Pith. sign in

REVIEW 3 major objections 4 minor 46 references

A single text-prompted natural image with controlled feature density can register a projector-camera pair as accurately as structured light, without the visual disruption.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 05:13 UTC pith:BYXJMMLR

load-bearing objection Solid niche engineering: controlled natural images for procam registration work better than random photos or structured light in the smooth-surface regime they actually test. the 3 major comments →

arxiv 2607.03046 v1 pith:BYXJMMLR submitted 2026-07-03 cs.CV

Text-to-Image Generation for Projector-Camera System Registration

Xinyu Chen , Yuqi Li , Jiabao Li , Pinyan Tang , Chong Wang , Aditi Majumder This is my paper
classification cs.CV
keywords Projector-Camera systemImage GenerationImage RegistrationControlNetORB featuresoptical flowspatial augmented reality
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Projector-camera systems need dense pixel correspondences so that projected content can be warped correctly onto real surfaces. The usual solution projects multiple structured-light patterns that are accurate for machines but meaningless and disruptive to human viewers. This paper shows that a single natural image, generated from a text prompt and deliberately shaped so that its local features (ORB points) are dense and uniformly distributed, can be projected once, captured by the camera, and fed to a transformer network that recovers the correspondence map. The network itself is trained only on synthetic pairs that model geometric warps, surface texture, ambient light and exposure changes. On both synthetic tests and two real setups the resulting end-point error is lower than that of standard optical-flow baselines and of ordinary Google-sourced natural images. A short user study further reports that viewers find the text-generated image less intrusive and more usable than structured light during live geometric recalibration. The practical payoff is that advertising, interactive AR and continuous display correction can keep their visual narrative intact while still obtaining high-precision registration.

Core claim

When a text-to-image diffusion model is conditioned on a blue-noise Gaussian map of ORB-feature locations, the resulting single natural image, together with a transformer registration network trained on a synthetic procam distortion dataset, produces projector-to-camera correspondences whose end-point error is lower than both pre-trained optical-flow models and ordinary natural images, across simulated and real smooth-surface procam configurations.

What carries the argument

Feature-distribution ControlNet: a Stable-Diffusion backbone whose generation is steered by a Gaussian blob map of desired ORB points (sampled by blue-noise placement), paired with a GMFlow-style transformer that predicts dense correspondences from the projected/captured pair.

Load-bearing premise

The hand-crafted synthetic pipeline of random affine, homography and thin-plate-spline warps plus simple multiplicative texture and gamma-exposure models is assumed to be faithful enough that a network trained only on those pairs will generalize to real projector optics and smooth physical surfaces.

What would settle it

Project one of the authors' generated images and a dense structured-light sequence onto a smooth but non-planar real object under ordinary room lighting; if the end-point error of the network prediction exceeds that of the decoded structured-light map by more than a few pixels, the claim that the synthetic training distribution is adequate collapses.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Advertising and interactive AR exhibits can insert a single thematically matching frame for geometric recalibration instead of flashing machine-readable patterns that break immersion.
  • Continuous geometric correction during movie or presentation playback becomes practical without noticeable interruption of the content stream.
  • The same controlled-generation idea can be extended to multi-projector systems and to joint geometric-plus-radiometric compensation once richer physics-based training data are available.
  • A public 4 500-pair synthetic procam dataset becomes a reusable benchmark for future registration networks.

Where Pith is reading between the lines

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

  • Because the feature-density map is an independent control signal, the same generation pipeline could be adapted to produce registration images that also satisfy radiometric or multi-view constraints without retraining the diffusion backbone from scratch.
  • If the smoothness assumption is relaxed by physics-based rendering of high-frequency geometry, the same network architecture may transfer to free-form projection surfaces that currently require multi-shot structured light.
  • Joint end-to-end training of the ControlNet generator and the registration network (instead of the current two-stage pipeline) is a natural next experiment that the authors themselves flag as future work.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper presents a text-to-image pipeline (ControlNet + Stable Diffusion conditioned on blue-noise ORB-Gaussian feature distribution maps) that produces a single natural image from a user text prompt, engineered to contain dense, uniformly distributed spatial features for procam registration. It also introduces a transformer-based registration network (spatial attention + cross-attention, trained on a 4 500-pair synthetic dataset that applies random affine/homography/TPS warps plus multiplicative texture, ambient, gamma and exposure models) that predicts dense projector-to-camera correspondences. Simulation results (Table 1, Fig. 7) and two real fixed-pose scenes (Table 2, Fig. 9) report lower end-point error than GMA/GMFlow baselines and than ordinary Google images; a 19-person user study claims superior naturalness and usability versus structured-light patterns for geometric calibration.

