Pith. sign in

REVIEW 4 major objections 6 minor 58 references

Modification Takes Courage: Seamless Image Stitching via Reference-Driven Inpainting

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims that treating image fusion and border straightening as reference-driven inpainting, trained on pseudo-stitched single photos, removes visible seams in uneven-hue and large-parallax stitching, and that multimodal language…

desk verdict A plausible stitching paper with a novel self-supervised training idea, but the claimed generalization to real parallax scenes is not yet supported. read the letter →

arxiv 2411.10309 v2 pith:I4B7DZ54 submitted 2024-11-15 cs.CV

Ziqi Xie , Xiao Lai , Weidong Zhao , Siqi Jiang , Xianhui Liu , Wenlong Hou This is my paper · ORCID
classification cs.CV
keywords imagestitchingreference-driveninpaintingdiffusionmodelself-supervisedlearningtext-to-imagequalityassessmentMLLM-basedmetricslargeparallax
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

Stitched panoramas often fail where the input photos differ in lighting or viewpoint: a visible seam or a ghosted object marks the boundary. The paper's central claim is that this failure disappears if the fusion and border-straightening steps are treated as one reference-driven inpainting task, with a deliberately large and strongly modified fusion region. To make that training feasible without labeled stitching data, the paper builds pseudo-stitched images from single photos by cutting out stitching-shaped masks and adding color shifts and affine shifts, then fine-tunes a text-to-image diffusion model to restore the original photo. It also argues that multimodal language models can score stitched-image quality automatically. A sympathetic reader would care because the method offers a path to seamless stitching in hard real-world cases and an automated way to measure what previous methods could only judge by small user studies.

What carries the argument

The load-bearing object is the reference-driven inpainting model $\epsilon_\theta(\cdot)$, implemented as a pretrained text-to-image diffusion model fine-tuned on a small number of trainable parameters. Its input concatenates a coarse-rectangled reference image with the target image, a mask marks the missing seam and border regions, and a special text prompt binds the generation to the stitching task; the output keeps the target side and fills the masked side. Training is powered by a pseudo-stitching scheme: the mask distribution of a real unlabeled stitching dataset is cut onto single-view images, with color jitter and random affine transforms simulating the two failure modes, and the model minimizes the standard denoising diffusion objective. The evaluation machinery is a pair of multimodal-language-model prompts, one scoring a single image on seams, brightness transitions, distortion, clarity, and abnormal content, and one comparing two stitched images.

What would settle it

Collect a few hundred real multi-view image pairs covering uneven hue and large parallax, have independent human raters choose between RDIStitcher and each baseline without knowing which is which, and check whether the automated scores agree with those choices; if RDIStitcher does not win a clear majority, or if its automated scores disagree with the raters, the central claims are falsified.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the fusion and rectangling stages of stitching can be collapsed into a reference-conditioned inpainting problem and that this model can be taught without ground-truth stitched pairs. Masks sampled from a real stitching dataset are applied to single-view images, augmented with color jitter to imitate uneven hue and random affine translations to imitate parallax misalignment, and the model is trained to recover the original image from this pseudo-stitched input. At inference, the target image is kept fixed while the model inpaints the seam and border regions using the reference image as guidance, accepting a larger modification area and stronger modification intensity than previous inpainting-based stitching. The claimed payoff is higher content coherence and seamless transitions in uneven-hue and large-parallax scenes, including zero-shot transfer to datasets never seen in training. For evaluation, the paper introduces a multimodal-large-language-model protocol: a single-image quality score covering seams, brightness transitions, distortion, clarity, and abnormal content, and a comparative score that chooses the better of two stitched images.

Load-bearing premise

The load-bearing premise is that a model trained to restore a single photo from pseudo-stitched versions of that same photo—masked, color-shifted, and affine-warped using masks sampled from real stitching data—learns a stitching concept that transfers to genuinely different multi-view pairs.

