Pith. sign in

REVIEW 4 major objections 4 minor 62 references

Customizable ROI-Based Deep Image Compression

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

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

Pith's one-line read Text prompts, not fixed masks, can drive ROI image compression.

desk verdict Useful integration of text-prompted ROI masks with latent mask attention, but the headline SOTA curves are run with ground-truth masks, so the customizable claim is not yet supported. read the letter →

arxiv 2507.00373 v3 pith:TMIXPYXR submitted 2025-07-01 cs.CV eess.IV

Jian Jin , Fanxin Xia , Feng Ding , Xinfeng Zhang , Meiqin Liu , Yao Zhao , Weisi Lin , Lili Meng This is my paper · ORCID
classification cs.CVeess.IV
keywords ROI-basedimagecompressionlearnedtext-controlledmaskacquisitionlatentattentionrate-distortionoptimizationqualitytrade-offfactorzero-shotmachinevisionevaluation
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

The paper tries to make ROI-based image compression customizable: a user types the thing they care about ("person", "fence", "grass") and that text becomes the mask that steers where bits go. It also gives the user a single number, the quality trade-off factor $\sigma \in [0,1]$, that sets how much quality the non-ROI gets relative to the ROI. The authors claim that fusing a latent mask prior with a latent rate-distortion prior improves both reconstruction quality and downstream object detection and instance segmentation at the same bitrate, compared with existing ROI codecs. The practical payoff is that ROI selection no longer needs a per-image segmentation step, and new categories can be handled without retraining.

What carries the argument

The load-bearing machinery is the LMA module's two-prior fusion in latent space. The Mask Representation sub-module turns the single-channel mask into a latent spatial prior $l$ with the same shape as the latent $y$, so mask guidance no longer crosses from image space to latent space; the Importance-map Generation sub-module produces a latent RDO prior $i$ that flags which latent details most affect reconstruction. Their element-wise sum forms attention map $s$, which is multiplied onto $y$; the CVA mechanism controls the mask's non-ROI value through $\sigma$, and a weighted distortion loss using the same mask reinforces the trade-off during training.

What would settle it

Run the pipeline on a held-out validation set twice, once with ground-truth masks and once with masks produced only from the text prompt, and compare the ROI-PSNR and detection mAP curves; if the text-mask curves fall below the ground-truth-mask curves by more than the margin the paper reports over its anchors, the practical text-customization claim fails even if the GT-mask experiments are correct.

Watch

Extended reading notes

Core claim

The central discovery is that the ROI mask can be replaced by a text-derived similarity map and still drive a deep codec effectively, provided the mask is converted into latent-space guidance rather than used as an image-space outline. The proposed pipeline builds a mask $m$ from prompt-to-pixel similarities, then learns a latent spatial prior $l$ from $m$ and a latent RDO prior $i$ from the image's intermediate features; adding $l$ and $i$ and multiplying the sum onto the latent representation tells the entropy model where bits matter. With ground-truth masks, this arrangement reports higher ROI-PSNR and average PSNR than the compared ROI codecs, and with uniform masks it still beats the anchors, indicating the prior fusion itself carries much of the benefit.

Load-bearing premise

The quantitative gains against the anchors come from substituting ground-truth segmentation masks for the text-produced similarities, so the whole customization story rests on text masks being accurate enough to reproduce those gains in practice.

Editorial extensions

If this is right

  • A user can switch ROI from "person" to "vehicle" for the same image without retraining or re-segmenting, changing only the text prompt.
  • Setting $\sigma$ small spends nearly all bits on the ROI; setting it near 1 approaches whole-image quality, giving one knob for the ROI/non-ROI trade-off.
  • Because the uniform-mask case still beats the anchors, the latent priors can benefit even users who never supply a meaningful mask.
  • On datasets without ground-truth masks, text prompts still produce usable masks, extending ROI compression to new categories.
  • Better boundary preservation in the ROI and suppression of non-ROI interference translate into higher mAP for object detection and instance segmentation at matched bitrate.