Significance. If the accuracy and generalization claims hold, the work supplies a practical alternative to multi-frame structured light for spatial AR, projection mapping and advertising displays: a single theme-matched natural image that simultaneously supports high-precision registration and avoids visual disruption. Concrete strengths include the released synthetic benchmark, the explicit joint optimization of naturalness and feature density, the demonstrated geometric-calibration application, and the user-study evidence of improved UX. These elements make the contribution useful even if later refinements are needed for non-smooth surfaces.

major comments (3)
  1. [Section 6 / Table 2 / Fig. 9] Section 6, Table 2 and Fig. 9: real-world validation consists of only two fixed-pose scenes (plaid cloth, art painting) whose ground truth is obtained by decoding a sparse 20 imes20 binary-dot matrix. That GT is itself sparse, subject to the same photometric and texture effects the network is meant to overcome, and does not sample dense correspondence fields, pose/lens variation or surface geometry beyond the two objects. Consequently the leap from the synthetic training distribution (Section 4.1) to the claim of “significantly improving registration accuracy across various procam configurations” remains under-supported.
  2. [Section 4.1, Eqs. (3)–(4)] Section 4.1, Eqs. (3)–(4): the synthetic pipeline relies on hand-chosen parameter ranges (γ1∈[1.5,2.5], γ2∈[0.4,0.66], β∈[0.8,1.25], α, etc.) and a simple multiplicative texture model. No quantitative match to real projector-camera optics or BRDF is shown; the Limitation section already concedes the model is inadequate for non-smooth surfaces. Because the registration network is trained exclusively on this distribution, the reported real-world superiority over GMA/GMFlow (Table 2) rests on an untested fidelity assumption that is load-bearing for the central claim.
  3. [Tables 1–2] Tables 1–2: end-point errors are reported without error bars, confidence intervals or statistical tests (apart from the Spearman analysis of feature count in Table 3). Given the large standard deviations of the baselines and the modest number of real scenes, the strength of the “significantly improving” language cannot be assessed from the published numbers alone.
minor comments (4)
  1. [Acknowledgments] Acknowledgments contain an incomplete placeholder sentence (“To Robert, for the bagels…”).
  2. [Section 3.1] Eq. (1) and surrounding text: the choice of ORB is justified only by a prior citation; a short ablation against SIFT or learned descriptors would clarify whether the control signal is optimal.
  3. [Figures 1, 4, 7] Figure captions (esp. Fig. 1, Fig. 4, Fig. 7) mix incomplete sentences and inconsistent notation for the color-coded flow visualization; a single legend would improve readability.
  4. [Section 3 / Fig. 4] No quantitative image-quality metrics (FID, CLIP score, etc.) are given for the generated images, making the naturalness–feature-richness trade-off hard to judge independently of registration error.

Circularity Check

0 steps flagged

No significant circularity: empirical ML pipeline with independent synthetic training and external baselines; minor self-citations are non-load-bearing.

full rationale

The paper's central claims (text-controlled feature-rich image generation via ControlNet/Stable Diffusion + transformer registration network trained on synthetic affine/homography/TPS + photometric overlays) are evaluated by end-point error against held-out synthetic GT (different random parameters) and sparse real-dot GT on two physical scenes, outperforming external pre-trained GMA/GMFlow models. No equation or result reduces by construction to a fitted input (Eqs. 1-4 define the synthesis process used for training data, not the reported test metrics). Self-citations to overlapping-author prior procam work ([16],[18]) appear only in related-work motivation and are not used to force uniqueness or close any derivation loop. Standard train/test on synthetic data plus real hardware checks constitute independent empirical content, not circularity. Score 1 only for the presence of non-essential self-references.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 1 invented entities

