REVIEW 4 major objections 5 minor 46 references
Consistent Time-of-Flight Depth Denoising via Graph-Informed Geometric Attention
T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash
Pith's one-line read Graph structures around a pixel stay similar across ToF frames even when depth values shift; fusing those graphs rather than depth features removes noise while preserving edges and temporal stability.
desk verdict Genuinely new graph-fusion idea with strong reported results, but the paper's key interpretability claim rests on a symmetric-PSD Laplacian that Eq. (6) does not yield. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is cross-frame graph fusion: intra-frame graphs $W^t$ encode eight-neighbor pixel correlations within a frame, the inter-frame attention graph $W^{t,t-1}$ connects each current-frame pixel to a $7\times7$ neighborhood in the previous frame, and the mapped graph $\hat{W}^{t-1}=W^{t,t-1}(W^{t-1}+I)(W^{t,t-1})^{\top}$ moves the reference graph along two-hop and three-hop paths. The fused graph $\tilde{W}^t=\Phi^{t,t-1}\hat{W}^{t-1}+W^t$ is weighted by a learned per-pixel confidence $\Phi^{t,t-1}$ so that unreliable mappings, for instance at occlusions, are down-weighted. This fused graph defines the graph Laplacian in the MAP smoothness prior, and unrolling the solution produces iterative low-pass graph filters whose weights come from attention-based graph learning, which is named the graph-informed geometric attention (GIGA) module.
What would settle it
An annotated dataset with occlusion masks and known inter-frame correspondences would settle the claim: run GIGA-ToF on occluded and fast-moving regions and compare against a variant that sets the reference graph contribution to zero ($W^{t,t-1}=0$). If fusing the mapped reference graph does not reduce error in exactly those regions, or if the learned confidence $\Phi^{t,t-1}$ does not down-weight mismapped pixels, then graph structures are not self-similar where temporal fusion matters most.
Extended reading notes
Core claim
The central discovery is temporal self-similarity of graph structure: depth values at corresponding pixels vary across frames, but the graph encoding which neighboring pixels are correlated stays similar, so a reference frame's graph can be transported to the current frame and fused with it. The transport is a mapped graph $\hat{W}^{t-1}=W^{t,t-1}(W^{t-1}+I)(W^{t,t-1})^{\top}$, combining two-hop and three-hop paths through the inter-frame attention graph, and the fusion is $\tilde{W}^t=\Phi^{t,t-1}\hat{W}^{t-1}+W^t$, where $\Phi^{t,t-1}$ is a learned diagonal confidence that down-weights unreliable mapping. The fused graph supplies the Laplacian for a graph-smoothness prior, and the likelihood comes from the ToF depth-noise distribution; the resulting MAP problem is unrolled into iterative filters whose kernels are the learned graph edge weights. The unrolled solution acts as a low-pass graph spectral filter, which the paper credits for its robustness to unseen real noise.
Load-bearing premise
The method must reliably carry the previous frame's pixel-correlation graph into the current frame, and the only protection against wrong mapping is a learned per-pixel confidence that was never trained or tested on occlusion labels; if that mapping misfires under occlusion or large motion, the fused graph injects wrong edges and the claimed sharpness and temporal stability would degrade.
Editorial extensions
If this is right
- If graph structures are as motion-invariant as claimed, multi-frame ToF denoising can drop explicit scene-flow alignment: the attention-based graph mapping absorbs the correspondence problem.
- Accuracy and temporal consistency improve in the same fused graph, so the usual tradeoff between per-frame sharpness and temporal stability need not apply to this class of methods.
- The unrolled MAP solution gives the network a graph-spectral interpretation as a low-pass filter, which is the paper's stated reason it transfers to real Kinectv2 noise despite synthetic-only training.
- Only the previous frame is used as reference with a forward-only update, so the same graph-fusion step can be chained over longer sequences; the paper reports the two-frame version and lists extension to more frames as future work.
Reading between the lines
- The paper leaves the confidence matrix $\Phi^{t,t-1}$ unvalidated against occlusion or large-displacement ground truth; an occlusion-annotated benchmark is the natural next test of the motion-invariant graph assumption.
