7. [7]

    Deep photo enhancer: Unpaired learning for image enhancement from photographs with gans

    Yu-Sheng Chen, Yu-Ching Wang, Man-Hsin Kao, and Yung- Yu Chuang. Deep photo enhancer: Unpaired learning for image enhancement from photographs with gans. InProc. IEEE CVPR, 2018. 6

    2018

  8. [8]

    Photowct2: Compact autoencoder for photorealistic style transfer resulting from blockwise training and skip connections of high-frequency residuals

    Tai-Yin Chiu and Danna Gurari. Photowct2: Compact autoencoder for photorealistic style transfer resulting from blockwise training and skip connections of high-frequency residuals. InProc. IEEE WACV, 2022. 1, 2, 6, 13, 14, 15

    2022

  9. [9]

    Stytr2: Image style transfer with transformers

    Yingying Deng, Fan Tang, Weiming Dong, Chongyang Ma, Xingjia Pan, Lei Wang, and Changsheng Xu. Stytr2: Image style transfer with transformers. InProc. IEEE CVPR, 2022. 2

    2022

  10. [10]

    A Learned Representation For Artistic Style

    Vincent Dumoulin, Jonathon Shlens, and Manjunath Kud- lur. A learned representation for artistic style.arXiv preprint arXiv:1610.07629, 2016. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  11. [11]

    Im- age style transfer using convolutional neural networks

    Leon A Gatys, Alexander S Ecker, and Matthias Bethge. Im- age style transfer using convolutional neural networks. In Proc. IEEE CVPR, 2016. 1, 2

    2016

  12. [12]

    Con- ditional sequential modulation for efficient global image re- touching

    Jingwen He, Yihao Liu, Yu Qiao, and Chao Dong. Con- ditional sequential modulation for efficient global image re- touching. InProc. ECCV, 2020. 6

    2020

  13. [13]

    Gaussian Error Linear Units (GELUs)

    Dan Hendrycks and Kevin Gimpel. Gaussian error linear units (GELUs).arXiv preprint arXiv:1606.08415, 2016. 4

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  14. [14]

    Deep preset: Blending and retouching photos with color style transfer

    Man M Ho and Jinjia Zhou. Deep preset: Blending and retouching photos with color style transfer. InProc. IEEE WACV, 2021. 2, 3, 6, 7, 8, 13, 14, 15

    2021

  15. [15]

    Exposure: A white-box photo post-processing framework.ACM Trans

    Yuanming Hu, Hao He, Chenxi Xu, Baoyuan Wang, and Stephen Lin. Exposure: A white-box photo post-processing framework.ACM Trans. Graphics, 37(2):1–17, 2018. 6

    2018

  16. [16]

    A software plat- form for manipulating the camera imaging pipeline

    Hakki Can Karaimer and Michael S Brown. A software plat- form for manipulating the camera imaging pipeline. InProc. ECCV, 2016. 3

    2016

  17. [17]

    Neural preset for color style transfer

    Zhanghan Ke, Yuhao Liu, Lei Zhu, Nanxuan Zhao, and Ryn- son WH Lau. Neural preset for color style transfer. InProc. IEEE CVPR, 2023. 1, 2, 3, 6, 7, 13, 14, 15

    2023

  18. [18]

    Supervised contrastive learning

    Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. Supervised contrastive learning. InProc. NeurIPS, 2020. 4, 5, 9

    2020

  19. [19]

    Lutformer: Lookup table transformer for image enhance- ment.Neurocomputing, 2025

    Jinwon Ko, Keunsoo Ko, Hanul Kim, and Chang-Su Kim. Lutformer: Lookup table transformer for image enhance- ment.Neurocomputing, 2025. 3

    2025

  20. [20]

    D-lut: Photorealistic style transfer via dif- fusion process

    Mujing Li, Guanjie Wang, Xingguang Zhang, Qifeng Liao, and Chenxi Xiao. D-lut: Photorealistic style transfer via dif- fusion process. InProc. IEEE WACV, 2025. 3

    2025

  21. [21]

    Universal style transfer via feature transforms

    Yijun Li, Chen Fang, Jimei Yang, Zhaowen Wang, Xin Lu, and Ming-Hsuan Yang. Universal style transfer via feature transforms. InProc. NeurIPS, 2017. 2

    2017

  22. [22]

    A closed-form solution to photorealistic image stylization

    Yijun Li, Ming-Yu Liu, Xueting Li, Ming-Hsuan Yang, and Jan Kautz. A closed-form solution to photorealistic image stylization. InProc. ECCV, 2018. 1, 2