The central accuracy claim rests on (a) the adequacy of a simple image-based distortion model for real procam optics, (b) a handful of hand-chosen ranges for gamma/exposure/ambient parameters, and (c) the assumption that ORB-feature density is a sufficient proxy for registration quality. No new physical entities are postulated; free parameters are the usual ML hyper-parameters plus the synthesis ranges.

free parameters (5)
  • γ1 (projector gamma) = [1.5, 2.5]
    Randomly sampled in [1.5, 2.5] during synthesis; range chosen by hand, not derived.
  • γ2 (camera gamma) = [0.4, 0.66]
    Randomly sampled in [0.4, 0.66]; hand-chosen interval.
  • β (exposure scale) = [0.8, 1.25]
    Randomly sampled in [0.8, 1.25]; hand-chosen.
  • feature sampling density & Gaussian scale range = 0.02 / [31,111]
    Density 0.02 and scale [31, 111] used for generation examples; chosen to produce 'high density and uniformity'.
  • α (ambient intensity)
    Additive ambient term in the reflectance equation; range not fully specified, treated as free.
axioms (4)
  • domain assumption Geometric distortion between projector and camera on a smooth surface can be adequately modeled by a composition of affine, homography and thin-plate-spline warps.
    Stated in Section 4.1; underpins the entire synthetic training set.
  • domain assumption Surface reflectance and ambient light act as a simple per-pixel multiplicative overlay plus additive constant after gamma correction.
    Equation (3); ignores inter-reflections, BRDF variation and projector non-linearity beyond gamma.
  • ad hoc to paper ORB feature density and spatial uniformity are sufficient control signals for generating images that maximize registration accuracy.
    Core design choice of Section 3; justified only by qualitative comparison and a small Spearman ablation.
  • domain assumption A network trained solely on the synthetic pairs generalizes to real hardware under the same smooth-surface regime.
    Implicit throughout Sections 5-6; acknowledged as a limitation in Section 7.
invented entities (1)
  • ORB-Gaussian feature distribution map used as ControlNet condition no independent evidence
    purpose: To force the text-to-image model to place rich, evenly spaced features that improve subsequent dense correspondence.
    Novel control signal for this application; no independent physical existence outside the generation pipeline.

pith-pipeline@v1.1.0-grok45 · 19735 in / 3037 out tokens · 29595 ms · 2026-07-12T05:13:36.376295+00:00 · methodology

0 comments
read the original abstract

Establishing correspondence between projector and camera images in a procam (projector + camera) system is essential for achieving high-resolution pixel matching, referred to as procam registration. The highest accuracy is typically obtained using structured light patterns (e.g., stripes or blobs). However, these methods are often inefficient and lack meaningful information for human viewers. Although some have explored the use of natural images, these often fail to provide a sufficient distribution of features to achieve comparable accuracy. Additionally, existing methods struggle to cope with environmental factors such as surface textures and variations in brightness due to ambient light or changes in camera exposure. To address these limitations, we propose a method based on deep neural networks. Our approach aims to generate a single natural image from text-based prompts that not only appears realistic but also possesses rich spatial features to enhance registration accuracy in procam applications. We have developed a deep neural network trained on a synthesized dataset that simulates potential geometric and photometric distortions encountered in a procam system illuminating a relatively smooth object (see Figure 1). Our trained network predicts the correspondence between projector and camera images, significantly improving registration accuracy across various procam configurations. By jointly considering the naturalness and feature richness of the projector images, our method minimizes visual disruptions in projected content without sacrificing precision. A user study confirms that our technique enhances perceived naturalness and usability compared to existing methods, validating its practical utility in real-world applications.

Figures

Figures reproduced from arXiv: 2607.03046 by Aditi Majumder, Chong Wang, Jiabao Li, Pinyan Tang, Xinyu Chen, Yuqi Li.