Reading between the lines

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

  • If text-mask quality is the bottleneck, pairing the same codec with a stronger open-vocabulary segmentation model should push text-driven results closer to the ground-truth-mask curves.
  • The uniform-mask result suggests the latent RDO prior may be the dominant contributor; a clean test would ablate the IG sub-module while keeping the mask channel constant and measuring the gain.
  • The same latent-space prior fusion is a candidate for video ROI coding, where object prompts could define preservation targets per frame.
  • The reported mAP above the source image is likely a de-emphasis effect: aggressive non-ROI compression removes distracting context; a background-blur or background-masking baseline could test whether compression is needed at all.
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

4 major / 4 minor

Summary. The paper proposes a customizable ROI-based deep image compression paradigm with three main components: a Text-controlled Mask Acquisition (TMA) module that derives ROI masks from a user text prompt via an Lseg/CLIP-based similarity map and an adjustable binarization step; a Customizable Value Assign (CVA) mechanism that encodes a user-chosen quality trade-off factor σ into the non-ROI mask values and the weighted distortion loss; and a Latent Mask Attention (LMA) module that fuses a latent spatial prior from the mask with a latent RDO prior (importance map) to guide the latent representation. The method is trained on COCO with category annotations and tested on COCO validation and Kodak. The quantitative rate-distortion (RD) and machine-vision results in Fig. 5 are obtained by replacing the similarity map in Eq. (2) with ground-truth COCO segmentation masks, while text-driven masks are demonstrated only qualitatively on a few Kodak images and with IoU on two examples in Fig. 3. The paper claims state-of-the-art ROI-PSNR, PSNR, and detection/segmentation mAP even when the same GT mask is supplied to all methods, and argues that the σ knob lets users manage the ROI/non-ROI quality trade-off.

Significance. If the full pipeline works as claimed, the paper addresses a genuinely useful gap: ROI-based deep image compression in which end users can define the ROI by a text prompt and control the ROI/non-ROI quality balance via a single parameter. The component ablations (MR, IG, fusion, η, σ) are informative and give some independent support for the architectural choices. The promise of public code and the zero-shot demonstrations on out-of-vocabulary text prompts are also strengths. However, because every quantitative RD and mAP curve substitutes GT masks for the proposed text-derived masks, the central claim that text-customizable ROI compression achieves the reported gains is not yet supported. The significance is therefore conditional on closing that evaluation gap with end-to-end text-mask experiments.

major comments (4)
  1. [§IV-A, Eq. (2), Fig. 5] The quantitative backbone of the paper is built on oracle masks: the text states 'we replace the p in Eq. (2) with the GT segmentation results while conducting our method.' Consequently, the PSNR, ROI-PSNR, and mAP curves in Fig. 5 measure the compression network with near-perfect masks, not the proposed text-controlled mask acquisition pipeline. Since the main contribution is text-customizable ROI selection, the results do not yet support the claim that users can achieve these gains by typing a prompt. Please add an end-to-end evaluation on COCO validation in which the TMA module (SG+AB) produces masks from text prompts (e.g., the COCO category names) and report mean mask IoU, PSNR/ROI-PSNR vs BPP, and detection/segmentation mAP vs BPP, overlaid on the GT-mask curves. This is the minimum evidence needed to connect the system-level claim to the experiments.
  2. [§III-B, Fig. 3, and §IV-D] Mask-quality evidence for TMA is limited to two images with IoU reported for a few η values, and no average over the dataset is given. Moreover, the choice η=0.85 is made by sweeping η on the COCO validation set, which is the same set on which the final RD and mAP numbers are reported; this is a form of tuning on the test set. Please report average mask IoU and boundary metrics over a hold-out split (or cross-validation), compare against the Lseg baseline without the AB sub-module, and show how end-to-end RD/mAP varies with η over the hold-out split. Without this, the reader cannot judge whether the qualitative examples in Figs. 1, 3, and 6 are representative.
  3. [§III-D] The training protocol states that stages 2 and 3 use COCO images 'together with the category annotations,' but it does not specify whether the mask m fed to the compression loss is the GT annotation mask or the output of the TMA module. If GT masks are used in training, there is a train/test distribution shift because inference uses Lseg-derived masks, and the decision in §IV-A to substitute GT masks at evaluation time is then a consequence of this mismatch. Please clarify the training mask source, and if GT masks are used, either fine-tune with on-the-fly TMA masks or quantify robustness by perturbing GT masks (e.g., dilation/erosion, random omissions) and measuring the resulting change in RD and mAP. This will bound the impact of TMA mask errors on the headline results.
  4. [§III-B Eq. (4) and §IV-D Fig. 11] The σ-based trade-off is enforced by construction: Eq. (2) assigns non-ROI mask values σ and Eq. (4) multiplies the distortion at each pixel by U(m), so increasing σ directly reduces the loss weight on non-ROI pixels. The trends in Fig. 11 are therefore a direct reflection of the loss definition rather than an emergent property of the learned system. To substantiate the claim of 'effective management' of the quality trade-off, report rate-distortion curves for several σ values at matched total BPP (e.g., by adjusting λ per σ) and report the measured bit allocation to ROI and non-ROI. This would demonstrate that σ is a meaningful user control beyond its explicit role in the objective.
minor comments (4)
  1. [Fig. 5 and Fig. 11] The RD and mAP curves do not include error bars or confidence intervals, and no tabulated mean values are provided; please state the number of seeds and add standard deviations or a table of representative operating points.
  2. [Eq. (2)-(3)] The notation Q(p; 1, σ) in Eq. (2) is not defined; the piecewise definition Q(phw) in Eq. (3) depends on phw and η, but the parameters 1 and σ are not explained there. Please rewrite the definition to clarify the role of each parameter.
  3. [Abstract and §IV-A] The abstract and Section IV-A use the phrase 'uniform mask' to describe the experimental setup; since the masks are COCO GT segmentation masks, please clarify whether 'uniform' means identical across methods or a constant-valued mask.
  4. [Throughout] Minor text issues include 'repeting' in Section III-B, 'tread-off' in Section III-B, 'iamge' in the caption of Fig. 7, and inconsistent spacing in 'CV A' in the abstract; please also make the range of phw and the role of η explicit in Eq. (3).

