Pith. sign in

REVIEW 3 major objections 4 minor 54 references

Self-Supervised Deep Depth Denoising

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A self-supervised autoencoder learns to denoise consumer depth maps from raw multi-view captures alone, with no clean depth ever shown, and reports the lowest RMSE and lowest KinectFusion error among filter and learned baselines on…

desk verdict Genuinely self-supervised depth denoising via multi-view photometric consistency, with strong controlled synthetic results, but the real-sensor outperformance claim rests on an unvalidated cross-sensor proxy and needs exact-GT confirmation. read the letter →

arxiv 1909.01193 v2 pith:TGEZPRWJ submitted 2019-09-03 cs.CV

Vladimiros Sterzentsenko , Leonidas Saroglou , Anargyros Chatzitofis , Spyridon Thermos , Nikolaos Zioulis , Alexandros Doumanoglou , Dimitrios Zarpalas , Petros Daras This is my paper · ORCID
classification cs.CV
keywords depthdenoisingself-supervisedlearningmulti-viewsupervisionforwardsplattingdifferentiablerenderingphotometricconsistencyRGB-Dsensors3Dreconstruction
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

This paper tries to establish that depth-map denoising for consumer sensors can be learned without any clean depth data. The proposed fully convolutional autoencoder is trained on raw multi-view RGB-D captures: since color is much cleaner than depth, the network's denoised depth predictions are used to forward-splat color from neighboring views into a target view, and the mismatch with the real target color becomes the training signal. Depth and surface-smoothness penalties regularize this photometric loss, and at inference only a single depth map is needed. On Intel RealSense D415 data the model reports the lowest RMSE (58.95 mm) and the lowest KinectFusion point-to-plane error (24.74 mm) against bilateral, joint-bilateral, rolling-guide, DRR, and DDRNet baselines.

What carries the argument

The central mechanism is differentiable multi-view forward splatting: each source pixel's color is reprojected through its predicted depth into the target view and spread over the four neighboring target pixels with a bilinear weight, a depth-confidence weight $w_d(D,p)=\exp(-D(p)/\sigma_D)$, and a radial-distortion weight $w_r(p)$. Accumulating these contributions from all non-target views and normalizing by the splatted weights produces a synthetic target color image $\hat{I}_t$, whose pixel-wise comparison with the real target image defines the photometric loss $L_{ph}$. Because splatting blends contributions rather than rasterizing, gradients flow to all contributing depth measurements, which implicitly handles occlusion; the photometric term is regularized by a BerHu depth residual $L_{depth}$ and a normal-smoothness surface loss $L_{surface}$.

What would settle it

Re-run the comparison on a static scene measured with an independent high-accuracy reference that does not involve a second consumer depth sensor, such as a laser-scanned or coded structured-light scan aligned to the D415 frame, and check whether the proposed network still beats bilateral filtering and rolling guidance in RMSE; additionally, plot per-distance error curves to see whether the gain is concentrated at ranges where the Kinect v2's bias differs most from the D415's.

Watch

Extended reading notes

Core claim

The paper's central claim is that view synthesis can replace ground truth in depth denoising: a single fully convolutional autoencoder, fed only raw depth, learns to suppress sensor noise because its predicted depths must be good enough to reconstruct the scene's observed color from other viewpoints. The authors implement this with multi-view forward splatting rather than inverse warping, so each source pixel contributes color to a target neighborhood with bilinear and confidence weights, gradients flow to every contributing depth measurement, and occlusions need no explicit depth testing. The self-supervised objective combines a photometric loss (a Charbonnier color term plus SSIM with a Tukey penalty) with a BerHu depth-residual regularizer and a surface-normal smoothness prior; partial convolutions handle the roughly 60% zero values of the RealSense high-accuracy profile. On D415 data the method achieves the lowest depth RMSE and the lowest KinectFusion point-to-plane RMSE among all compared methods, transfers qualitatively to Kinect v2 data without retraining, and runs at about 11 ms per frame.

Load-bearing premise

The quantitative evaluation assumes that depth maps from a Kinect v2 placed next to the D415 are a reliable close-to-ground-truth reference, so that projective alignment and the systematic bias between the two sensors are small enough not to decide the reported error differences.