- The same principle could extend beyond ToF: any representation invariant to per-pixel value shifts but sensitive to local geometry, such as surface normals or local shape indices, could be fused across frames instead of depth features.
- Since the mapped-graph formula is a linear transport of edge weights, composing it over multiple reference frames would yield a recurrent graph-state update, potentially handling variable-length video without redesign.
- Because the unrolled step is a graph Laplacian low-pass filter, one could compare the spectral response of fused and single-frame graphs to quantify how much cross-frame information contributes; the paper does not report such an analysis.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes GIGA-ToF, a multi-frame Time-of-Flight depth denoising network that fuses graph structures across frames rather than depth features. The authors define intra-frame graphs on ToF raw data, map the reference frame's graph into the current frame via a cross-frame geometric attention mechanism, and form a MAP problem with a graph Laplacian smoothness prior on the fused graph. The solution is unrolled into iterative graph filters, yielding an interpretable network. Experiments on a self-generated synthetic DVToF dataset report state-of-the-art MAE, AbsRel, δ1, and TEPE, with qualitative generalization results on real Kinect v2 data.
Significance. The core idea of exploiting temporal self-similarity of graph structures instead of depth features is novel and potentially impactful for temporal ToF denoising. The reported improvements are large (at least 37.9% MAE and 13.2% TEPE over the best compared methods), and the claimed cross-dataset generalization to Kinect v2 is practically important. The paper also ships source code and an algorithm unrolling formulation, which are positive for reproducibility. However, the theoretical claims about the MAP objective and graph spectral filtering currently rest on a symmetry/PSD property that the construction does not satisfy, and the empirical claims lack error bars, so the significance can only be assessed after these issues are resolved.
major comments (4)
- [Sec. 4.2, Eq. (6); Sec. 5.2] The fused graph Wtilde^t = Phi^{t,t-1} What^{t-1} + W^t is not symmetric when Phi^{t,t-1} is a spatially varying diagonal matrix, because the transpose is What^{t-1} Phi^{t,t-1} + W^t, which equals Wtilde^t only if Phi commutes with What^{t-1}. Consequently the graph Laplacian Ltilde^t = Dtilde^t - Wtilde^t in Eq. (9) is not symmetric, and the quadratic form x^T Ltilde x is not non-negative in general; for a two-node example with W_12=a and W_21=b, x=(1,2)^T gives x^T L x = -a+2b, which can be negative. The claim in Sec. 5.2 that the graph Laplacian is symmetric positive semi-definite and that the solution is a low-pass graph spectral filter is therefore false. The MAP prior in Eq. (11) and the unrolled iteration in Eq. (14) are not supported by the stated construction. The paper should modify the fusion to preserve symmetry and PSD, e.g., Phi^{1/2} What Phi^{1/2} + W^t, and re-derive the unrolling accordingly.
- [Sec. 4.3, Eq. (12)] The fidelity term in Eq. (12) is not the MAP objective from Eq. (11). Whereas Eq. (7) and Eq. (24) depend on the observed noisy measurements y_i^t and y_q^t, Eq. (12) uses ||(Xa^{r-1})^{-1} xq^{r-1} ⊙ (xi - xi^{r-1})||^2, which effectively replaces y_i^t with the previous estimate xi^{r-1} and y_q^t with xq^{r-1}. The alternating minimization described in the text therefore does not solve the MAP problem stated in Eq. (11). The authors should justify this replacement as a proximal or surrogate approximation, or re-derive the unrolled filter from the original data fidelity term.
- [Sec. 6.1 and Sec. 6.2, Table 1] All quantitative claims are based on a single run of a self-generated synthetic dataset with no standard deviations, number of seeds, or statistical tests. The central performance claims ('outperforming other methods by at least 37.9% in MAE and 13.2% in TEPE') should be accompanied by multiple training runs with different seeds and a report of variance, especially because the dataset generation involves randomly generated camera paths.