Editorial extensions

If this is right

  • Stitching pipelines can drop the separate supervised rectangling stage: fusion and border straightening become one inpainting step trained from unlabeled photo collections.
  • Hard cases that previously forced a choice between visible seams and distorted content can be handled by permitting a larger edited region, as long as the reference image stays coherent with the target.
  • The self-supervised concept transfers zero-shot to unseen stitching datasets, so new domains such as aerial, medical, or remote-sensing mosaics could be stitched without collected ground-truth pairs.
  • Stitched-image quality can be assessed automatically at scale with multimodal-language-model scores, replacing small manual user studies in future work.

Reading between the lines

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

  • The two data augmentations could be isolated in ablation: training with color jitter only and with affine shifts only would test whether each simulated failure mode transfers independently to its real counterpart.
  • The same pseudo-stitching recipe should transfer to other generative backbones, since the mask-distribution trick is not tied to the chosen text-to-image model.
  • If the central claim holds, the practical trade-off is that users accept generative edits inside the fusion region; the consistency numbers in the paper measure global similarity, not whether thin structures such as wires or pillars are semantically preserved.
  • The multimodal-language-model metrics, once validated on a larger human study than the 50-image pilot described in the supplement, could become the default evaluation instrument for unsupervised stitching research.
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 / 6 minor

Summary. This paper proposes RDIStitcher, a diffusion-based image stitching method that reformulates fusion and rectangling as reference-driven inpainting. A Stable Diffusion inpainting model is fine-tuned with LoRA using a self-supervised scheme in which single-view images are combined with masks from UDIS-D to create pseudo-stitching pairs; color jitter and affine shifts simulate hue differences and misalignment. The authors also propose two MLLM-based evaluation metrics (SIQS and MICQS) and report experiments on UDIS-D and zero-shot traditional datasets, claiming state-of-the-art performance in uneven-hue and large-parallax scenarios.

Significance. If the claims are borne out, the paper would contribute a new training paradigm for inpainting-based stitching, including what the authors describe as the first unsupervised training method for the rectangling problem, as well as a reusable automated evaluation metric for stitched images. The release of code and the use of five-seed reporting with standard deviations are positive aspects of the work. However, the significance is conditional on two load-bearing premises: that the self-supervised pseudo-pairs teach a concept that transfers to true multi-view stitching, and that the proposed MLLM metrics are a reliable proxy for human perception. The current evidence for both premises is suggestive but not conclusive.

