Pith. sign in

REVIEW 4 major objections 6 minor 52 references

DeQA-Doc: Adapting DeQA-Score to Document Image Quality Assessment

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Adapting an MLLM image-quality scorer to document images reaches Final Score 0.9288 on DIQA-5000.

desk verdict Useful and honest adaptation of DeQA-Score to documents, but Eq. (4) is sign-flipped as written and the test phase only has internal comparisons. read the letter →

arxiv 2507.12796 v1 pith:6XRHS7OH submitted 2025-07-17 cs.CV

Junjie Gao , Runze Liu , Yingzhe Peng , Shujian Yang , Jin Zhang , Kai Yang , Zhiyuan You This is my paper · ORCID
classification cs.CV
keywords documentimagequalityassessmentmulti-modallargelanguagemodelscoredistributionregressionsoftlabelpseudo-varianceDIQA-5000ensembleinferenceresolution-flexiblevisionencoding
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

DeQA-Doc sets out to show that a multimodal-large-language-model quality scorer can be moved from natural images to document images without redesigning the core. The paper adapts DeQA-Score, which regresses a continuous quality score from a predicted distribution over five quality levels, to the document domain by supplying soft labels even though DIQA-5000 lacks annotator variance. It does this with a fixed pseudo-variance borrowed from natural-image datasets, or with linear interpolation between adjacent levels, and it relaxes CLIP's fixed resolution so high-resolution document layouts are not crushed into 448 by 448. On the DIQA-5000 test set the full ensemble of mPLUG-Owl2 and Qwen2.5-VL variants reaches a Final Score of 0.9288, which the paper reports as a clear improvement over its internally re-implemented Q-Align and DeQA-Score baselines. If this holds, MLLMs offer a practical route to document-quality assessment for OCR, digitization, and archival.

What carries the argument

The load-bearing mechanism is score-distribution regression with soft labels. DeQA-Score discretizes a Gaussian MOS distribution over five rating levels, applies a linear post-adjustment so the discrete expectation equals the mean score, and trains the MLLM's level token against this distribution with a KL divergence loss while other tokens use standard cross-entropy. DeQA-Doc keeps this mechanism and replaces the missing variance in two ways: a fixed pseudo-standard-deviation of 20 percent of the score range, and linear interpolation between the two adjacent levels around the mean score. To handle large document images, it removes the absolute position embeddings from the CLIP vision encoder or switches to Qwen2.5-VL, which accepts original resolution. At inference, predicted probabilities over the five levels are converted to a mean score, and ensemble inference averages the predicted distributions across models and prompt variants.

What would settle it

Measure per-image annotator variance on a random subset of DIQA-5000 by collecting multiple quality ratings per document. If the true standard deviation differs substantially from the assumed 0.2 times the score range, or if retraining with the measured variances fails to reproduce the reported gains over Q-Align and DeQA-Score, the central claim would be undercut.

Watch

Extended reading notes

Core claim

The central claim is that document image quality can be scored accurately by an MLLM trained with distribution-based soft labels, even when the only supervision is a mean opinion score. DeQA-Doc treats quality as a discrete five-level distribution and supervises the level token with KL divergence against a soft label; the paper's contribution is to construct that soft label without per-image variance data and to make the vision encoder resolution-flexible. On the DIQA-5000 benchmark, the best configuration — averaging predictions from mPLUG-Owl2 and Qwen2.5-VL variants, with KonIQ pretraining and LoRA or full fine-tuning at 1024 resolution or original resolution — achieves a Final Score of 0.9288, outperforming the paper's own Q-Align baseline at 0.8523 and DeQA-Score baseline at 0.8849 on the validation split. The paper interprets this as evidence that a single MLLM-based scoring framework can generalize across image and document quality, and that ensembling complementary backbones plus preserving high input resolution matters more than prompt variation.

Load-bearing premise

The method assumes that human score variance on document images is roughly the same as on natural images — about 20 percent of the score range — even though the DIQA-5000 dataset provides no variance to check this against.