- [Sec. 5.1 and Sec. 5.2] The paper does not specify how the mapping confidence matrix Phi^{t,t-1} is computed or how its non-negative diagonal structure is enforced in the network, despite this matrix being load-bearing for the fusion in Eq. (6). The text only states that the GIGA module computes W^{t-1}, W^{t,t-1}, and Phi^{t,t-1} at 1/8 scale. Without this architectural detail, the method is not fully reproducible and the asymmetry issue cannot be assessed empirically.
minor comments (5)
- [Sec. 4.3, Eq. (12)] The notation 'fix x_q^t = y_q^t = x_q^{t,r-1}' conflates the observed measurement with the previous iterate; please clarify the initialization and the update order so that the reader can track which quantity is fixed in each alternating step.
- [Sec. 6] There is a typo in the first paragraph of Sec. 6: 'syntheic' should be 'synthetic'.
- [References and Table 1] UDA is cited as [1] in the text but as [2] in Table 1; the authors should reconcile whether they are referring to the CVPR 2019 paper or the IEEE TPAMI 2022 journal version.
- [Sec. 6.1] The description of DVToF dataset generation is brief; for reproducibility, please provide details of the random camera paths, the range of motions, the scene content, and the exact noise model parameters beyond citing [34].
- [Fig. 3] The output dimensions and module boundaries in Figure 3 are difficult to read at print resolution; please increase the font size or separate the three components more clearly.
Circularity Check
No circular derivation found: the cross-frame graph fusion is a new construction evaluated against external baselines, and the author-overlapping citations are used only as standard components, not as the source of the claimed prediction.
full rationale
I walked the paper's derivation chain from the intra-frame graph construction (Eqs. 4.1), through the mapped graph composition (Eq. 5) and cross-frame fusion (Eq. 6), to the MAP objective (Eq. 11) and its unrolled update (Eq. 14). The central claim is that fusing motion-invariant graph structures improves temporal consistency and spatial sharpness. This claim is not circular: the fused graph is a weighted combination of a mapped previous-frame graph and a current-frame graph, with the confidence diagonal and attention weights learned end-to-end, and the resulting depth accuracy and TEPE are measured against external baselines (Table 1) and ablations (Table 2). The paper does cite the authors' prior GLRUN work [17] for the unrolled GLR module and for the ToF noise distribution, but those are used as building blocks rather than as the output being predicted. The noise-likelihood term is re-derived in the supplementary material (Sec. 8), and the unrolled GLR is itself included as a separate baseline in the experiments, so the citation is not load-bearing for the new graph-fusion contribution. I also note that the skeptic's objection about the asymmetric graph Laplacian and the PSD claim is a mathematical correctness concern, not a circularity: even if Eq. (6) makes the Laplacian nonsymmetric, that does not make the paper's prediction reduce to its inputs by construction. No equation in the paper is equivalent to a fitted parameter renamed as a prediction, and no self-citation chain forces the central result. The paper is self-contained against external benchmarks, so the appropriate circularity score is 0.
Assumptions & free parameters
free parameters (4)
- GLR prior weight lambda (and pixel-wise Lambda) =
learned end-to-end, no single reported value
- Inter-frame neighborhood size q =
q = 7
- Unrolled iteration numbers R and P =
R = 2, P not specified in main text
- Feature scales for graph computation =
1/8 for inter-frame, 1/2 for intra-frame
assumptions (5)
- domain assumption The in-phase x_i and quadrature x_q ToF measurements are corrupted by additive white Gaussian noise.
- standard math The depth noise is small, gamma << 1, so the error-function terms in Eq. (20) can be dropped.
- domain assumption Graph structures are temporally self-similar across frames even when depth values shift.
- domain assumption The 2-hop/3-hop path composition in Eq. (5) correctly maps the reference frame graph to the current frame.
- ad hoc to paper Unrolled gradient descent with R=2 and P inner iterations approximately solves the MAP problem in Eq. (11).