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: the σ tradeoff is definitionally encoded in Eq. (4), while the LMA module gains rest on independent ablations.

  1. self definitional [Sec. III-B (Eqs. 2-4) and Sec. IV-D 'Setting of σ']
    "By assigning different values of σ in m, users can effectively manage the reconstruction quality trade-off between ROI and non-ROI in ˆx. ... It can be observed that as the value of σ increases, the PSNR across all BPP points increases, while the ROI-PSNR decreases."

    The mask m in Eq. (2) is defined to be σ on the non-ROI, and Eq. (4) sets D = ||(x−x̂) ⊗ U(m)||². Thus σ is not an independently discovered control variable but the literal pixel weight in the training loss: increasing σ mechanically increases the MSE penalty on non-ROI pixels, which by construction raises non-ROI/total PSNR and, under a rate constraint, shifts bits away from the ROI. Reporting the PSNR/ROI-PSNR trend as experimental validation of the CVA mechanism is therefore restating the definition of the loss rather than testing an emergent prediction.

full rationale

The central compression gains attributed to the LMA module are supported by direct ablations (MR vs bilinear downsampling, w/IG vs w/o IG, A+M vs M+M/A+A) against a public baseline, so those claims are not circular. There is no load-bearing self-citation or imported uniqueness argument. The most serious weakness is the Sec. IV-A substitution of COCO GT segmentation masks for the TMA text-derived similarity p when reporting SOTA RD and mAP numbers; that is a real support gap for the text-customization headline, but it is an evidential mismatch rather than a circular derivation, so it does not by itself raise the circularity score. The only genuine circularity is the σ trade-off: because σ enters Eq. (2) directly as the non-ROI mask value and Eq. (4) as the distortion weight, the observed monotonic effect of σ is an algebraic consequence of the loss definition. This affects one of the three claimed contributions but leaves the others with independent support, so the overall circularity is partial rather than total.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method depends on two adjustable scalars, eta tuned on the evaluation split and sigma chosen by the user, and on three domain assumptions about transferred zero-shot masking, GT-mask proxy validity, and the loss construction. No new physical entities are introduced; the latent priors are trained feature maps inside the network.

free parameters (2)
  • Binarization threshold eta = 0.85
    Controls which similarity values become ROI; chosen after experiments on the COCO validation split (Sec. IV-D), which is also the main evaluation split.
  • Quality trade-off factor sigma = User-set in [0,1]; default 0.01
    Sets non-ROI mask value and the weight in the distortion loss (Eqs. 2 and 4); not learned and not compared against an external optimum.
assumptions (3)
  • domain assumption CLIP/Lseg zero-shot pixel-text similarity transfers to arbitrary semantic categories and datasets, such as 'house', 'grass', and 'fence' on Kodak.
    The SG sub-module inherits Lseg's zero-shot ability, but this is validated only by example images (Sec. IV-B), not by a quantitative mask metric.
  • domain assumption COCO ground-truth segmentation masks are a valid proxy for the text-derived masks in the quantitative benchmark.
    Sec. IV-A replaces p in Eq. (2) with GT segmentation results for all anchor comparisons, so the main numbers do not exercise the TMA module.
  • ad hoc to paper The weighted distortion loss of Eq. (4), which reuses the mask m that depends on sigma and eta, correctly encodes user quality preference and produces the desired bit allocation.
    The trade-off response is partly imposed by this loss, not independently measured; bit allocation behavior is shown in Fig. 10.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Customizable ROI-Based Deep Image Compression." pith.science (2026). https://pith.science/paper/TMIXPYXR