Editorial extensions

If this is right

  • Document quality assessment can be driven by an MLLM trained only on mean opinion scores, so variance-free datasets are no longer a barrier to distribution-based soft-label training.
  • Removing absolute position embeddings from a CLIP-based vision encoder lets a fixed-resolution MLLM process 1024 or 1536 resolution documents, with 1024 performing best in the paper's ablations.
  • Averaging across mPLUG-Owl2 and Qwen2.5-VL variants yields consistent gains over any single model, while averaging across ten prompt templates yields no measurable gain.
  • Pretraining on natural-image quality data (KonIQ) and using LoRA tuning each give a small further improvement, suggesting cross-domain transfer works in both directions.
  • The full ensemble's 0.9288 Final Score on the DIQA-5000 test set is the paper's evidence that MLLM-based scoring can serve OCR, digitization, and archival pipelines.

Reading between the lines

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

  • If document annotators have noticeably different score variance than natural-image annotators, the fixed 20-percent pseudo-variance could be miscalibrated; the paper does not test this directly, so a measurement of per-image variance on DIQA-5000 would be the natural next experiment.
  • Because prompt ensemble did not help while model ensemble did, the main source of DeQA-Doc's robustness may be architectural diversity rather than prompt diversity — a hypothesis the paper leaves implicit.
  • The resolution-flexible score-distribution recipe could plausibly transfer to other high-resolution visual quality tasks such as screen-content, medical, or archival image assessment, since none of the components is document-specific.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes DeQA-Doc, an adaptation of the authors' earlier DeQA-Score MLLM-based image quality scorer to the document quality assessment task. Since DIQA-5000 provides only mean opinion scores without per-image variance, the authors construct soft labels either by assigning a fixed pseudo standard deviation of 20% of the score range or by linear interpolation between adjacent quality levels. They also relax input-resolution constraints by removing absolute position embeddings from the CLIP encoder or by switching to Qwen2.5-VL, and they combine model and prompt ensembles at inference. On the DIQA-5000 validation and hidden test splits, the paper reports final scores up to 0.9288 for an ensemble of mPLUG-Owl2 and Qwen2.5-VL variants, and claims significant improvements over existing baselines.

Significance. If the results are reproducible, the paper offers a practical recipe for adapting score-distribution-regression MLLMs to document images and demonstrates that a MOS-only dataset can be used with an assumed variance model. The release of code and model weights is a strength, as is the use of an external benchmark (DIQA-5000). However, the significance of the empirical claim is currently limited by the absence of external baselines on the hidden test set, the lack of uncertainty estimates, and a training objective that is written with an incorrect sign in a load-bearing equation.

major comments (4)
  1. [Section 3.2, Eq. (4)] The KL divergence loss is written as L_kl = Σ_i p_i log(ppred_i / p_i), which equals -KL(p || ppred). Minimizing this expression maximizes the KL divergence between the soft label and the model prediction. Because the level token receives only this KL loss, a model trained literally as specified would be pushed away from the target soft label and could not produce the reported correlations. Please correct the sign or reverse the argument order, and verify in the released code that the implemented loss matches the corrected formula. As currently written, the central training objective is inconsistent with the reported results.
  2. [Section 4.4, Tables 5–7] The test-phase evaluation compares only internal variants of DeQA-Doc. No published external baseline (for example, Q-Align, DeQA-Score, or a traditional DIQA method) is evaluated on the hidden DIQA-5000 test set in the same protocol. The only comparison with Q-Align and DeQA-Score appears in Table 1 on the validation split. The abstract's claim that DeQA-Doc 'significantly outperforms existing baselines' is therefore not supported by the test-phase evidence presented here; please add external baseline scores on the same test protocol or temper the claim accordingly.
  3. [Section 3.1.2] The pseudo standard deviation is fixed at 20% of the score range based on statistics from natural-image datasets (KonIQ, SPAQ, KADID). No evidence is given that document-image annotation variance is similar, and no sensitivity analysis with other pseudo-variance values is reported. Since the pseudo-variance soft label is consistently preferred over linear interpolation (Table 3), the reported advantage could depend on the imported 20% constant. Please report results for a range of pseudo standard deviation values (for example, 0.1, 0.2, and 0.3 of the score range) and, if possible, validate the assumption on a document dataset that contains per-item variance information.
  4. [Section 4.2] The text states that 'these four score results ... are obtained through submitting inferred quality scores to the official workshop website. Therefore, the SRCC and PLCC results are not available.' This is contradictory because Tables 1–7 report dimension scores that are derived from SRCC and PLCC. Please clarify how the reported numbers were actually obtained and whether the final score in Eq. (7) is computed from independently computed correlations or from a single submitted score. In addition, no error bars or significance tests are provided for any table; differences of 0.002–0.008 (for example, m0 versus m1 in Table 5) are presented as improvements without statistical support.
minor comments (6)
  1. [Abstract and Section 3.2] The phrase 'resolution constrains' should be 'resolution constraints', and 'challenge lays in' should be 'challenge lies in'.
  2. [Section 4.3, Table 4] The text says LoRA yields improvements across all attributes, but LoRA's overall score (0.9054) is lower than full fine-tuning's overall score (0.9063). Please correct this description.
  3. [Section 4.4, Table 6] The '5-fold ensemble' is not described precisely; please specify how the five models are trained (for example, different seeds or data subsets) and how their predictions are combined.
  4. [Section 4.1] The pretraining on KonIQ used in Table 5 (model m3) is not described in the implementation details; please specify the pretraining data split, input resolution, loss, and number of epochs.
  5. [References] Reference [30] is cited as 'B Series'; this should be a proper citation, for example ITU-R Recommendation BT.500, with the title and year.
  6. [Section 4.3] The sentence describing Table 1 says 'The soft label in DeQA-Score is calculated by pseudo variance.' Q-Align uses one-hot labels rather than pseudo-variance soft labels, so the wording should be clarified to avoid implying that the same label construction applies to both methods.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction: DeQA-Doc's benchmark numbers rest on external DIQA-5000 labels; DeQA-Score self-citations are explicit assumptions, not forced predictions (Eq. (4) sign issue is a separate correctness risk).