major comments (4)
  1. [Sec. 3.3, Eqs. (3)-(5)] The self-supervised training signal is structurally different from the test-time task. In Eq. (3), both pseudo-views are produced by masking the same single image Isg, so the reference always contains the exact pixels to be recovered. The only simulated difficulties are color jitter (Eq. 4) and a global affine shift (Eq. 5); there is no local parallax, no object occlusion, and no content that exists in only one of the two views. The model can therefore succeed by copying from the reference or by invoking its generative prior, without ever learning to combine two genuinely different observations. The paper's claim of strong generalization capabilities to real uneven-hue and large-parallax pairs (Abstract, Sec. 4.4) rests on the unverified assumption that the Stable Diffusion prior converts this single-image masking task into true two-view stitching. I recommend a quantitative test of this transfer, for example on real pairs with known correspondences or with objects visible in only one view, comparing RDIStitcher against a model trained on actual two-view pseudo-pairs.
  2. [Table 3; Sec. 4.2 and 4.6] The main quantitative evidence for superiority uses the authors' own MLLM metrics. These metrics are validated on only 50 hand-crafted images rated by four volunteers (Supp. Sec. G), and the reported correlations are moderate (SRCC 0.728 for SIQS-Q, 0.634 for SIQS-G; PLCC 0.685 and 0.525). On the UDIS-D test set, the gaps between RDIStitcher and the strongest baseline are 0.05-0.30 points on a 10-point scale, with overlapping standard deviations (e.g., SIQS-Q 9.54±0.07 vs 9.49±0.05; SIQS-G 9.39±0.05 vs 9.33±0.04). This does not support the claim of significant improvement. The paper should report paired statistical tests, effect sizes, and ideally validate the metrics on a larger, more diverse set of stitched images before using them as the primary endpoint.
  3. [Sec. 4.1] The comparison with SRStitcher is unequal: SRStitcher is a frozen, zero-shot method, while RDIStitcher is fine-tuned on UDIS-D training data. The authors acknowledge this (the comparison is somewhat less fair), but the central claim of outperforming the inpainting-based SOTA is drawn from this comparison. Since SRStitcher also uses a smaller modification area, the comparison cannot isolate the effect of modification area or intensity, which is the paper's organizing principle. To make the claim convincing, the authors should fine-tune SRStitcher under equivalent conditions (or a controlled variant with matched modification area) or explicitly restrict the claim to zero-shot transfer.
  4. [Sec. 4.5, Table 4] The consistency evaluation measures PSNR/SSIM between RDIStitcher's output and UDIS's output, not against ground truth or the original input images. High agreement with UDIS is not by itself evidence of content preservation; it may indicate that RDIStitcher makes only small changes relative to a particular baseline. The conclusion that the method preserves content consistency is therefore not supported by these numbers. I suggest evaluating fidelity to the original reference and target images in non-overlapping regions, or using correspondence-based metrics, and reporting the variance across the five seeds for these results as well.
minor comments (6)
  1. [Supp. E.1 and Supp. D.2] There are typos in the supplementary material: 'cnfiguration' should be 'configuration' and 'we don not test' should be 'we do not test'.
  2. [Sec. 3.2] The notation Concat(·) and the exact tensor layout of the model input are not formally defined; please specify the channel dimensions and ordering explicitly rather than relying on Fig. 3.
  3. [Fig. 4] The meaning of the 'Both Good' and 'Both Bad' categories in the stacked bar charts is not defined in the caption; please state how these labels were assigned.
  4. [Sec. 1 and 3.3] The paper claims to be the first unsupervised training method for the rectangling problem, but the related work section does not discuss prior unsupervised rectangling attempts; if none exist, state this explicitly in the related work section.
  5. [Eq. (1)] The symbol I is used both for the input image and for the identity matrix; please use a distinct symbol such as I_3 for the identity matrix to avoid ambiguity.
  6. [Tables 3 and 5] The use of red highlighting to indicate the best result is not accessible in printed or color-blind formats; please add boldface or symbols in addition to color.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation identified; the self-supervised transfer and MLLM-based evaluation are empirically grounded assumptions rather than definitional reductions.

full rationale

The paper's derivation chain is not circular. The self-supervised training in Sec. 3.3 constructs pseudo-stitching pairs from a single image via Eq. 3, but this is a pretext task explicitly stated as a hypothesis, not a prediction equivalent to its inputs. The claimed generalization to real multi-view stitching is an empirical claim tested on held-out and zero-shot datasets; whether the transfer succeeds is a correctness risk, not a circularity. The MLLM-based metrics (Sec. 4.2) are proposed by the authors, but they are calibrated against human scores on the hand-crafted Dquality set (Table 5 and Supplementary Sec. G), and the main comparisons also include CCS from SRStitcher and a real-user vote (Fig. 4), providing external grounding. The use of RDIStitcher-R to form the UDIS+R and UDISplus+R baselines is a comparison-protocol choice made necessary by the lack of unsupervised rectangling baselines, not a definitional identity. Self-citations to the authors' prior SRStitcher work supply components such as coarse rectangling and gradient masks, as well as baselines, but no load-bearing authority or forbidden alternative is imported. The paper's own limitations, such as the small 50-image, four-volunteer metric calibration and the MLLMs' limited grasp of 'artifact' concepts, are acknowledged in Supplementary Secs. F.5 and G, and affect robustness rather than circularity. Overall, no step reduces by construction to its own inputs, so the circularity score is low.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The method rests on a small set of hand-chosen hyperparameters and two domain assumptions about transfer learning and MLLM-based evaluation. No new physical entities are introduced; the paper is an empirical engineering contribution.