@misc{pith2026250700373,
  author       = {Pith},
  title        = {Pith review of: Customizable ROI-Based Deep Image Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TMIXPYXR}},
  note         = {Machine review of arXiv:2507.00373}
}
read the original abstract

Region of Interest (ROI)-based image compression optimizes bit allocation by prioritizing ROI for higher-quality reconstruction. However, as the users (including human clients and downstream machine tasks) become more diverse, ROI-based image compression needs to be customizable to support various preferences. For example, different users may define distinct ROI or require different quality trade-offs between ROI and non-ROI. Existing ROI-based image compression schemes predefine the ROI, making it unchangeable, and lack effective mechanisms to balance reconstruction quality between ROI and non-ROI. This work proposes a paradigm for customizable ROI-based deep image compression. First, we develop a Text-controlled Mask Acquisition (TMA) module, which allows users to easily customize their ROI for compression by just inputting the corresponding semantic \emph{text}. It makes the encoder controlled by text. Second, we design a Customizable Value Assign (CVA) mechanism, which masks the non-ROI with a changeable extent decided by users instead of a constant one to manage the reconstruction quality trade-off between ROI and non-ROI. Finally, we present a Latent Mask Attention (LMA) module, where the latent spatial prior of the mask and the latent Rate-Distortion Optimization (RDO) prior of the image are extracted and fused in the latent space, and further used to optimize the latent representation of the source image. Experimental results demonstrate that our proposed customizable ROI-based deep image compression paradigm effectively addresses the needs of customization for ROI definition and mask acquisition as well as the reconstruction quality trade-off management between the ROI and non-ROI.

Figures

Figures reproduced from arXiv: 2507.00373 by the authors.

Figure 1
Figure 1. The masks and their corresponding reconstructed images with our pro [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Architecture of the proposed customizable ROI-based deep image compression paradigm. It is achieved in two phases. Phase 1: mask representation and fusion process. Phase 2: ROI-based deep image compression process. convolutional layer and obtains the latent representation of the source image, termed y. Then, an element-wise multiplication operation is applied on y and s. After that, an optimized latent representatio… view at source ↗
Figure 3
Figure 3. Visualization of our masks obtained by assigning different values [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Visualization of the variables in our paradigm. (a) and (b) are the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: RD performance on the COCO validation dataset. (a) is the reconstruction quality result of the entire image (PSNR [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The BPP, PSNR and ROI-PSNR comparisons among (a) the source image, (b) the reconstructed image obtained with the baseline [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 8
Figure 8. Figure 8: RD performance comparisons by setting the entire image as the ROI [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: RD performance comparisons with different [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Visualization of the reconstructed images, their associated masks, latent spatial prior, and attention maps with different settings of [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: RD performance comparisons with different [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]
Figure 12
Figure 12. Figure 12: Visualization result of the sub-module MR and IG ablation studies. (a) is the source image. (b) and (e) are our reconstructed images and the latent [PITH_FULL_IMAGE:figures/full_fig_p010_12.png]
Figure 13
Figure 13. Figure 13: RD performance of the IG ablation on the COCO validation dataset. [PITH_FULL_IMAGE:figures/full_fig_p010_13.png]
Figure 14
Figure 14. Figure 14: The detailed architectures of different fusion methods. [PITH_FULL_IMAGE:figures/full_fig_p011_14.png]
Figure 15
Figure 15. Figure 15: RD performance of different fusion methods on COCO validation [PITH_FULL_IMAGE:figures/full_fig_p011_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

62 extracted references · 62 canonical work pages

  1. [1]

    Salient object detection: A survey,

    A. Borji, M.-M. Cheng, Q. Hou, H. Jiang, and J. Li, “Salient object detection: A survey,” Computational visual media , vol. 5, pp. 117–150,

    work page

  2. [2]

    Multi-scale interactive network for salient object detection,

    Y . Pang, X. Zhao, L. Zhang, and H. Lu, “Multi-scale interactive network for salient object detection,” in 2020 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2020, pp. 9410–9419. 1

    work page 2020

  3. [3]