Editorial extensions

If this is right

  • Training requires only raw, synchronized multi-view RGB-D captures with known poses, so new sensor setups can be covered without building synthetic clean-depth pairs the way the DRR and DDRNet pipelines do.
  • Because inference consumes a single depth map and takes roughly 11 ms per frame on a GTX 1080, the denoiser can be embedded in real-time reconstruction or tracking pipelines.
  • The denoiser transfers to another sensor (Kinect v2) that it was never trained on, while the supervised CNN baselines fail to generalize to D415 data.
  • Denoised depth improves downstream 3D reconstruction: KinectFusion point-to-plane RMSE drops to 24.74 mm, and Poisson reconstructions preserve finer surface detail than those built from filter outputs.

Reading between the lines

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

  • The same forward-splatting supervision could be redirected to depth completion or depth super-resolution rather than denoising, since the loss never requires the depth itself to be clean, only the color.
  • If the Kinect-v2 proxy carries a systematic sensor bias, part of the reported margin may be bias removal rather than noise suppression; evaluating against two unrelated sensors as proxies in the same benchmark would separate those effects.
  • The hand-crafted confidence weights (exponential in depth and a radial FoV model) are a natural candidate for a learned confidence network, which could adapt the supervision to textureless regions and sensor-specific distortion in one step.
  • Multi-view splatting supervision assumes photometric consistency, so strong non-Lambertian surfaces or moving content across views would stress it; per-pixel learned weights or a temporal consistency term are testable extensions.
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

3 major / 4 minor

Summary. The paper proposes a fully convolutional autoencoder for depth-map denoising trained without ground-truth depth. During training, multiple calibrated RGB-D views are used: the network predicts denoised depth for each view, depth-image-based rendering forward-splats colors from source views into each target view, and a photometric loss between the synthesized color and the real target color, combined with depth and surface-normal regularizers, supervises the network. At inference only a single noisy depth map is needed. The authors collect a multi-view Intel RealSense D415 dataset and compare against bilateral, joint bilateral, rolling-guidance filters, DRR, and DDRNet on direct depth metrics, KinectFusion, and Poisson reconstruction, with a supplementary ablation study and InteriorNet experiments.

Significance. If the evaluation were validated, the method would be a practical contribution: it removes the ground-truth bottleneck for depth denoising, is sensor-agnostic in principle, and the ablation gives some evidence that the photometric loss improves on a plain autoencoder. The paper releases code and a new RGB-D corpus, and the self-supervision mechanism, based on forward splatting with depth-confidence weighting, is clearly described. The main risk is the Kinect v2 evaluation proxy, which is load-bearing for the headline claim that the model outperforms both traditional filtering and data-driven methods.

major comments (3)
  1. [Section 4 (Evaluation Methodology), Table 1] The quantitative comparison to BF, JBF, RGF, DRR, and DDRNet rests entirely on treating Kinect v2 depth as "close to ground truth" for D415 data after stereo calibration and projective association. The reported advantages are small (MAE 25.11 mm vs 26.11 mm for BF and 26.60 mm for RGF), and cross-sensor bias, resolution differences, and residual misalignment at depth edges can easily produce errors of the same order. No error bars, confidence intervals, significance tests, or validation of the association accuracy are reported, so the 1 mm MAE margin and even the 14 mm RMSE margin cannot be distinguished from proxy artifacts. The central claim that the model "outperformed both traditional filtering and data-driven methods" in direct denoising therefore needs an evaluation against exact geometry or an independently validated target, with uncertainty quantification.
  2. [Section 4 (Results), Table 1] The normal-angle metrics contradict the unqualified outperformance claim: DRR achieves a lower mean angle error (30.23 vs 32.09 degrees) and higher 10/20/30-degree accuracy at every threshold, and RGF is also better at the 20 and 30 degree thresholds. The paper's own text acknowledges ranking third on surface errors, yet the abstract and conclusion say the model outperformed both traditional filtering and data-driven methods. This claim needs to be qualified to the metrics on which the improvement actually holds, or the normal-angle comparison needs to be addressed with an appropriate aggregate metric.
  3. [Supplementary Table 2 (bottom), Section A.5] The ablation does not support the stated conclusion that depth regularization "aids photometric supervision by constraining it": the P-only setting has RMSE 58.30 mm versus 58.95 mm for the full P+D+N model, and P-only also has lower or comparable MAE than P+D, while the differences among all ablation rows are only a few millimeters. Because the ablation is used to justify the final loss design and the photometric-supervision contribution, the authors should report repeated runs with variance, significance tests, or at least a metric-by-metric discussion that acknowledges the non-monotonic results.
minor comments (4)
  1. [Section 4 (Results)] The text refers to "Table ?? (columns 2-7)" and "Table ?? (last column)"; these references should be replaced with the actual table number.
  2. [Abstract] The word "avalable" in the abstract should be corrected to "available".
  3. [Section 3.1, Eq. (2)] The notation "....pt" for the four neighboring target pixels is confusing and appears malformed; a standard notation such as the four pixel coordinates obtained by flooring and ceiling x and y would be clearer.
  4. [Supplementary Section A.2] The description of the retrained DDRNet-TC/DDRNet-PC models using "splatted depth" as near ground truth should be specified more precisely, including how the splatted depth is computed and how the near-ground-truth quality was assessed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the denoising model is trained from multi-view photometric consistency and geometric priors, and the reported evaluations use external benchmarks or the K2 proxy rather than the paper's own training targets.