free parameters (5)
  • pcj = 0.25
    Probability of color jitter augmentation (Eq. 4), chosen by hand to simulate uneven hue.
  • pat = 0.25
    Probability of affine augmentation (Eq. 5), chosen to simulate parallax misalignment.
  • R = 3
    Radius for Telea inpainting in coarse rectangling (Eq. 6).
  • Kd and Kg = 10 and 15
    Kernel sizes for dilation and Gaussian blur in gradient mask generation (Eq. 7).
  • LoRA rank = 8
    Rank of LoRA adaptation for the U-Net and text encoder, chosen via limited ablation (Table 8).
assumptions (3)
  • domain assumption Pseudo-stitched single-view images teach generalizable stitching concepts
    Sec. 3.3 hypothesizes that restoring a single image from masked and augmented pieces transfers to real multi-view stitching. This is the load-bearing assumption of the training method.
  • domain assumption MLLM scores correlate with human perception of stitching quality
    Sec. 4.2 and G validate on a 50-image, 4-volunteer dataset, which is a weak external check.
  • domain assumption UDIS warping produces sufficiently undistorted aligned images
    Sec. 3.1 states UDIS's W(·) is preferred over UDIS++ because the latter introduces local distortions; this choice affects all inputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modification Takes Courage: Seamless Image Stitching via Reference-Driven Inpainting." pith.science (2026). https://pith.science/paper/I4B7DZ54

@misc{pith2026241110309,
  author       = {Pith},
  title        = {Pith review of: Modification Takes Courage: Seamless Image Stitching via Reference-Driven Inpainting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I4B7DZ54}},
  note         = {Machine review of arXiv:2411.10309}
}
read the original abstract

Current image stitching methods often produce noticeable seams in challenging scenarios such as uneven hue and large parallax. To tackle this problem, we propose the Reference-Driven Inpainting Stitcher (RDIStitcher), which reformulates the image fusion and rectangling as a reference-based inpainting model, incorporating a larger modification fusion area and stronger modification intensity than previous methods. Furthermore, we introduce a self-supervised model training method, which enables the implementation of RDIStitcher without requiring labeled data by fine-tuning a Text-to-Image (T2I) diffusion model. Recognizing difficulties in assessing the quality of stitched images, we present the Multimodal Large Language Models (MLLMs)-based metrics, offering a new perspective on evaluating stitched image quality. Compared to the state-of-the-art (SOTA) method, extensive experiments demonstrate that our method significantly enhances content coherence and seamless transitions in the stitched images. Especially in the zero-shot experiments, our method exhibits strong generalization capabilities. Code: https://github.com/yayoyo66/RDIStitcher

Figures

Figures reproduced from arXiv: 2411.10309 by the authors.

Figure 1
Figure 1. Different solutions to image fusion in image stitching. Our method reformulates the fusion and rectangling tasks as a reference [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. A user experience survey of the recon-based method [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The framework of RDIStitcher. (a) Training. For the sake of clarity in the presentation, the input images and masks are simplified. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Multi-image comparative evaluation results. We decide [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative evaluation results. The upper half of the dotted line displays the results on UDIS-D, and the lower half is the results [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: The framework of RDIStitcher-R. C.2. Ablation Study of RDIStitcher-R RDIStitcher-R operates by fine-tuning the pre-trained T2I model to establish the relationship between the coarse rect￾angling image and the detailed image. Table.6 shows the ablations of USID+RDIStitc…
Figure 7
Figure 7. Figure 7: Qualitative results on the different seeds. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Qualitative results on the different metrics. [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]
Figure 9
Figure 9. Figure 9: Additional qualitative results for the challenging scenarios. [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]
Figure 10
Figure 10. Figure 10: Additional qualitative results on APAPdataset [ [PITH_FULL_IMAGE:figures/full_fig_p017_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 52 canonical work pages

  1. [1]

    A new approach for image stitching technique using dynamic time warping (dtw) algorithm towards scoliosis x-ray diagnosis

    Somaya Adwan, Iqbal Alsaleh, and Rasha Majed. A new approach for image stitching technique using dynamic time warping (dtw) algorithm towards scoliosis x-ray diagnosis. Measurement, 84:32–46, 2016. 3

    work page 2016

  2. [2]

    Stable diffusion 2 inpainting

    Stability AI. Stable diffusion 2 inpainting. https : / / huggingface . co / stabilityai / stable - diffusion-2-inpainting, 2023. 2, 4, 6, 1

    work page 2023

  3. [3]