full rationale

DeQA-Doc is a direct adaptation of DeQA-Score, and the paper repeatedly cites DeQA-Score for the soft-label construction, the 20% pseudo-variance statistic, the linear-interpolation surrogate, and the equivalence of full-vocabulary and closed-set softmax. These are self-citations because Zhiyuan You is an author of both works. I do not count them as circular steps: each is an explicit assumption or a transfer from external natural-image IQA datasets (KonIQ/SPAQ/KADID), and the reported correlations are computed against DIQA-5000's human MOS on a held-out test set that is not constructed by the paper. No equation reduces a reported prediction to a fitted input; Eq. (2) merely forces the soft label's mean to equal the MOS (a label-construction property), and Eq. (5) defines the predicted mean from model probabilities. The paper honestly notes that prompt ensemble gave no gain (Section 3.3, Table 7). Separately, Eq. (4) is written as sum p_i log(ppred_i/p_i), which equals -KL(p||ppred); this is a potential sign error that would make the stated objective repulsive, but it is a correctness/implementation issue, not a circularity, and would need code inspection to resolve. I therefore assign 2 for the self-citation density rather than for a circular reduction.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central method relies on a fixed pseudo-variance ratio (0.2 of score range) and a selected input resolution, both chosen without direct evidence for document images. Key assumptions are the Gaussian score distribution, the linear mapping of five quality levels to [1,5], and that removing absolute position embeddings preserves layout information.

free parameters (2)
  • pseudo standard deviation ratio = 0.2 x (max - min) of score range
    Fixed pseudo std used to construct Gaussian soft labels when dataset variance is unavailable; borrowed from average std ratio in KonIQ, SPAQ, KADID as cited in DeQA-Score.
  • input resolution = 1024x1024
    Selected from validation ablation (448, 1024, 1536) as the best trade-off; a hyperparameter affecting performance.
assumptions (4)
  • domain assumption Gaussian distribution of human quality scores: x ~ N(mu, sigma^2)
    Standard assumption in IQA, used by DeQA-Score; invoked in Section 3.1.1 to construct soft labels.
  • ad hoc to paper Pseudo variance of 20% of the score range approximates true document score variance
    The paper assigns a fixed std of 0.2*(max-min) to DIQA-5000 because variance is missing; this is a heuristic borrowed from natural image datasets, not verified on documents. Section 3.1.2.
  • ad hoc to paper Removing absolute position embeddings from CLIP preserves document layout information
    The paper removes absolute position embeddings to support large resolutions; this assumes relative spatial relations remain sufficient, which is untested for layout-sensitive document images. Section 3.2.
  • domain assumption Quality levels map linearly to MOS scale [1,5]
    The five levels are assigned centers 1 to 5 with equal width; this discretization follows DeQA-Score and assumes equal intervals. Section 3.1.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DeQA-Doc: Adapting DeQA-Score to Document Image Quality Assessment." pith.science (2026). https://pith.science/paper/6XRHS7OH