full rationale

The paper's derivation chain is self-contained and non-circular. The training objective in Eqs. (6)-(11) supervises the autoencoder by forward-splatting color information from other views and computing photometric, depth, and surface losses; ground-truth depth is not used as supervision. The denoised depth map is not defined in terms of the evaluation metric, and no fitted parameter is renamed as a prediction. The quantitative evaluation uses Kinect v2 depth as a close-to-ground-truth proxy for D415 depth, which is an external measurement source rather than an input to the training loss; any weakness in this proxy is a benchmark-validity concern, not circularity. The self-citations for capture ([44]) and multi-sensor calibration ([38]) support experimental infrastructure rather than the core denoising claim. The ablation study and the InteriorNet evaluation provide additional independent evidence. Therefore no step in the paper reduces, by construction or by self-citation, to its own inputs.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The method is an empirical deep learning pipeline. The central claim relies on multi-view photometric consistency, accurate calibration, and several hand-chosen weighting functions and loss hyperparameters, not on a derivation from first principles. No new physical entities are introduced.

free parameters (8)
  • loss weight lambda1 = 0.85
    Weight of photometric loss in Eq. 6, chosen by hand and central to training.
  • loss weight lambda2 = 0.1
    Weight of depth regularization in Eq. 6.
  • loss weight lambda3 = 0.05
    Weight of surface normal regularization in Eq. 6.
  • alpha photometric loss mixing = 0.85
    Mixes color and SSIM terms in Eq. 9.
  • sigma_D depth confidence scale = 3
    Depth threshold and scale in exponential confidence weight exp(-D/sigma_D); set to 3 meters.
  • Charbonnier gamma = 0.447
    Robust penalty constant in Eq. 7.
  • Tukey penalty c = 2.2
    Robust penalty parameter in Eq. 8.
  • learning rate = 0.0002
    Adam optimizer learning rate used in training.
assumptions (5)
  • domain assumption Multi-view photometric consistency: the same 3D surface point has the same color in all views.
    Basis of the photometric loss Lph in Eqs. 7-9. The paper acknowledges real scenes violate it and uses robust penalties, but the supervision signal still depends on this assumption.
  • domain assumption Camera poses and intrinsics are known accurately from calibration.
    Eq. 1 reprojects pixels using poses and intrinsics; calibration errors directly corrupt the splatted images and thus the supervision.
  • ad hoc to paper Forward splatting with exponential depth weights implicitly handles occlusions and visibility.
    The paper argues background depths contribute minimal gradients due to exp(-D/sigma_D), but this is an heuristic rather than a proven visibility test.
  • domain assumption Color images are clean and aligned to depth via the infrared stream.
    Self-supervision treats color as the high-quality signal; color misalignment or noise would bias the training signal.
  • ad hoc to paper The generic FoV radial distortion model and exponential confidence weights adequately capture sensor noise characteristics.
    Eq. 3 and the related confidence weights are chosen rather than derived from measured sensor noise; they affect which pixels dominate the splatting.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self-Supervised Deep Depth Denoising." pith.science (2026). https://pith.science/paper/TGEZPRWJ

@misc{pith2026190901193,
  author       = {Pith},
  title        = {Pith review of: Self-Supervised Deep Depth Denoising},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TGEZPRWJ}},
  note         = {Machine review of arXiv:1909.01193}
}
read the original abstract