Figure 1
Figure 1. Figure 1: Text-generated non-pattern projected images are used for geometric calibration of the projector-camera system while [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Random Radius(Scale) of the Gaussian Points Randomly Generated Blue Noise σ∈[5,15) σ∈[10,15) σ∈[2,18) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: The generation process of Gaussian distribution [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The control network architecture for natural image generation. The network consists of two parallel branches: one [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The figure displays images generated using our proposed non-pattern projector image generation technique, alongside [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The figure illustrates the overall framework of the data synthesis process. On the left, randomly selected natural [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Framework of the Deep Registration Network: The [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The warped results obtained by applying the predicted correspondence from different models to generated projector [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualize the errors of the sample according to [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Projecting images onto different textured objects and comparing the predicted correspondence results of the [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Geometric calibration results after projecting images onto surfaces with different textures. We compared five models [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Results of paired t-test comparing user ratings between proposed and traditional structured light methods. Our [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

46 extracted references · 4 linked inside Pith

  1. [1]

    Mostafa Abdou, Artur Kulmizev, Daniel Hershcovich, Stella Frank, Ellie Pavlick, and Anders Søgaard. 2021. Can language models encode perceptual structure without grounding? a case study in color.arXiv preprint arXiv:2109.06129(2021)

    Pith/arXiv arXiv 2021

  2. [2]

    Ali A Al-Temeemy and Shahad A Al-Saqal. 2021. Laser-based structured light technique for 3D reconstruction using extreme laser stripes extraction method with global information extraction.Optics & Laser Technology138 (2021), 106897

    2021

  3. [3]

    Anson. 2023. NeoplasticismMix. https://www.liblib.art/modelinfo/ 4fafc0b7fb3c46988115e72fb05f145b

    2023

  4. [4]

    Tianle Cheng, Long Qin, Yulong Li, Jialong Hou, and Changyan Xiao. 2024. An adaptive multi-exposure scheme for the structured light profilometry of highly reflective surfaces using complementary binary gray code.IEEE Transactions on Instrumentation and Measurement(2024)

    2024

  5. [5]

    Paul E Debevec, Camillo J Taylor, and Jitendra Malik. 2023. Modeling and rendering architecture from photographs: A hybrid geometry-and image-based approach. InSeminal Graphics Papers: Pushing the Boundaries, Volume 2. 465–474

    2023

  6. [6]

    Xin Dong, Haibin Ling, and Bingyao Huang. 2023. Adaptive Color Structured Light for Calibration and Shape Reconstruction. In2023 IEEE International Sym- posium on Mixed and Augmented Reality (ISMAR). IEEE, 1240–1249

    2023

  7. [7]

    Yotam Erel, Daisuke Iwai, and Amit H Bermano. 2023. Neural projection mapping using reflectance fields.IEEE Transactions on Visualization and Computer Graphics (2023)

    2023

  8. [8]

    Naoki Hashimoto, Ryo Koizumi, and Daisuke Kobayashi. 2017. Dynamic projec- tion mapping with a single IR camera.International Journal of Computer Games Technology2017, 1 (2017), 4936285

    2017

  9. [9]

    Naoki Hashimoto and Kazuma Yoshimura. 2021. Radiometric compensation for non-rigid surfaces by continuously estimating inter-pixel correspondence.The Visual Computer37, 1 (2021), 175–187

    2021

  10. [10]

    Bingyao Huang and Haibin Ling. 2019. Compennet++: End-to-end full projec- tor compensation. InProceedings of the IEEE/CVF International Conference on Computer Vision. 7165–7174

    2019

  11. [11]

    Muhammad Twaha Ibrahim and Aditi Majumder. 2023. Spatially Augmented Reality on Non-rigid Dynamic Surfaces. In2023 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW). IEEE, 1009–1010

    2023

  12. [12]

    Shihao Jiang, Dylan Campbell, Yao Lu, Hongdong Li, and Richard Hartley. 2021. Learning to estimate hidden motions with global motion aggregation. InProceed- ings of the IEEE/CVF international conference on computer vision. 9772–9781

    2021

  13. [13]

    Yuta Kageyama, Daisuke Iwai, and Kosuke Sato. 2024. Efficient distortion-free neural projector deblurring in dynamic projection mapping.IEEE Transactions on Visualization and Computer Graphics(2024)