@misc{pith2026250712796,
  author       = {Pith},
  title        = {Pith review of: DeQA-Doc: Adapting DeQA-Score to Document Image Quality Assessment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6XRHS7OH}},
  note         = {Machine review of arXiv:2507.12796}
}
read the original abstract

Document quality assessment is critical for a wide range of applications including document digitization, OCR, and archival. However, existing approaches often struggle to provide accurate and robust quality scores, limiting their applicability in practical scenarios. With the rapid progress in Multi-modal Large Language Models (MLLMs), recent MLLM-based methods have achieved remarkable performance in image quality assessment. In this work, we extend this success to the document domain by adapting DeQA-Score, a state-of-the-art MLLM-based image quality scorer, for document quality assessment. We propose DeQA-Doc, a framework that leverages the visual language capabilities of MLLMs and a soft label strategy to regress continuous document quality scores. To adapt DeQA-Score to DeQA-Doc, we adopt two complementary solutions to construct soft labels without the variance information. Also, we relax the resolution constrains to support the large resolution of document images. Finally, we introduce ensemble methods to further enhance the performance. Extensive experiments demonstrate that DeQA-Doc significantly outperforms existing baselines, offering accurate and generalizable document quality assessment across diverse degradation types. Codes and model weights are available in https://github.com/Junjie-Gao19/DeQA-Doc.

Figures

Figures reproduced from arXiv: 2507.12796 by the authors.

Figure 1
Figure 1. (a) MLLMs need discrete level tokens for training. Thus, continuous scores must be converted into discrete level tokens as training labels. (b) Soft label in DeQA-Score [46]. DeQA-Score discretizes the estimated Gaussian distribution of the score. The soft label better preserves the relationships be￾tween levels, e.g., “fair” is closer to “good” than “excellent”. This figure is adapted from DeQA-Score [46]. {1, 2, 3… view at source ↗
Figure 2
Figure 2. Framework of our DeQA-Doc trained with soft label. For the “<level>” token, the KL divergence loss is calculated between predicted probabilities and our soft label. For other tokens, common cross-entropy loss for next token prediction is calculated. This figure is borrowed from DeQA-Score [46]. 3.2. Training with Soft Label As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Qualitative results of our DeQA-Doc on the DIQA-5000 dataset. From left to right, each column is roughly the “bad”, “poor”, “fair”, “good”, “excellent” documents in sequence. Ablation study on image resolution. To accommodate high-resolution document images, we remove the abso￾lute position embeddings from the CLIP-based vision encoder in mPLUG-Owl2-7B. This modification relaxes the strict resolution constraint (typ… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 48 canonical work pages

  1. [1]

    A new document image quality assess- ment method based on hast derivations

    Alireza Alaei. A new document image quality assess- ment method based on hast derivations. In International Conference on Document Analysis and Recognition (IC- DAR), 2019. 1, 2

    work page 2019

  2. [2]

    Document image quality assessment based on improved gradient magnitude similarity deviation

    Alireza Alaei, Donatello Conte, and Romain Raveaux. Document image quality assessment based on improved gradient magnitude similarity deviation. In International Conference on Document Analysis and Recognition (IC- DAR), 2015. 1, 2

    work page 2015

  3. [3]

    Document image quality assessment: A survey

    Alireza Alaei, Vinh Bui, David Doermann, and Uma- pada Pal. Document image quality assessment: A survey. ACM Computing Surveys, 2023. 1, 2

    work page 2023

  4. [4]

    Qwen2.5-VL technical report

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wen- bin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, Humen Zhong, Yuanzhi Zhu, Mingkun Yang, Zhaohai Li, Jianqiang Wan, Pengfei Wang, Wei Ding, Zheren Fu, Yiheng Xu, Jiabo Ye, Xi Zhang, Tianbao Xie, Zesen Cheng, Hang Zhang, Zhibo Yang, Haiyang Xu, and Junyang Lin. Qwen2.5-VL technical report....

    arXiv 2025

  5. [5]

    Q-Ground: Image quality grounding with large multi-modality models

    Chaofeng Chen, Yang Sensen, Haoning Wu, Liang Liao, Zicheng Zhang, Annan Wang, Wenxiu Sun, Qiong Yan, and Weisi Lin. Q-Ground: Image quality grounding with large multi-modality models. In ACM MM, 2024. 2

    work page 2024

  6. [6]