    Hi- danet: Rgb-d salient object detection via hierarchical depth awareness,

    Z. Wu, G. Allibert, F. Meriaudeau, C. Ma, and C. Demonceaux, “Hi- danet: Rgb-d salient object detection via hierarchical depth awareness,” IEEE Transactions on Image Processing, vol. 32, pp. 2160–2173, 2023. 1

    work page 2023

  4. [4]

    One-shot video object segmentation,

    S. Caelles, K.-K. Maninis, J. Pont-Tuset, L. Leal-Taix ´e, D. Cremers, and L. Van Gool, “One-shot video object segmentation,” in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 5320–5329. 1

    work page 2017

  5. [5]

    Just noticeable difference for deep machine vision,

    J. Jin, X. Zhang, X. Fu, H. Zhang, W. Lin, J. Lou, and Y . Zhao, “Just noticeable difference for deep machine vision,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 6, pp. 3452– 3461, 2021. 1

    work page 2021

  6. [6]

    Target-aware camera placement for large-scale video surveillance,

    H. Wu, Q. Zeng, C. Guo, T. Zhao, and C. Wen Chen, “Target-aware camera placement for large-scale video surveillance,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 12, pp. 13 338–13 348, 2024. 1

    work page 2024

  7. [7]

    Image compression using object-based re- gions of interest,

    S. Han and N. Vasconcelos, “Image compression using object-based re- gions of interest,” in 2006 International Conference on Image Processing (ICIP), 2006, pp. 3097–3100. 1

    work page 2006

  8. [8]

    Roi-based deep image compression with swin transformers,

    B. Li, J. Liang, H. Fu, and J. Han, “Roi-based deep image compression with swin transformers,” in 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2023, pp. 1–5. 1, 2, 3, 6

    work page 2023

Show all 62 references
  1. [9]

    Transformer-based variable-rate image compression with region-of- interest control,

    C.-H. Kao, Y .-C. Weng, Y .-H. Chen, W.-C. Chiu, and W.-H. Peng, “Transformer-based variable-rate image compression with region-of- interest control,” in 2023 IEEE International Conference on Image Processing (ICIP), 2023, pp. 2960–2964. 1, 2, 3, 6, 8

    2023

  2. [10]

    Focussing learned image compression to semantic classes for v2x applications,

    J. L ¨ohdefink, A. B ¨ar, N. M. Schmidt, F. H ¨uger, P. Schlicht, and T. Fin- gscheidt, “Focussing learned image compression to semantic classes for v2x applications,” in 2020 IEEE Intelligent Vehicles Symposium (IV) , 2020, pp. 1641–1648. 1, 2

    2020

  3. [11]

    A new image codec paradigm for human and machine uses,

    S. Chen, J. Jin, L. Meng, W. Lin, Z. Chen, T.-S. Chang, Z. Li, and H. Zhang, “A new image codec paradigm for human and machine uses,” arXiv preprint arXiv:2112.10071 , 2021. 1, 2

    2021 arXiv

  4. [12]

    Variable rate roi image compression optimized for visual quality,

    Y . Ma, Y . Zhai, C. Yang, J. Yang, R. Wang, J. Zhou, K. Li, Y . Chen, and R. Wang, “Variable rate roi image compression optimized for visual quality,” in 2021 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021, pp. 1936–1940. 1, 2

    2021

  5. [13]

    Fast roi-based hevc coding for surveillance videos,

    H. Xue, Y . Zhang, and Y . Wei, “Fast roi-based hevc coding for surveillance videos,” in 2016 19th International Symposium on Wireless Personal Multimedia Communications (WPMC) , 2016, pp. 299–304. 1

    2016

  6. [14]

    Adaptive positive sample selection and dynamic soft label assignment for keypoint detection,

    W. Tang, S. Chen, M. Wang, M. Saad Shakeel, J. Jin, W. Kang, and W. Lin, “Adaptive positive sample selection and dynamic soft label assignment for keypoint detection,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 12, pp. 12 665–12 675, 2024. 1

    2024

  7. [15]

    Semantic lens: Instance- centric semantic alignment for video super-resolution,

    Q. Tang, Y . Zhao, M. Liu, J. Jin, and C. Yao, “Semantic lens: Instance- centric semantic alignment for video super-resolution,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 6, 2024, pp. 5154–5161. 1

    2024

  8. [16]