Depth perception is considered an invaluable source of information for various vision tasks. However, depth maps acquired using consumer-level sensors still suffer from non-negligible noise. This fact has recently motivated researchers to exploit traditional filters, as well as the deep learning paradigm, in order to suppress the aforementioned non-uniform noise, while preserving geometric details. Despite the effort, deep depth denoising is still an open challenge mainly due to the lack of clean data that could be used as ground truth. In this paper, we propose a fully convolutional deep autoencoder that learns to denoise depth maps, surpassing the lack of ground truth data. Specifically, the proposed autoencoder exploits multiple views of the same scene from different points of view in order to learn to suppress noise in a self-supervised end-to-end manner using depth and color information during training, yet only depth during inference. To enforce selfsupervision, we leverage a differentiable rendering technique to exploit photometric supervision, which is further regularized using geometric and surface priors. As the proposed approach relies on raw data acquisition, a large RGB-D corpus is collected using Intel RealSense sensors. Complementary to a quantitative evaluation, we demonstrate the effectiveness of the proposed self-supervised denoising approach on established 3D reconstruction applications. Code is avalable at https://github.com/VCL3D/DeepDepthDenoising

Figures

Figures reproduced from arXiv: 1909.01193 by the authors.

Figure 1
Figure 1. An abstract representation of the proposed method. Our [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Our multi-view forward splatting scheme is illustrated. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Detailed network architecture of the proposed depth denoising method. The network receives raw depth information from all [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Collected training set samples showing the captured con [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Qualitative results using KinectFusion [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results using Poisson reconstruction. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison using K2 data. In this experiment we opt to use an aggregated metric that handles surface and geometry information jointly, point-to￾plane. Instead of relying on the nearest neighbor for dis￾tance computation, we calculate the Least Square Planes…
Figure 9
Figure 9. Figure 9: Denoising results using original DDRNet[ [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Projected depth maps to 3D domain, after denoising [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: Quantitative results of learning-based methods in rendered images from InteriorNet layouts. The first row and last row show the [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]
Figure 12
Figure 12. Figure 12: Reconstruction results of KinectFusion scans. It is worth mentioning that even noisy raw input can be reconstructed into a high [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: Poisson reconstruction sample. BF and JBF lead to low quality reconstruction due their inability to understand the global context [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 14
Figure 14. Figure 14: Poisson reconstruction sample with the sensors placed higher (looking downwards) and slightly further away from the target. [PITH_FULL_IMAGE:figures/full_fig_p017_14.png]
Figure 15
Figure 15. Figure 15: Poisson reconstruction sample using the setup described in Fig. [PITH_FULL_IMAGE:figures/full_fig_p018_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 50 canonical work pages

  1. [1]

    Deep learning for multi-path error removal in ToF sensors

    Gianluca Agresti and Pietro Zanuttigh. Deep learning for multi-path error removal in ToF sensors. In ECCVW, pages 410–426, 2018

    work page 2018

  2. [2]

    Alexiadis, Nikolaos Zioulis, Dimitrios Zarpalas, and Petros Daras

    Dimitrios S. Alexiadis, Nikolaos Zioulis, Dimitrios Zarpalas, and Petros Daras. Fast deformable model-based human per- formance capture and FVV using consumer-grade RGB-D sensors. Pattern Recognition, 79:260–278, 2018

    work page 2018

  3. [3]

    Barron and Jitendra Malik

    Jonathan T. Barron and Jitendra Malik. Intrinsic scene prop- erties from a single RGB-D image. In CVPR, pages 17–24, 2013

    work page 2013

  4. [4]

    Ro- bust intrinsic and extrinsic calibration of RGB-D cameras

    Filippo Basso, Emanuele Menegatti, and Alberto Pretto. Ro- bust intrinsic and extrinsic calibration of RGB-D cameras. IEEE Transactions on Robotics, (99):01–18, 2018

    work page 2018

  5. [5]

    Noise2Self: Blind denoising by self-supervision

    Joshua Batson and Loic Royer. Noise2Self: Blind denoising by self-supervision. In ICML, 2019

    work page 2019

  6. [6]

    Robust optimization for deep regression

    Vasileios Belagiannis, Christian Rupprecht, Gustavo Carneiro, and Nassir Navab. Robust optimization for deep regression. In ICCV, pages 2830–2838, 2015

    work page 2015

  7. [7]