    2024

  14. [14]

    Nasim Kayhan and Shervan Fekri-Ershad. 2021. Content based image retrieval based on weighted fusion of texture and color features derived from modified local binary patterns and local neighborhood difference patterns.Multimedia Tools and Applications80, 21 (2021), 32763–32790. , Xinyu Chen, Yuqi Li, Jiabao Li, Pinyan Tang, Chong Wang, and Aditi Majumder

    2021

  15. [15]

    Sungtae Kim, Jurim Jeon, Yangjin Kim, Naohiko Sugita, and Mamoru Mitsuishi

  16. [16]

    Design and assessment of phase-shifting algorithms in optical interfer- ometer.International Journal of Precision Engineering and Manufacturing-Green Technology10, 2 (2023), 611–634

    2023

  17. [17]

    Yuqi Li, Niguang Bao, Qingshu Yuan, and Dongming Lu. 2012. Real-time contin- uous geometric calibration for projector-camera system under ambient illumina- tion. In2012 International Conference on Virtual Reality and Visualization. IEEE, 7–12

    2012

  18. [18]

    Yuqi Li, Dongming Lu, and Lei Zhao. 2013. Robust estimation of spectral re- flectance by a projector-camera system.Chinese Optics Letters11, 11 (2013), 113301

    2013

  19. [19]

    Yuqi Li, Wenting Yin, Jiabao Li, and Xijiong Xie. 2023. Physics-based efficient full projector compensation using only natural images.IEEE Transactions on Visualization and Computer Graphics(2023)

    2023

  20. [20]

    Zhengqi Li and Noah Snavely. 2018. Megadepth: Learning single-view depth prediction from internet photos. InProceedings of the IEEE conference on computer vision and pattern recognition. 2041–2050

    2018

  21. [21]

    Jia-Wei Liao, Winston Wang, Tzu-Sian Wang, Li-Xuan Peng, Cheng-Fu Chou, and Jun-Cheng Chen. 2024. Diffusion-based Aesthetic QR Code Generation via Scanning-Robust Perceptual Guidance.arXiv preprint arXiv:2403.15878(2024)

    Pith/arXiv arXiv 2024

  22. [22]

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. 2021. Swin transformer: Hierarchical vision transformer us- ing shifted windows. InProceedings of the IEEE/CVF international conference on computer vision. 10012–10022

    2021

  23. [23]

    Xingyu Lu, Lei Sun, Diyang Gu, Zhijie Xu, and Kaiwei Wang. 2024. SGE: Struc- tured Light System Based on Gray Code with an Event Camera.arXiv preprint arXiv:2403.07326(2024)

    Pith/arXiv arXiv 2024

  24. [24]

    Yiyang Ma, Huan Yang, Wenjing Wang, Jianlong Fu, and Jiaying Liu. 2023. Uni- fied multi-modal latent diffusion for joint subject and text conditional image generation.arXiv preprint arXiv:2303.09319(2023)

    Pith/arXiv arXiv 2023

  25. [25]

    Sagi Marcovich, Tuvi Etzion, and Eitan Yaakobi. 2021. Balanced de Bruijn se- quences. In2021 IEEE International Symposium on Information Theory (ISIT). IEEE, 1528–1533

    2021

  26. [26]

    Hao Meng, Qi Xue, and Bang Liu. 2022. Research of color-coded structured light based on convolutional code. In2nd International Conference on Laser, Optics and Optoelectronic Technology (LOPET 2022), Vol. 12343. SPIE, 128–134

    2022

  27. [27]

    Yue Ming, Xuyang Meng, Chunxiao Fan, and Hui Yu. 2021. Deep learning for monocular depth estimation: A review.Neurocomputing438 (2021), 14–33

    2021

  28. [28]

    Megumi Nakao, Fei Tong, Mitsuhiro Nakamura, and Tetsuya Matsuda. 2021. Image-to-graph convolutional network for deformable shape reconstruction from a single projection image. InMedical Image Computing and Computer As- sisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 2021, Proceedings, Part IV 24....