    Break-a-scene: Extracting multi- ple concepts from a single image

    Omri Avrahami, Kfir Aberman, Ohad Fried, Daniel Cohen- Or, and Dani Lischinski. Break-a-scene: Extracting multi- ple concepts from a single image. In SIGGRAPH Asia 2023 Conference Papers, pages 1–12, 2023. 6

    work page 2023

  4. [4]

    Qwen-vl: A versatile vision-language model for un- derstanding, localization, text reading, and beyond

    Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. Qwen-vl: A versatile vision-language model for un- derstanding, localization, text reading, and beyond. arXiv preprint arXiv:2308.12966, 2023. 6, 3

    arXiv 2023

  5. [5]

    Stitching and sur- face reconstruction from endoscopic image sequences: a re- view of applications and methods

    Tobias Bergen and Thomas Wittenberg. Stitching and sur- face reconstruction from endoscopic image sequences: a re- view of applications and methods. IEEE journal of biomed- ical and health informatics, 20(1):304–321, 2014. 3

    work page 2014

  6. [6]

    Leftrefill: Filling right canvas based on left refer- ence through generalized text-to-image diffusion model

    Chenjie Cao, Yunuo Cai, Qiaole Dong, Yikai Wang, and Yanwei Fu. Leftrefill: Filling right canvas based on left refer- ence through generalized text-to-image diffusion model. In CVPR, pages 7705–7715, 2024. 3, 6

    work page 2024

  7. [7]

    Topiq: A top-down approach from semantics to distortions for image quality assessment

    Chaofeng Chen, Jiadi Mo, Jingwen Hou, Haoning Wu, Liang Liao, Wenxiu Sun, Qiong Yan, and Weisi Lin. Topiq: A top-down approach from semantics to distortions for image quality assessment. IEEE Transactions on Image Processing, 33:2404–2418, 2024. 8

    work page 2024

  8. [8]

    Deep Seam Prediction for Image Stitching Based on Selection Consistency Loss

    Senmao Cheng, Fan Yang, Zhi Chen, Nanjun Yuan, and Wenbing Tao. Deep seam prediction for image stitch- ing based on selection consistency loss. arXiv preprint arXiv:2302.05027, 2023. 2, 3

    work page Pith review arXiv 2023

Show all 58 references
  1. [9]

    Incremental transformer structure enhanced image inpainting with mask- ing positional encoding

    Qiaole Dong, Chenjie Cao, and Yanwei Fu. Incremental transformer structure enhanced image inpainting with mask- ing positional encoding. In CVPR, pages 11358–11368,

  2. [10]

    Con- structing image panoramas using dual-homography warping

    Junhong Gao, Seon Joo Kim, and Michael S Brown. Con- structing image panoramas using dual-homography warping. In CVPR, pages 49–56. IEEE, 2011. 1

    2011

  3. [11]

    Chatglm: A family of large language models from glm-130b to glm-4 all tools

    Team GLM, Aohan Zeng, Bin Xu, Bowen Wang, Chenhui Zhang, Da Yin, Diego Rojas, Guanyu Feng, Hanlin Zhao, Hanyu Lai, et al. Chatglm: A family of large language models from glm-130b to glm-4 all tools. arXiv preprint arXiv:2406.12793, 2024. 6, 3

    2024 arXiv

  4. [12]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. In CVPR, pages 16000–16009, 2022. 4

    2022

  5. [13]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. In NeurIPS, pages 6840–6851,

  6. [14]

    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, 2022. 4, 2

    2022

  7. [15]

    Automatic color image stitch- ing using quaternion rank-1 alignment

    Jiaxue Li and Yicong Zhou. Automatic color image stitch- ing using quaternion rank-1 alignment. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19720–19729, 2022. 2, 3

    2022

  8. [16]

    Parallax-tolerant image stitching based on robust elastic warping

    Jing Li, Zhengming Wang, Shiming Lai, Yongping Zhai, and Maojun Zhang. Parallax-tolerant image stitching based on robust elastic warping. IEEE TMM, 20(7):1672–1687, 2017. 5, 1, 7

    2017

  9. [17]

    Quasi-homography warps in image stitching

    Nan Li, Yifang Xu, and Chao Wang. Quasi-homography warps in image stitching. IEEE TMM , 20(6):1365–1375,