    Fast MRF optimization with application to depth reconstruction

    Qifeng Chen and Vladlen Koltun. Fast MRF optimization with application to depth reconstruction. In CVPR, pages 3914–3921, 2014

    work page 2014

  8. [8]

    LiDAR-Video driving dataset: Learning driving policies effectively

    Yiping Chen, Jingkang Wang, Jonathan Li, Cewu Lu, Zhipeng Luo, Han Xue, and Cheng Wang. LiDAR-Video driving dataset: Learning driving policies effectively. pages 5870–5878, 2018

    work page 2018

Show all 54 references
  1. [9]

    Fast and accurate deep network learning by Exponential Linear Units (ELUs)

    Djork-Arn ´e Clevert, Thomas Unterthiner, and Sepp Hochre- iter. Fast and accurate deep network learning by Exponential Linear Units (ELUs). In ICLR, 2016

    2016

  2. [10]

    BundleFusion

    Angela Dai, Matthias Nießner, Michael Zollh ¨ofer, Shahram Izadi, and Christian Theobalt. BundleFusion. ACM Trans- actions on Graphics, 36(4):1, 2017

    2017

  3. [11]

    Understanding the dif- ficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the dif- ficulty of training deep feedforward neural networks. In Proc. International Conference on Artificial Intelligence and Statistics, pages 249–256, 2010

    2010

  4. [12]

    Learning dynamic guidance for depth image enhancement

    Shuhang Gu, Wangmeng Zuo, Shi Guo, Yunjin Chen, Chongyu Chen, and Lei Zhang. Learning dynamic guidance for depth image enhancement. In CVPR, pages 712–721, 2017

    2017

  5. [13]

    Real-time geometry, albedo, and motion re- construction using a single RGB-D camera

    Kaiwen Guo, Feng Xu, Tao Yu, Xiaoyang Liu, Qionghai Dai, and Yebin Liu. Real-time geometry, albedo, and motion re- construction using a single RGB-D camera. ACM Transac- tions on Graphics, 36(3):32, 2017

    2017

  6. [14]

    Tackling 3D ToF artifacts through learning and the FLAT dataset

    Qi Guo, Iuri Frosio, Orazio Gallo, Todd Zickler, and Jan Kautz. Tackling 3D ToF artifacts through learning and the FLAT dataset. In ECCV, pages 368–383, 2018

    2018

  7. [15]

    Robust image filtering using joint static and dynamic guidance

    Bumsub Ham, Minsu Cho, and Jean Ponce. Robust image filtering using joint static and dynamic guidance. In CVPR, pages 4823–4831, 2015

    2015

  8. [16]

    High qual- ity shape from a single RGB-D image under uncalibrated natural illumination

    Yudeog Han, Joon-Young Lee, and In So Kweon. High qual- ity shape from a single RGB-D image under uncalibrated natural illumination. In ICCV, pages 1617–1624, 2013

    2013

  9. [17]

    Identity mappings in deep residual networks

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Identity mappings in deep residual networks. In ECCV, pages 630–645, 2016

    2016

  10. [18]

    Joint depth and color camera calibration with distortion correc- tion

    Daniel Herrera, Juho Kannala, and Janne Heikkil ¨a. Joint depth and color camera calibration with distortion correc- tion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 34(10):2058–2064, 2012

    2012

  11. [19]

    Spatial transformer networks

    Max Jaderberg, Karen Simonyan, Andrew Zisserman, and Koray Kavukcuoglu. Spatial transformer networks. In NIPS, pages 2017–2025. 2015

    2017

  12. [20]

    Reconstruction-based pair- wise depth dataset for depth image enhancement using CNN

    Junho Jeon and Seungyong Lee. Reconstruction-based pair- wise depth dataset for depth image enhancement using CNN. In ECCV, pages 438–454, 2018

    2018

  13. [21]

    Screened poisson surface reconstruction

    Michael Kazhdan and Hugues Hoppe. Screened poisson surface reconstruction. ACM Transactions on Graphics , 32(3):29, 2013

    2013

  14. [22]

    Intel (R) realsense (TM) stereoscopic depth cameras

    Leonid Keselman, John Iselin Woodfill, Anders Grunnet- Jepsen, and Achintya Bhowmik. Intel (R) realsense (TM) stereoscopic depth cameras. In CVPRW, pages 1267–1276, 2017

    2017

  15. [23]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In ICLR, 2015

    2015

  16. [24]