    2021

  29. [29]

    Yuji Oyamada and Hideo Saito. 2007. Focal pre-correction of projected image for deblurring screen image. In2007 IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 1–8

    2007

  30. [30]

    Il-Jun Pak, Chol-Su Kim, Jong-Chol Kang, and Ju-Ryong Son. 2021. Verification of phase measurement error sources in phase-shifting interferometry with four step phase-shifting algorithms.Applied Optics60, 13 (2021), 3856–3864

    2021

  31. [31]

    Hanhoon Park, Moon-Hyun Lee, Byung-Kuk Seo, Yoonjong Jin, and Jong-Il Park

  32. [32]

    InProceedings of the 2nd International Conference on Virtual Reality(Beijing, China)(ICVR’07)

    Content adaptive embedding of complementary patterns for nonintru- sive direct-projected augmented reality. InProceedings of the 2nd International Conference on Virtual Reality(Beijing, China)(ICVR’07). Springer-Verlag, Berlin, Heidelberg, 132–141

  33. [33]

    Parinya Punpongsanon, Daisuke Iwai, and Kosuke Sato. 2015. Softar: Visually manipulating haptic softness perception in spatial augmented reality.IEEE transactions on visualization and computer graphics21, 11 (2015), 1279–1288

    2015

  34. [34]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 10684–10695

    2022

  35. [35]

    Ethan Rublee, Vincent Rabaud, Kurt Konolige, and Gary Bradski. 2011. ORB: An efficient alternative to SIFT or SURF. In2011 International conference on computer vision. Ieee, 2564–2571

    2011

  36. [36]

    Toru Takahashi, Norihito Numa, Takafumi Aoki, and Satoshi Kondo. 2008. A geometric correction method for projected images using sift feature points. In Proceedings of the 5th ACM/IEEE International Workshop on Projector camera systems. 1–2

    2008

  37. [37]

    Yuwei Wang, Lu Liu, Jun Wu, Xiangcheng Chen, and Yajun Wang. 2020. Spatial binary coding method for stripe-wise phase unwrapping.Applied optics59, 14 (2020), 4279–4285

    2020

  38. [38]

    Zhenzhou Wang, Qi Zhou, and YongCan Shuang. 2020. Three-dimensional reconstruction with single-shot structured light dot pattern and analytic solutions. Measurement151 (2020), 107114

    2020

  39. [39]

    Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, and Dacheng Tao. 2022. GMFlow: Learning Optical Flow via Global Matching. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 8121–8130

    2022

  40. [40]

    Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming-Hsuan Yang. 2023. Diffusion models: A comprehensive survey of methods and applications.Comput. Surveys56, 4 (2023), 1–39

    2023

  41. [41]

    Jing Ye and Changhe Zhou. 2021. Time-resolved coded structured light for 3D measurement.Microwave and Optical Technology Letters63, 1 (2021), 5–12

    2021

  42. [42]

    Jinghui Zeng, Wenxin Ma, Wei Jia, Yucheng Li, Haowen Li, Xuewen Liu, and Mingkui Tan. 2022. Self-unwrapping phase-shifting for fast and accurate 3-D shape measurement.IEEE Transactions on Instrumentation and Measurement71 (2022), 1–12

    2022

  43. [43]

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. 2023. Adding conditional con- trol to text-to-image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision. 3836–3847

    2023

  44. [44]

    Yabing Zheng, Minghui Duan, Zheng Sun, Yi Jin, and Jinjin Zheng. 2022. Pre- shifted phase coding approach for absolute phase retrieval. InEighth Symposium on Novel Photoelectronic Detection Technology and Applications, Vol. 12169. SPIE, 2923–2928

    2022

  45. [45]

    Xiao Zhou, Yu Kang, Tingting Zhang, and Xingang Mou. 2021. M-array based on non-zero maps.IEEE Access9 (2021), 58467–58477

    2021

  46. [46]

    Stefanie Zollmann, Tobias Langlotz, and Oliver Bimber. 2007. Passive-active geometric calibration for view-dependent projections onto arbitrary surfaces. JVRB-Journal of Virtual Reality and Broadcasting4, 6 (2007)

    2007