    Jnd-based perceptual optimization for learned image compression,

    F. Ding, J. Jin, L. Meng, and W. Lin, “Jnd-based perceptual optimization for learned image compression,” arXiv preprint arXiv:2302.13092, 2023. 1

    2023 arXiv

  9. [17]

    Let there be light: Improved traffic surveillance via detail preserving night-to- day transfer,

    L. Fu, H. Yu, F. Juefei-Xu, J. Li, Q. Guo, and S. Wang, “Let there be light: Improved traffic surveillance via detail preserving night-to- day transfer,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 12, pp. 8217–8226, 2022. 2

    2022

  10. [18]

    Jpeg2000 roi coding method with perfect fine-grain accuracy and lossless recovery,

    J. Bartrina-Rapesta, J. Serra-Sagrist `a, F. Aul ´ı-Llin`as, and J. Mu ˜noz G ´omez, “Jpeg2000 roi coding method with perfect fine-grain accuracy and lossless recovery,” in 2009 Conference Record of the Forty-Third Asilomar Conference on Signals, Systems and Computers , 2009, ...

    2009

  11. [19]

    Automatic roi selection for jpeg2000 compression of remote sensing images,

    F. Wang, J. Wei, N. Zheng, S. Du, and B. Gao, “Automatic roi selection for jpeg2000 compression of remote sensing images,” in 2007 International Conference on Semantic Computing (ICSC) , 2007, pp. 615–621. 2

    2007

  12. [20]

    Roi based error concealment of com- pressed object based image using qim data hiding and wavelet trans- form,

    A. Phadikar and S. P. Maity, “Roi based error concealment of com- pressed object based image using qim data hiding and wavelet trans- form,” IEEE Transactions on Consumer Electronics , vol. 56, no. 2, pp. 971–979, 2010. 2

    2010

  13. [21]

    Object-based image coding: A learning- driven revisit,

    Q. Xia, H. Liu, and Z. Ma, “Object-based image coding: A learning- driven revisit,” in 2020 IEEE International Conference on Multimedia and Expo (ICME) , 2020, pp. 1–6. 2, 3

    2020

  14. [22]

    Object-aware image compres- sion with adversarial learning,

    Y . Dul, N. Zhaol, Y . Duan, and C. Han, “Object-aware image compres- sion with adversarial learning,” in 2019 IEEE International Conference on Communications in China (ICCC) , 2019, pp. 804–808. 2

    2019

  15. [23]

    A mathematical theory of communication,

    C. E. Shannon, “A mathematical theory of communication,” The Bell system technical journal , vol. 27, no. 3, pp. 379–423, 1948. 2

    1948

  16. [24]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in 2021 International conference on machine learning (ICML) , 2021, pp. 8748–8763. 2

    2021

  17. [25]

    Asymmetric learned image compression with multi-scale residual block, importance scaling, and post-quantization filtering,

    H. Fu, F. Liang, J. Liang, B. Li, G. Zhang, and J. Han, “Asymmetric learned image compression with multi-scale residual block, importance scaling, and post-quantization filtering,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 33, no. 8, pp. 4309–4321, 2...

    2023

  18. [26]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, L. Bourdev, R. Girshick, J. Hays, P. Perona, D. Ramanan, C. L. Zitnick, and P. Doll ´ar, “Microsoft coco: Common objects in context,” 2015. 3, 6

    2015

  19. [27]

    Kodak lossless true color image suite,

    R. Franzen, “Kodak lossless true color image suite,” 1999. [Online]. Available: http://r0k.us/graphics/kodak 1, 3, 6

    1999

  20. [28]

    End-to-end optimized image compression,

    J. Ball ´e, V . Laparra, and E. P. Simoncelli, “End-to-end optimized image compression,” in 2017 International Conference on Learning Representations (ICLR), 2017, pp. 2961–2987. 3

    2017

  21. [29]

    Variational image compression with a scale hyperprior,

    J. Ball ´e, D. Minnen, S. Singh, S. J. Hwang, and N. Johnston, “Variational image compression with a scale hyperprior,” in 2018 International Conference on Learning Representations (ICLR) , 2018, pp. 4961–5007. 3

    2018

  22. [30]

    Joint autoregressive and hier- archical priors for learned image compression,

    D. Minnen, J. Ball ´e, and G. Toderici, “Joint autoregressive and hier- archical priors for learned image compression,” in 2018 International Conference on Neural Information Processing Systems (NIPS) , 2018, pp. 10 794–10 803. 3

    2018

  23. [31]