  10. [18]

    Mat: Mask-aware transformer for large hole image in- painting

    Wenbo Li, Zhe Lin, Kun Zhou, Lu Qi, Yi Wang, and Jiaya Jia. Mat: Mask-aware transformer for large hole image in- painting. In CVPR, pages 10758–10768, 2022. 3

    2022

  11. [19]

    Single-perspective warps in natural image stitching

    Tianli Liao and Nan Li. Single-perspective warps in natural image stitching. IEEE TIP, 29:724–735, 2019. 5, 1, 7

    2019

  12. [20]

    Smoothly varying affine stitch- ing

    Wen-Yan Lin, Siying Liu, Yasuyuki Matsushita, Tian-Tsong Ng, and Loong-Fah Cheong. Smoothly varying affine stitch- ing. In CVPR, pages 345–352. IEEE, 2011. 1

    2011

  13. [21]

    Unsupervised deep homography: A fast and robust homography estimation model

    Ty Nguyen, Steven W Chen, Shreyas S Shivakumar, Camillo Jose Taylor, and Vijay Kumar. Unsupervised deep homography: A fast and robust homography estimation model. IEEE Robotics and Automation Letters , 3(3):2346– 2353, 2018. 2, 3

    2018

  14. [22]

    A view-free image stitching network based on global homography

    Lang Nie, Chunyu Lin, Kang Liao, Meiqin Liu, and Yao Zhao. A view-free image stitching network based on global homography. Journal of Visual Communication and Image Representation, 73:102950, 2020. 2, 3

    2020

  15. [23]

    Unsupervised deep image stitching: Reconstructing stitched features to images

    Lang Nie, Chunyu Lin, Kang Liao, Shuaicheng Liu, and Yao Zhao. Unsupervised deep image stitching: Reconstructing stitched features to images. IEEE TIP, 30:6184–6197, 2021. 1, 2, 3, 4, 5, 6, 7

    2021

  16. [24]

    Deep rectangling for image stitching: A learning base- line

    Lang Nie, Chunyu Lin, Kang Liao, Shuaicheng Liu, and Yao Zhao. Deep rectangling for image stitching: A learning base- line. In CVPR, pages 5740–5748, 2022. 3, 6, 2

    2022

  17. [25]

    Learn- ing edge-preserved image stitching from multi-scale deep homography

    Lang Nie, Chunyu Lin, Kang Liao, and Yao Zhao. Learn- ing edge-preserved image stitching from multi-scale deep homography. Neurocomputing, 491:533–543, 2022. 2, 3

    2022

  18. [26]

    Parallax-tolerant unsupervised deep image stitching

    Lang Nie, Chunyu Lin, Kang Liao, Shuaicheng Liu, and Yao Zhao. Parallax-tolerant unsupervised deep image stitching. In ICCV, pages 7399–7408, 2023. 1, 2, 3, 5, 6

    2023

  19. [27]

    Unsupervised learning of visual representations by solving jigsaw puzzles

    Mehdi Noroozi and Paolo Favaro. Unsupervised learning of visual representations by solving jigsaw puzzles. In ECCV, pages 69–84. Springer, 2016. 4

    2016

  20. [28]

    Open source computer vision library, 2015

    OpenCV . Open source computer vision library, 2015. 5

    2015

  21. [29]

    Pytorch: an imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. Pytorch: an imperative style, high-performance deep learning library. In NeurIPS, pages 8026–8037, 2019. 5

    2019

  22. [30]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In CVPR, pages 22500–22510, 2023. 4

    2023

  23. [31]

    Blindly assess image qual- ity in the wild guided by a self-adaptive hyper network

    Shaolin Su, Qingsen Yan, Yu Zhu, Cheng Zhang, Xin Ge, Jinqiu Sun, and Yanning Zhang. Blindly assess image qual- ity in the wild guided by a self-adaptive hyper network. In CVPR, pages 3667–3676, 2020. 2, 6 9

    2020

  24. [32]