    Toward generalized image quality assessment: Relaxing the per- fect reference quality assumption

    Du Chen, Tianhe Wu, Kede Ma, and Lei Zhang. Toward generalized image quality assessment: Relaxing the per- fect reference quality assumption. In CVPR, 2025. 2

    work page 2025

  7. [7]

    Image quality assessment: Unifying structure and texture similarity

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P Simon- celli. Image quality assessment: Unifying structure and texture similarity. IEEE TPAMI, 2020

    work page 2020

  8. [8]

    Locally adaptive structure and texture similarity for image quality assessment

    Keyan Ding, Yi Liu, Xueyi Zou, Shiqi Wang, and Kede Ma. Locally adaptive structure and texture similarity for image quality assessment. In ACM MM, 2021. 2

    work page 2021

Show all 52 references
  1. [9]

    Perceptual quality assessment of smart- phone photography

    Yuming Fang, Hanwei Zhu, Yan Zeng, Kede Ma, and Zhou Wang. Perceptual quality assessment of smart- phone photography. In CVPR, 2020. 3

    2020

  2. [10]

    Shift-tolerant perceptual similarity metric

    Abhijay Ghildyal and Feng Liu. Shift-tolerant perceptual similarity metric. In ECCV, 2022. 2

    2022

  3. [11]

    KonIQ-10K: An ecologically valid database for deep learning of blind image quality assessment

    Vlad Hosu, Hanhe Lin, Tamas Sziranyi, and Dietmar Saupe. KonIQ-10K: An ecologically valid database for deep learning of blind image quality assessment. IEEE TIP, 2020. 2, 3, 5, 7

    2020

  4. [12]

    LoRA: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. In ICLR, 2021. 5

    2021

  5. [13]

    Decoupled weight de- cay regularization

    Loshchilov Ilya and Hutter Frank. Decoupled weight de- cay regularization. In ICLR, 2019. 5

    2019

  6. [14]

    PIPAL: a large-scale image quality assessment dataset for perceptual image restoration

    Gu Jinjin, Cai Haoming, Chen Haoyu, Ye Xiaoxing, Jimmy S Ren, and Dong Chao. PIPAL: a large-scale image quality assessment dataset for perceptual image restoration. In ECCV, 2020. 2

    2020

  7. [15]

    A deep learning approach to document image quality as- sessment

    Le Kang, Peng Ye, Yi Li, and David Doermann. A deep learning approach to document image quality as- sessment. In International Conference on Image Pro- cessing (ICIP), 2014. 1, 2

    2014

  8. [16]

    MUSIQ: Multi-scale image quality transformer

    Junjie Ke, Qifei Wang, Yilin Wang, Peyman Milanfar, and Feng Yang. MUSIQ: Multi-scale image quality transformer. In CVPR, 2021. 2

    2021

  9. [17]

    Kumar, F

    J. Kumar, F. Chen, and David Doermann. Sharpness es- timation for document and scene images. In ICPR, 2012. 1, 2

    2012

  10. [18]

    CG-DIQA: No- reference document image quality assessment based on character gradient

    Hongyu Li, Fan Zhu, and Junhua Qiu. CG-DIQA: No- reference document image quality assessment based on character gradient. In International Conference on Pat- tern Recognition (ICPR), 2018. 1, 2

    2018

  11. [19]

    KADID- 10K: A large-scale artificially distorted iqa database

    Hanhe Lin, Vlad Hosu, and Dietmar Saupe. KADID- 10K: A large-scale artificially distorted iqa database. In International Conference on Quality of Multimedia Ex- perience (QoMEX), 2019. 3

    2019

  12. [20]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. In NeurIPS, 2024. 4

    2024

  13. [21]

    A deep transfer learning ap- proach to document image quality assessment

    Tan Lu and Ann Dooms. A deep transfer learning ap- proach to document image quality assessment. In Inter- national Conference on Document Analysis and Recog- nition (ICDAR), 2019. 1, 2

    2019

  14. [22]

    Learning a no-reference quality metric for single-image super-resolution

    Chao Ma, Chih-Yuan Yang, Xiaokang Yang, and Ming- Hsuan Yang. Learning a no-reference quality metric for single-image super-resolution. Computer Vision and Im- age Understanding, 2017. 2