    Cohen, Dani Lischinski, and Matt Uyttendaele

    Johannes Kopf, Michael F. Cohen, Dani Lischinski, and Matt Uyttendaele. Joint bilateral upsampling. ACM Transactions on Graphics, 26(3), 2007

    2007

  17. [25]

    Noise2V oid - Learning denoising from single noisy images

    Alexander Krull, Tim-Oliver Buchholz, and Florian Jug. Noise2V oid - Learning denoising from single noisy images. In CVPR, pages 2129–2137, 2019

    2019

  18. [26]

    Data- driven depth map refinement via multi-scale sparse represen- tation

    HyeokHyen Kwon, Yu-Wing Tai, and Stephen Lin. Data- driven depth map refinement via multi-scale sparse represen- tation. In CVPR, pages 159–167, 2015

    2015

  19. [27]

    Deeper depth prediction with fully convolutional residual networks

    Iro Laina, Christian Rupprecht, Vasileios Belagiannis, Fed- erico Tombari, and Nassir Navab. Deeper depth prediction with fully convolutional residual networks. In 3DV, pages 239–248, 2016

    2016

  20. [28]

    The adaptive BerHu penalty in robust regression

    Sophie Lambert-Lacroix and Laurent Zwald. The adaptive BerHu penalty in robust regression. Journal of Nonparamet- ric Statistics, 28(3):487–514, 2016

    2016

  21. [29]

    InteriorNet: Mega-scale multi- sensor photo-realistic indoor scenes dataset

    Wenbin Li, Sajad Saeedi, John McCormac, Ronald Clark, Dimos Tzoumanikas, Qing Ye, Yuzhong Huang, Rui Tang, and Stefan Leutenegger. InteriorNet: Mega-scale multi- sensor photo-realistic indoor scenes dataset. InBMVC, 2018

    2018

  22. [30]

    Joint image filtering with deep convolutional net- works

    Yijun Li, Jia-Bin Huang, Narendra Ahuja, and Ming-Hsuan Yang. Joint image filtering with deep convolutional net- works. IEEE Transactions on Pattern Analysis and Machine Intelligence, page 1, 2019

    2019

  23. [31]

    Reda, Kevin J

    Guilin Liu, Fitsum A. Reda, Kevin J. Shih, Ting-Chun Wang, Andrew Tao, and Bryan Catanzaro. Image inpainting for ir- regular holes using partial convolutions. In ECCV, 2018

    2018

  24. [32]

    Depth enhancement via low-rank matrix completion

    Si Lu, Xiaofeng Ren, and Feng Liu. Depth enhancement via low-rank matrix completion. In CVPR, pages 3390–3397, 2014

    2014

  25. [33]

    Kim, Xin Tong, and Diego Gutierrez

    Julio Marco, Quercus Hernandez, Adolfo Mu ˜noz, Yue Dong, Adrian Jarabo, Min H. Kim, Xin Tong, and Diego Gutierrez. DeepToF: off-the-shelf real-time correction of multipath in- terference in time-of-flight imaging. ACM Transactions on Graphics, 36(6):1–12, 2017

    2017

  26. [34]

    Plane fitting and depth variance based upsampling for noisy depth map from 3D-ToF cameras in real-time

    Kazuki Matsumoto, Francois De Sorbier, and Hideo Saito. Plane fitting and depth variance based upsampling for noisy depth map from 3D-ToF cameras in real-time. In ICPRAM. Science and and Technology Publications, 2015. 9

    2015

  27. [35]

    Dictionary learning from incomplete data for efficient image restoration

    Valeriya Naumova and Karin Schnass. Dictionary learning from incomplete data for efficient image restoration. In EU- SIPCO, pages 1425–1429, 2017

    2017

  28. [36]

    Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J

    Richard A. Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J. Davison, Push- meet Kohi, Jamie Shotton, Steve Hodges, and Andrew Fitzgibbon. KinectFusion: Real-time dense surface mapping and tracking. In ISMAR, pages 127–136, 2011

    2011

  29. [37]

    Bruckstein

    Roy Or - El, Guy Rosman, Aaron Wetzler, Ron Kimmel, and Alfred M. Bruckstein. RGBD-fusion: Real-time high precision depth recovery. In CVPR, pages 5407–5416, 2015

    2015

  30. [38]

    Markerless structure-based multi-sensor calibration for free viewpoint video capture

    Alexandros Papachristou, Nikolaos Zioulis, Dimitrios Zarpalas, and Petros Daras. Markerless structure-based multi-sensor calibration for free viewpoint video capture. In Proc. International Conference in Central Europe on Com- puter Graphics, Visualization and Computer Vision ...