    Realfill: Reference-driven generation for authentic image completion

    Luming Tang, Nataniel Ruiz, Qinghao Chu, Yuanzhen Li, Aleksander Holynski, David E Jacobs, Bharath Hariharan, Yael Pritch, Neal Wadhwa, Kfir Aberman, et al. Realfill: Reference-driven generation for authentic image completion. ACM TOG, 43(4):1–12, 2024. 3

    2024

  25. [33]

    An image inpainting technique based on the fast marching method

    Alexandru Telea. An image inpainting technique based on the fast marching method. Journal of graphics tools , 9(1): 23–34, 2004. 5

    2004

  26. [34]

    Drone image stitching using local mesh-based bundle adjustment and shape-preserving transform

    Qi Wan, Jun Chen, Linbo Luo, Wenping Gong, and Long- sheng Wei. Drone image stitching using local mesh-based bundle adjustment and shape-preserving transform. IEEE Transactions on Geoscience and Remote Sensing , 59(8): 7027–7037, 2020. 3

    2020

  27. [35]

    Ex- ploring clip for assessing the look and feel of images

    Jianyi Wang, Kelvin CK Chan, and Chen Change Loy. Ex- ploring clip for assessing the look and feel of images. In AAAI, pages 2555–2563, 2023. 2, 6

    2023

  28. [36]

    A spatial arrangement preservation based stitch- ing method via geographic coordinates of uav for farmland remote sensing image

    Jiaxin Wang, Peng Du, Shuqin Yang, Zhitao Zhang, and Jifeng Ning. A spatial arrangement preservation based stitch- ing method via geographic coordinates of uav for farmland remote sensing image. IEEE Transactions on Geoscience and Remote Sensing, 2024. 3

    2024

  29. [37]

    Sdr: stepwise deep rectangling model for stitched images

    Mengsi Wang, Yuan Mei, Lichun Yang, Bin Tian, and Kai- jun Wu. Sdr: stepwise deep rectangling model for stitched images. The Vis. Comput., pages 1–15, 2024. 3, 6

    2024

  30. [38]

    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, Qiong Yan, Xiongkuo Min, Guang- tao Zhai, and Weisi Lin. Q-align: teaching lmms for visual scoring via discrete text-defined levels. 2024. 8

    2024

  31. [39]

    A comprehensive study of multimodal large language mod- els for image quality assessment

    Tianhe Wu, Kede Ma, Jie Liang, Yujiu Yang, and Lei Zhang. A comprehensive study of multimodal large language mod- els for image quality assessment. In ECCV, 2024. 6, 8

    2024

  32. [40]

    Rectanglinggan: Deep rectangling for stitched image via image inpainting

    Ziqi Xie, Weidong Zhao, Xianhui Liu, Zhen Xu, Jian Zhao, and Gong Gao. Rectanglinggan: Deep rectangling for stitched image via image inpainting. Knowledge-Based Sys- tems, page 112129, 2024. 3, 6

    2024

  33. [41]

    Paint by example: Exemplar-based image editing with diffusion mod- els

    Binxin Yang, Shuyang Gu, Bo Zhang, Ting Zhang, Xuejin Chen, Xiaoyan Sun, Dong Chen, and Fang Wen. Paint by example: Exemplar-based image editing with diffusion mod- els. In CVPR, pages 18381–18391, 2023. 3

    2023

  34. [42]

    As-projective-as-possible image stitching with mov- ing dlt

    Julio Zaragoza, Tat-Jun Chin, Michael S Brown, and David Suter. As-projective-as-possible image stitching with mov- ing dlt. In CVPR, pages 2339–2346, 2013. 5, 1, 7

    2013

  35. [43]

    High-resolution image inpainting with iterative confidence feedback and guided upsampling

    Yu Zeng, Zhe Lin, Jimei Yang, Jianming Zhang, Eli Shecht- man, and Huchuan Lu. High-resolution image inpainting with iterative confidence feedback and guided upsampling. In ECCV, pages 1–17. Springer, 2020. 3

    2020

  36. [44]

    Parallax-tolerant image stitching

    Fan Zhang and Feng Liu. Parallax-tolerant image stitching. In CVPR, pages 3262–3269, 2014. 1

    2014

  37. [45]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In ICCV, pages 3836–3847, 2023. 6

    2023

  38. [46]