    2017

  15. [23]

    Unified blind quality assessment of compressed natural, graphic, and screen content images

    Xiongkuo Min, Kede Ma, Ke Gu, Guangtao Zhai, Zhou Wang, and Weisi Lin. Unified blind quality assessment of compressed natural, graphic, and screen content images. IEEE TIP, 2017. 1

    2017

  16. [24]

    completely blind

    Anish Mittal, Rajiv Soundararajan, and Alan C Bovik. Making a “completely blind” image quality analyzer. IEEE Sign. Process. Letters, 2012. 2

    2012

  17. [25]

    Training language models to follow instructions with hu- man feedback

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with hu- man feedback. In NeurIPS, 2022. 4

    2022

  18. [26]

    PieAPP: Perceptual image-error assess- ment through pairwise preference

    Ekta Prashnani, Hong Cai, Yasamin Mostofi, and Pradeep Sen. PieAPP: Perceptual image-error assess- ment through pairwise preference. In CVPR, 2018. 2

    2018

  19. [27]

    Learning transferable visual models from natural lan- guage supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sas- try, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural lan- guage supervision. In ICML, 2021. 1, 4

    2021

  20. [28]

    Blind image quality assessment: A natural scene statis- tics approach in the dct domain

    Michele A Saad, Alan C Bovik, and Christophe Charrier. Blind image quality assessment: A natural scene statis- tics approach in the dct domain. IEEE TIP, 2012. 2

    2012

  21. [29]

    Identity documents image quality assessment

    Daniel Schulz, Jose Maureira, Juan Tapia, and Christoph Busch. Identity documents image quality assessment. In European Signal Processing Conference (EUSIPCO),

  22. [30]

    Methodology for the subjective assessment of the quality of television pictures

    B Series. Methodology for the subjective assessment of the quality of television pictures. Recommendation ITU- R BT, 2012. 2

    2012

  23. [31]

    A method of image quality assessment for text recogni- tion on camera-captured and projectively distorted docu- ments

    Julia Shemiakina, Elena Limonova, Natalya Skoryuk- ina, Vladimir V Arlazarov, and Dmitry P Nikolaev. A method of image quality assessment for text recogni- tion on camera-captured and projectively distorted docu- ments. Mathematics, 2021. 1

    2021

  24. [32]

    LLaMA: Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth´ee Lacroix, Bap- tiste Rozi`ere, Naman Goyal, Eric Hambro, Faisal Azhar, et al. LLaMA: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. 4

    2023 arXiv

  25. [33]

    Exploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Exploring clip for assessing the look and feel of images. In AAAI, 2023. 2

    2023

  26. [34]

    Image quality assessment: from error visi- bility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visi- bility to structural similarity. IEEE TIP, 2004. 2

    2004

  27. [35]

    Q-Bench: A bench- mark for general-purpose foundation models on low- level vision

    Haoning Wu, Zicheng Zhang, Erli Zhang, Chaofeng Chen, Liang Liao, Annan Wang, Chunyi Li, Wenxiu Sun, Qiong Yan, Guangtao Zhai, et al. Q-Bench: A bench- mark for general-purpose foundation models on low- level vision. In ICLR, 2024. 2

    2024

  28. [36]

    Q-Instruct: Improv- ing low-level visual abilities for multi-modality founda- tion models

    Haoning Wu, Zicheng Zhang, Erli Zhang, Chaofeng Chen, Liang Liao, Annan Wang, Kaixin Xu, Chunyi Li, Jingwen Hou, Guangtao Zhai, et al. Q-Instruct: Improv- ing low-level visual abilities for multi-modality founda- tion models. In CVPR, 2024. 1, 2

    2024

  29. [37]

    Q-Align: Teaching LMMs for visual scoring via discrete text-defined levels

    Haoning Wu, Zicheng Zhang, Weixia Zhang, Chaofeng Chen, Liang Liao, Chunyi Li, Yixuan Gao, Annan Wang, Erli Zhang, Wenxiu Sun, et al. Q-Align: Teaching LMMs for visual scoring via discrete text-defined levels. In ICML, 2024. 2, 4, 5, 7

    2024

  30. [38]