    2018

  23. [23]

    Lsdir: A large scale dataset for image restoration

    Yawei Li, Kai Zhang, Jingyun Liang, Jiezhang Cao, Ce Liu, Rui Gong, Yulun Zhang, Hao Tang, Yun Liu, Denis Deman- dolx, et al. Lsdir: A large scale dataset for image restoration. InProc. IEEE CVPR, 2023. 6

    2023

  24. [24]

    PPR10K: A large-scale portrait photo retouch- ing dataset with human-region mask and group-level consis- tency

    Jie Liang, Hui Zeng, Miaomiao Cui, Xuansong Xie, and Lei Zhang. PPR10K: A large-scale portrait photo retouch- ing dataset with human-region mask and group-level consis- tency. InProc. IEEE CVPR, 2021. 6

    2021

  25. [25]

    Enhanced deep residual networks for single image super-resolution

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. Enhanced deep residual networks for single image super-resolution. InProc. IEEE CVPR Workshops, pages 136–144, 2017. 6

    2017

  26. [26]

    Adacm: adaptive colormlp for real-time universal photo-realistic style transfer

    Tianwei Lin, Honglin Lin, Fu Li, Dongliang He, Wenhao Wu, Meiling Wang, Xin Li, and Yong Liu. Adacm: adaptive colormlp for real-time universal photo-realistic style transfer. InProc. AAAI, 2023. 3

    2023

  27. [27]

    Adaattn: Revisit attention mechanism in arbitrary neural style transfer

    Songhua Liu, Tianwei Lin, Dongliang He, Fu Li, Meiling Wang, Xin Li, Zhengxing Sun, Qian Li, and Errui Ding. Adaattn: Revisit attention mechanism in arbitrary neural style transfer. InProc. IEEE ICCV, 2021. 2

    2021

  28. [28]

    Sgdr: Stochastic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. InProc. ICLR, 2017. 9

    2017

  29. [29]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. InProc. ICLR, 2019. 9

    2019

  30. [30]

    Structure-preserving super resolution with gradient guidance

    Cheng Ma, Yongming Rao, Yean Cheng, Ce Chen, Jiwen Lu, and Jie Zhou. Structure-preserving super resolution with gradient guidance. InProc. IEEE CVPR, 2020. 9

    2020

  31. [31]

    Visualizing data using t-sne.Journal of machine learning research, 9,

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of machine learning research, 9,

  32. [32]

    The relationship between photo retouching and color grading

    Isabelle Magrin-Chagnolleau. The relationship between photo retouching and color grading. Technical report, Chap- man University Presidential Fellows Research, 2022. 1

    2022

  33. [33]

    Deeplpf: Deep local para- metric filters for image enhancement

    Sean Moran, Pierre Marza, Steven McDonagh, Sarah Parisot, and Gregory Slabaugh. Deeplpf: Deep local para- metric filters for image enhancement. InProc. IEEE CVPR,

  34. [34]

    Comparing im- ages using color coherence vectors

    Greg Pass, Ramin Zabih, and Justin Miller. Comparing im- ages using color coherence vectors. InProc. ACM Multime- dia, 1997. 6 22

    1997

  35. [35]

    Film: Visual reasoning with a general conditioning layer

    Ethan Perez, Florian Strub, Harm De Vries, Vincent Du- moulin, and Aaron Courville. Film: Visual reasoning with a general conditioning layer. InProc. AAAI, 2018. 4

    2018

  36. [36]

    N- dimensional probability density function transfer and its ap- plication to color transfer

    Francois Pitie, Anil C Kokaram, and Rozenn Dahyot. N- dimensional probability density function transfer and its ap- plication to color transfer. InProc. IEEE ICCV, 2005. 2

    2005

  37. [37]

    Au- tomated colour grading using colour distribution transfer

    Franc ¸ois Piti´e, Anil C Kokaram, and Rozenn Dahyot. Au- tomated colour grading using colour distribution transfer. Comput. Vis. Image Understand., 107:123–137, 2007

    2007

  38. [38]

    Color transfer between images.IEEE CGA, 21:34– 41, 2002

    Erik Reinhard, Michael Adhikhmin, Bruce Gooch, and Peter Shirley. Color transfer between images.IEEE CGA, 21:34– 41, 2002. 2

    2002

  39. [39]