    Learned lossless image compression with a hyperprior and discretized gaussian mixture likeli- hoods,

    Z. Cheng, H. Sun, M. Takeuchi, and J. Katto, “Learned lossless image compression with a hyperprior and discretized gaussian mixture likeli- hoods,” in 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , 2020, pp. 2158–2162. 3

    2020

  24. [32]

    Learned image compression with gaussian- laplacian-logistic mixture model and concatenated residual modules,

    H. Fu, F. Liang, J. Lin, B. Li, M. Akbari, J. Liang, G. Zhang, D. Liu, C. Tu, and J. Han, “Learned image compression with gaussian- laplacian-logistic mixture model and concatenated residual modules,” IEEE Transactions on Image Processing, vol. 32, pp. 2063–2076, 2023. 3

    2023

  25. [33]

    Efficient contextformer: Spatio-channel window attention for fast context model- ing in learned image compression,

    A. B. Koyuncu, P. Jia, A. Boev, E. Alshina, and E. Steinbach, “Efficient contextformer: Spatio-channel window attention for fast context model- ing in learned image compression,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 8, pp. 7498–7511, 2024. 3

    2024

  26. [34]

    Transformer-based image compression,

    M. Lu, P. Guo, H. Shi, C. Cao, and Z. Ma, “Transformer-based image compression,” in 2022 Data Compression Conference (DCC) , 2022, pp. 469–469. 3, 6, 8

    2022

  27. [35]

    Webp-jpeg transcoding detection by spotting re-compression artifacts with cnn-vit for processing dual- domain features,

    B. Li, J. Shi, W. Li, and H. Li, “Webp-jpeg transcoding detection by spotting re-compression artifacts with cnn-vit for processing dual- domain features,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 12, pp. 12 535–12 549, 2024. 3

    2024

  28. [36]

    Improved lossy image compres- sion with priming and spatially adaptive bit rates for recurrent networks,

    N. Johnston, D. Vincent, D. Minnen, M. Covell, S. Singh, T. Chinen, S. Jin Hwang, J. Shor, and G. Toderici, “Improved lossy image compres- sion with priming and spatially adaptive bit rates for recurrent networks,” in 2018 IEEE Conference on Computer Vision and Pattern Recogni...

    2018

  29. [37]

    Learning convolutional networks for content-weighted image compression,

    M. Li, W. Zuo, S. Gu, D. Zhao, and D. Zhang, “Learning convolutional networks for content-weighted image compression,” in 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 3214–3223. 3

    2018

  30. [38]

    Auto-weighted layer representation based view synthesis distortion estimation for 3-d video coding,

    J. Jin, X. Zhang, L. Meng, W. Lin, J. Liang, H. Zhang, and Y . Zhao, “Auto-weighted layer representation based view synthesis distortion estimation for 3-d video coding,” IEEE Transactions on Multimedia , vol. 25, pp. 5775–5788, 2022. 3

    2022

  31. [39]

    Gan-based image compression with improved rdo process,

    F. Xia, J. Jin, L. Meng, F. Ding, and H. Zhang, “Gan-based image compression with improved rdo process,” in International Conference on Image and Graphics . Springer, 2023, pp. 361–372. 3

    2023

  32. [40]

    End-to-end optimized roi image compression,

    C. Cai, L. Chen, X. Zhang, and Z. Gao, “End-to-end optimized roi image compression,” IEEE Transactions on Image Processing , vol. 29, pp. 3442–3457, 2019. 3

    2019

  33. [41]

    Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,

    L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille, “Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 40, no. 4, pp. 834–848,

  34. [42]

    Recovering realistic texture in image super-resolution by deep spatial feature transform,

    X. Wang, K. Yu, C. Dong, and C. Change Loy, “Recovering realistic texture in image super-resolution by deep spatial feature transform,” in JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13 2018 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018,...