    Towards open-ended vi- sual quality comparison

    Haoning Wu, Hanwei Zhu, Zicheng Zhang, Erli Zhang, Chaofeng Chen, Liang Liao, Chunyi Li, Annan Wang, Wenxiu Sun, Qiong Yan, et al. Towards open-ended vi- sual quality comparison. In ECCV, 2024. 1, 2

    2024

  31. [39]

    A comprehensive study of multimodal large lan- guage models for image quality assessment

    Tianhe Wu, Kede Ma, Jie Liang, Yujiu Yang, and Lei Zhang. A comprehensive study of multimodal large lan- guage models for image quality assessment. In ECCV,

  32. [40]

    No-reference document image quality as- sessment based on high order image statistics

    Jingtao Xu, Peng Ye, Qiaohong Li, Yong Liu, and David Doermann. No-reference document image quality as- sessment based on high order image statistics. In IEEE International Conference on Image Processing (ICIP) ,

  33. [41]

    MANIQA: Multi-dimension attention network for no- reference image quality assessment

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. MANIQA: Multi-dimension attention network for no- reference image quality assessment. In CVPRW, 2022. 2

    2022

  34. [42]

    Document image quality assessment: A brief survey

    Peng Ye and David Doermann. Document image quality assessment: A brief survey. In International Conference on Document Analysis and Recognition, 2013. 1, 2

    2013

  35. [43]

    mPLUG-Owl2: Revolutionizing multi-modal large lan- guage model with modality collaboration

    Qinghao Ye, Haiyang Xu, Jiabo Ye, Ming Yan, Anwen Hu, Haowei Liu, Qi Qian, Ji Zhang, and Fei Huang. mPLUG-Owl2: Revolutionizing multi-modal large lan- guage model with modality collaboration. In CVPR,

  36. [44]

    Descriptive image quality assessment in the wild

    Zhiyuan You, Jinjin Gu, Zheyuan Li, Xin Cai, Kai- wen Zhu, Chao Dong, and Tianfan Xue. Descriptive image quality assessment in the wild. arXiv preprint arXiv:2405.18842, 2024. 1

    2024

  37. [45]

    Depicting beyond scores: Ad- vancing image quality assessment through multi-modal language models

    Zhiyuan You, Zheyuan Li, Jinjin Gu, Zhenfei Yin, Tian- fan Xue, and Chao Dong. Depicting beyond scores: Ad- vancing image quality assessment through multi-modal language models. In ECCV, 2024

    2024

  38. [46]

    Teaching large language models to regress ac- curate image quality scores using score distribution

    Zhiyuan You, Xin Cai, Jinjin Gu, Tianfan Xue, and Chao Dong. Teaching large language models to regress ac- curate image quality scores using score distribution. In CVPR, 2025. 1, 2, 3, 4, 5, 7

    2025

  39. [47]

    FSIM: A feature similarity index for image quality as- sessment

    Lin Zhang, Lei Zhang, Xuanqin Mou, and David Zhang. FSIM: A feature similarity index for image quality as- sessment. IEEE TIP, 2011. 2

    2011

  40. [48]

    An efficient transformer–cnn network for document image binariza- tion

    Lina Zhang, Kaiyuan Wang, and Yi Wan. An efficient transformer–cnn network for document image binariza- tion. Electronics, 2024. 1

    2024

  41. [49]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In CVPR, 2018. 2

    2018

  42. [50]

    Uncertainty-aware blind image quality assessment in the laboratory and wild

    Weixia Zhang, Kede Ma, Guangtao Zhai, and Xiaokang Yang. Uncertainty-aware blind image quality assessment in the laboratory and wild. IEEE TIP, 2021. 2

    2021

  43. [51]

    Blind image quality assessment via vision-language correspondence: A multitask learn- ing perspective

    Weixia Zhang, Guangtao Zhai, Ying Wei, Xiaokang Yang, and Kede Ma. Blind image quality assessment via vision-language correspondence: A multitask learn- ing perspective. In CVPR, 2023. 1, 2

    2023

  44. [52]

    Adaptive image qual- ity assessment via teaching large multimodal model to compare

    Hanwei Zhu, Haoning Wu, Yixuan Li, Zicheng Zhang, Baoliang Chen, Lingyu Zhu, Yuming Fang, Guangtao Zhai, Weisi Lin, and Shiqi Wang. Adaptive image qual- ity assessment via teaching large multimodal model to compare. arXiv preprint arXiv:2405.19298, 2024. 1

    2024 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.