    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 Transactions on Image Pro- cessing, 30:3474–3486, 2021. 8

    2021

  39. [47]

    Gpt-4v (ision) as a general- ist evaluator for vision-language tasks

    Xinlu Zhang, Yujie Lu, Weizhi Wang, An Yan, Jun Yan, Lianke Qin, Heng Wang, Xifeng Yan, William Yang Wang, and Linda Ruth Petzold. Gpt-4v (ision) as a general- ist evaluator for vision-language tasks. arXiv preprint arXiv:2311.01361, 2023. 6

    2023 arXiv

  40. [48]

    Multi-lora composition for image generation

    Ming Zhong, Yelong Shen, Shuohang Wang, Yadong Lu, Yizhu Jiao, Siru Ouyang, Donghan Yu, Jiawei Han, and Weizhu Chen. Multi-lora composition for image generation. arXiv preprint arXiv:2402.16843, 2024. 6

    2024 arXiv

  41. [49]

    Recdiffu- sion: Rectangling for image stitching with diffusion models

    Tianhao Zhou, Haipeng Li, Ziyi Wang, Ao Luo, Chen-Lin Zhang, Jiajun Li, Bing Zeng, and Shuaicheng Liu. Recdiffu- sion: Rectangling for image stitching with diffusion models. In CVPR, pages 2692–2701, 2024. 3, 6, 2

    2024

  42. [50]

    Transfill: Reference-guided image inpainting by merging multiple color and spatial transformations

    Yuqian Zhou, Connelly Barnes, Eli Shechtman, and Sohrab Amirghodsi. Transfill: Reference-guided image inpainting by merging multiple color and spatial transformations. In CVPR, pages 2266–2276, 2021. 3

    2021

  43. [51]

    Reconstructing the image stitching pipeline: Integrat- ing fusion and rectangling into a unifiedinpainting model

    Xie Ziqi, Zhao Weidong, Liu Xianhui, Zhao Jian, and Jia Ning. Reconstructing the image stitching pipeline: Integrat- ing fusion and rectangling into a unifiedinpainting model. In NeurIPS, 2024. 1, 2, 3, 5, 6, 7 10 Modification Takes Courage: Seamless Image Stitching via Refere...

    2024

  44. [52]

    The comparisons in- clude UDIS+stable diffusion inpainting model [2] without 1 Table 6

    on the UDIS-D test set. The comparisons in- clude UDIS+stable diffusion inpainting model [2] without 1 Table 6. Ablation study of RDIStitcher-R. The best performer is highlighted by red. Method UDIS+SD2 UDIS+RWCR UDIS+R CCS (%) 85.97 ± 1.33 87.68± 1.17 89.64± 0.96 fine-tuning ...

  45. [53]

    Therefore, the construction of prompts can significantly impact the evaluation performance of the MLLMs

    is relatively insensitive, not displaying a significant score reduction. Therefore, the construction of prompts can significantly impact the evaluation performance of the MLLMs. Concept limitation In the early design, we intended to in- clude artifacts in the evaluation criter...

  46. [54]

    Whether there are seams in the image (2 points). score 2: the image is smooth without obvious boundaries or misalignment; score 1: there are slightly visible boundaries in the image, but overall look well; score 0: there are obvious borders or dislocations in the image, affect...

  47. [55]

    Whether there are brightness transitions in the image (2 points). score 2: the brightness transition of image is smooth; score 1: the light and shade changes in the image are a bit unnatural; score 0: the light and shade changes in the image are very abrupt

  48. [56]

    Whether there are distortions in the image (2 points). score 2: no distortion in the image; score 1: there are a few structural anomalies of straight lines in the image; score 0: there are noticeably distortions, such as distorted pillar, brick, and building construction

  49. [57]

    Whether the image is clear and blurred (2 points). score 2: the image is clear, the details are visible, and there is no blur; score 1: the resolution of the image is good, but slightly blurred; score 0: the image is blurred and the details are not clear

  50. [58]

    Whether the image is natural (2 points). score 2: the image is natural with out abnormal content; score 1: there are some places in the image that is not in harmony with the main content; score 0: There are a lot of abnormal content in the image such as strange texture and non...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.