    2021

  35. [43]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in 2021 IEEE International Conference on Computer Vision (ICCV), 2021, pp. 9992–10 002. 3

    2021

  36. [44]

    Integrating language-derived ap- pearance elements with visual cues in pedestrian detection,

    S. Park, H. Kim, and Y . M. Ro, “Integrating language-derived ap- pearance elements with visual cues in pedestrian detection,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 34, no. 9, pp. 7975–7985, 2024. 3

    2024

  37. [45]

    Jewitt, J

    C. Jewitt, J. Bezemer, and K. O’Halloran, Introducing multimodality . Routledge, 2016. 3

    2016

  38. [46]

    Detecting everything in the open world: Towards universal object detection,

    Z. Wang, Y . Li, X. Chen, S.-N. Lim, A. Torralba, H. Zhao, and S. Wang, “Detecting everything in the open world: Towards universal object detection,” in 2023 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 11 433–11 443. 3

    2023

  39. [47]

    Learning to prompt for open-vocabulary object detection with vision-language model,

    Y . Du, F. Wei, Z. Zhang, M. Shi, Y . Gao, and G. Li, “Learning to prompt for open-vocabulary object detection with vision-language model,” in 2022 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 14 064–14 073. 3

    2022

  40. [48]

    Ecffnet: Effective and consistent feature fusion network for rgb-t salient object detection,

    W. Zhou, Q. Guo, J. Lei, L. Yu, and J.-N. Hwang, “Ecffnet: Effective and consistent feature fusion network for rgb-t salient object detection,” IEEE Transactions on Circuits and Systems for Video Technology , vol. 32, no. 3, pp. 1224–1235, 2022. 3

    2022

  41. [49]

    Language as queries for referring video object segmentation,

    J. Wu, Y . Jiang, P. Sun, Z. Yuan, and P. Luo, “Language as queries for referring video object segmentation,” in 2022 Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 4974–4984. 3

    2022

  42. [50]

    Open- vocabulary object segmentation with diffusion models,

    Z. Li, Q. Zhou, X. Zhang, Y . Zhang, Y . Wang, and W. Xie, “Open- vocabulary object segmentation with diffusion models,” in 2023 Pro- ceedings of the IEEE International Conference on Computer Vision (ICCV), 2023, pp. 7667–7676. 3

    2023

  43. [51]

    Video object segmen- tation: a compressed domain approach,

    R. Babu, K. Ramakrishnan, and S. Srinivasan, “Video object segmen- tation: a compressed domain approach,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 14, no. 4, pp. 462–474, 2004. 3

    2004

  44. [52]

    Atzsl: Defensive zero- shot recognition in the presence of adversaries,

    X. Zhang, S. Gui, J. Jin, Z. Zhu, and Y . Zhao, “Atzsl: Defensive zero- shot recognition in the presence of adversaries,” IEEE Transactions on Multimedia, vol. 26, pp. 15–27, 2023. 3

    2023

  45. [53]

    Clipasso: Semantically-aware object sketching,

    Y . Vinker, E. Pajouheshgar, J. Y . Bo, R. C. Bachmann, A. H. Bermano, D. Cohen-Or, A. Zamir, and A. Shamir, “Clipasso: Semantically-aware object sketching,” ACM Transactions on Graphics (TOG), vol. 41, no. 4, pp. 1–11, 2022. 3

    2022

  46. [54]

    Open-vocabulary object detection via vision and language knowledge distillation,

    X. Gu, T.-Y . Lin, W. Kuo, and Y . Cui, “Open-vocabulary object detection via vision and language knowledge distillation,” in 2021 International Conference on Learning Representations (ICLR) , 2021. 3

    2021

  47. [55]

    Language-driven semantic segmentation,

    B. Li, K. Q. Weinberger, S. J. Belongie, V . Koltun, and R. Ranftl, “Language-driven semantic segmentation,” in 2022 International Con- ference on Learning Representations (ICLR) , 2022. 3, 4

    2022

  48. [56]

    A survey of text similarity approaches,

    W. H. Gomaa and A. A. Fahmy, “A survey of text similarity approaches,” international journal of Computer Applications , vol. 68, no. 13, 2013. 4

    2013

  49. [57]

    Flicker 2w,

    J. Liu, G. Lu, Z. Hu, and D. Xu, “Flicker 2w,” 2019. [Online]. Available: https://github.com/liujiaheng/CompressionData 6

    2019

  50. [58]

    Mlic++: Linear complexity multi-reference entropy modeling for learned image compression,

    W. Jiang and R. Wang, “Mlic++: Linear complexity multi-reference entropy modeling for learned image compression,” in 2023 International Conference on Machine Learning Workshop (ICMLW) , 2023. 6

    2023

  51. [59]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: Towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 39, no. 6, pp. 1137– 1149, 2017. 6

    2017

  52. [60]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r-cnn,” in 2017 IEEE International Conference on Computer Vision (ICCV) , 2017, pp. 2980–2988. 6

    2017

  53. [61]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778. 6

    2016

  54. [2019]

    1 JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12

    2021

Pith tools

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