    2018

  31. [39]

    Automatic differentiation in PyTorch

    Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Al- ban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in PyTorch. 2017

    2017

  32. [40]

    U- Net: Convolutional networks for biomedical image segmen- tation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U- Net: Convolutional networks for biomedical image segmen- tation. In MICCAI, pages 234–241, 2015

    2015

  33. [41]

    The xbox one system on a chip and Kinect sensor

    John Sell and Patrick O’Connor. The xbox one system on a chip and Kinect sensor. IEEE Micro, 34(2):44–53, 2014

    2014

  34. [42]

    Ju Shen and Sen-ching S. Cheung. Layer depth denois- ing and completion for structured-light RGB-D cameras. In CVPR, pages 1187–1194, 2013

    2013

  35. [43]

    Mutual- structure for joint filtering

    Xiaoyong Shen, Chao Zhou, Li Xu, and Jiaya Jia. Mutual- structure for joint filtering. In ICCV, 2015

    2015

  36. [44]

    A low-cost, flexible and portable volumetric capturing system

    Vladimiros Sterzentsenko, Antonis Karakottas, Alexandros Papachristou, Nikolaos Zioulis, Alexandros Doumanoglou, Dimitrios Zarpalas, and Petros Daras. A low-cost, flexible and portable volumetric capturing system. In SITIS, pages 200–207, 2018

    2018

  37. [45]

    Deqing Sun, Stefan Roth, and Michael J. Black. Secrets of optical flow estimation and their principles. In CVPR, pages 2432–2439, 2010

    2010

  38. [46]

    A precision analysis of camera distortion models

    Zhongwei Tang, Rafael Grompone von Gioi, Pascal Monasse, and Jean-Michel Morel. A precision analysis of camera distortion models. IEEE Transactions on Image Pro- cessing, 26(6):2694–2704, 2017

    2017

  39. [47]

    CNN-SLAM: Real-time dense monocular SLAM with learned depth prediction

    Keisuke Tateno, Federico Tombari, Iro Laina, and Nassir Navab. CNN-SLAM: Real-time dense monocular SLAM with learned depth prediction. In CVPR, pages 6565–6574, 2017

    2017

  40. [48]

    Bilateral filtering for gray and color images

    Carlo Tomasi and Roberto Manduchi. Bilateral filtering for gray and color images. In ICCV, pages 839–846, 1998

    1998

  41. [49]

    Layer-structured 3D scene inference via view synthesis

    Shubham Tulsiani, Richard Tucker, and Noah Snavely. Layer-structured 3D scene inference via view synthesis. In ECCV, pages 302–317, 2018

    2018

  42. [50]

    Real- time shading-based refinement for consumer depth cameras

    Chenglei Wu, Michael Zollh ¨ofer, Matthias Nießner, Marc Stamminger, Shahram Izadi, and Christian Theobalt. Real- time shading-based refinement for consumer depth cameras. ACM Transactions on Graphics, 33(6):1–10, 2014

    2014

  43. [51]

    DDRNet: Depth map denoising and refinement for consumer depth cameras using cascaded CNNs

    Shi Yan, Chenglei Wu, Lizhen Wang, Feng Xu, Liang An, Kaiwen Guo, and Yebin Liu. DDRNet: Depth map denoising and refinement for consumer depth cameras using cascaded CNNs. In ECCV, pages 155–171, 2018

    2018

  44. [52]

    Shading-based shape refinement of RGB-D images

    Lap-Fai Yu, Sai-Kit Yeung, Yu-Wing Tai, and Stephen Lin. Shading-based shape refinement of RGB-D images. In CVPR, pages 1415–1422, 2013

    2013

  45. [53]

    Rolling guidance filter

    Qi Zhang, Xiaoyong Shen, Li Xu, and Jiaya Jia. Rolling guidance filter. In ECCV, pages 815–830, 2014

    2014

  46. [54]

    spraying

    Zhengyou Zhang. A flexible new technique for camera cali- bration. IEEE Transactions on pattern analysis and machine intelligence, 22, 2000. 10 Supplementary A. Introduction In this supplementary material we complement our orig- inal manuscript with additional quantitative and ...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.