    Imagenet large scale visual recognition challenge.Int

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, et al. Imagenet large scale visual recognition challenge.Int. J. Comput. Vis., 115: 211–252, 2015. 10

    2015

  40. [40]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zh- moginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. InProc. IEEE CVPR, 2018. 3, 9

    2018

  41. [41]

    Video color grading via look-up table generation

    Seunghyun Shin, Dongmin Shin, Jisu Shin, Hae-Gon Jeon, and Joon-Young Lee. Video color grading via look-up table generation. InProc. IEEE ICCV, 2025. 3

    2025

  42. [42]

    Very deep convo- lutional networks for large-scale image recognition.Proc

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition.Proc. ICLR, 2015. 10

    2015

  43. [43]

    Ldc: Lightweight dense cnn for edge detection.IEEE Access, 10:68281–68290, 2022

    Xavier Soria, Gonzalo Pomboza-Junez, and Angel Domingo Sappa. Ldc: Lightweight dense cnn for edge detection.IEEE Access, 10:68281–68290, 2022. 6

    2022

  44. [44]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. Attention is all you need. InProc. NeurIPS,

  45. [45]

    Underexposed photo enhance- ment using deep illumination estimation

    Ruixing Wang, Qing Zhang, Chi-Wing Fu, Xiaoyong Shen, Wei-Shi Zheng, and Jiaya Jia. Underexposed photo enhance- ment using deep illumination estimation. InProc. IEEE CVPR, 2019. 6

    2019

  46. [46]

    Cap- vstnet: Content affinity preserved versatile style transfer

    Linfeng Wen, Chengying Gao, and Changqing Zou. Cap- vstnet: Content affinity preserved versatile style transfer. In Proc. IEEE CVPR, 2023. 2, 6, 13, 14, 15

    2023

  47. [47]

    Google landmarks dataset v2-a large-scale benchmark for instance-level recognition and retrieval

    Tobias Weyand, Andre Araujo, Bingyi Cao, and Jack Sim. Google landmarks dataset v2-a large-scale benchmark for instance-level recognition and retrieval. InProc. IEEE CVPR, 2020. 6

    2020

  48. [48]

    Adaint: Learning adaptive intervals for 3d lookup tables on real-time image enhancement

    Canqian Yang, Meiguang Jin, Xu Jia, Yi Xu, and Ying Chen. Adaint: Learning adaptive intervals for 3d lookup tables on real-time image enhancement. InProc. IEEE CVPR, 2022. 3

    2022

  49. [49]

    Seplut: Separable image-adaptive lookup tables for real-time image enhancement

    Canqian Yang, Meiguang Jin, Yi Xu, Rui Zhang, Ying Chen, and Huaida Liu. Seplut: Separable image-adaptive lookup tables for real-time image enhancement. InProc. ECCV,

  50. [50]

    Filter style transfer between photos

    Jonghwa Yim, Jisung Yoo, Won-joon Do, Beomsu Kim, and Jihwan Choe. Filter style transfer between photos. InProc. ECCV, 2020. 2, 3

    2020

  51. [51]

    Photorealistic style transfer via wavelet transforms

    Jaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang, and Jung-Woo Ha. Photorealistic style transfer via wavelet transforms. InProc. IEEE ICCV, 2019. 1, 2, 7

    2019

  52. [52]

    Learning image-adaptive 3d lookup tables for high perfor- mance photo enhancement in real-time.IEEE Trans

    Hui Zeng, Jianrui Cai, Lida Li, Zisheng Cao, and Lei Zhang. Learning image-adaptive 3d lookup tables for high perfor- mance photo enhancement in real-time.IEEE Trans. Pattern Anal. Mach. Intell., 44(4):2058–2073, 2020. 3, 6

    2058

  53. [53]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProc. IEEE CVPR, 2018. 6, 9

    2018

  54. [54]

    Research on color management and color grading applications in new media movies.Journal of New Media Arts and Technology, 7(1):45–53, 2025

    Yifan Zhang and Lei Wang. Research on color management and color grading applications in new media movies.Journal of New Media Arts and Technology, 7(1):45–53, 2025. 1

    2025

  55. [55]

    Do- main enhanced arbitrary image style transfer via contrastive learning

    Yuxin Zhang, Fan Tang, Weiming Dong, Haibin Huang, Chongyang Ma, Tong-Yee Lee, and Changsheng Xu. Do- main enhanced arbitrary image style transfer via contrastive learning. InACM SIGGRAPH 2022 conference proceedings,

    2022