Cite this review
Pith. "Pith review of Consistent Time-of-Flight Depth Denoising via Graph-Informed Geometric Attention." pith.science (2026). https://pith.science/paper/LOOJ6WBW
@misc{pith2026250623542,
author = {Pith},
title = {Pith review of: Consistent Time-of-Flight Depth Denoising via Graph-Informed Geometric Attention},
year = {2026},
howpublished = {\url{https://pith.science/paper/LOOJ6WBW}},
note = {Machine review of arXiv:2506.23542}
}
read the original abstract
Depth images captured by Time-of-Flight (ToF) sensors are prone to noise, requiring denoising for reliable downstream applications. Previous works either focus on single-frame processing, or perform multi-frame processing without considering depth variations at corresponding pixels across frames, leading to undesirable temporal inconsistency and spatial ambiguity. In this paper, we propose a novel ToF depth denoising network leveraging motion-invariant graph fusion to simultaneously enhance temporal stability and spatial sharpness. Specifically, despite depth shifts across frames, graph structures exhibit temporal self-similarity, enabling cross-frame geometric attention for graph fusion. Then, by incorporating an image smoothness prior on the fused graph and data fidelity term derived from ToF noise distribution, we formulate a maximum a posterior problem for ToF denoising. Finally, the solution is unrolled into iterative filters whose weights are adaptively learned from the graph-informed geometric attention, producing a high-performance yet interpretable network. Experimental results demonstrate that the proposed scheme achieves state-of-the-art performance in terms of accuracy and consistency on synthetic DVToF dataset and exhibits robust generalization on the real Kinectv2 dataset. Source code will be released at \href{https://github.com/davidweidawang/GIGA-ToF}{https://github.com/davidweidawang/GIGA-ToF}.
Figures
Figures from the paper (10 more)
Reference graph
Works this paper leans on
-
[1]
Unsupervised domain adaptation for tof data denoising with adversarial learning
Gianluca Agresti, Henrik Schaefer, Piergiorgio Sartor, and Pietro Zanuttigh. Unsupervised domain adaptation for tof data denoising with adversarial learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5584–5593, 2019. 2, 3, 7
work page 2019
-
[2]
Unsupervised domain adapta- tion of deep networks for tof depth refinement
Gianluca Agresti, Henrik Sch ¨afer, Piergiorgio Sartor, Yalcin Incesu, and Pietro Zanuttigh. Unsupervised domain adapta- tion of deep networks for tof depth refinement. IEEE Trans- actions on Pattern Analysis and Machine Intelligence , 44 (12):9195–9208, 2022. 2, 7
work page 2022
-
[3]
Jonathan T Barron and Ben Poole. The fast bilateral solver. In European conference on computer vision, pages 617–632. Springer, 2016. 7
work page 2016
-
[4]
Ayush Bhandari and Ramesh Raskar. Signal processing for time-of-flight imaging sensors: An introduction to inverse problems in computational 3-d imaging. IEEE Signal Pro- cessing Magazine, 33(5):45–58, 2016. 1
work page 2016
-
[5]
Faquan Chen, Rendong Ying, Jianwei Xue, Fei Wen, and Peilin Liu. A configurable and real-time multi-frequency 3d image signal processor for indirect time-of-flight sensors. IEEE Sensors Journal, 22(8):7834–7845, 2022. 1
work page 2022
-
[6]
Very power efficient neu- ral time-of-flight
Yan Chen, Jimmy Ren, Xuanye Cheng, Keyuan Qian, Luyang Wang, and Jinwei Gu. Very power efficient neu- ral time-of-flight. In Proceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision , pages 2257–2266, 2020. 1, 2
work page 2020
-
[7]
Graph spectral image processing
Gene Cheung, Enrico Magli, Yuichi Tanaka, and Michael K Ng. Graph spectral image processing. Proceedings of the IEEE, 106(5):907–930, 2018. 2
work page 2018
-
[8]
Interpretable lightweight transformer via unrolling of learned graph smoothness pri- ors
Tam Thuc Do, Parham Eftekhar, Seyed Alireza Hosseini, Gene Cheung, and Philip Chou. Interpretable lightweight transformer via unrolling of learned graph smoothness pri- ors. Advances in Neural Information Processing Systems , 37:6393–6416, 2025. 6
work page 2025
Show all 46 references
-
[9]
Exploiting dual-correlation for multi-frame time-of- flight denoising
Guanting Dong, Yueyi Zhang, Xiaoyan Sun, and Zhiwei Xiong. Exploiting dual-correlation for multi-frame time-of- flight denoising. In European Conference on Computer Vi- sion, pages 473–489. Springer, 2024. 1, 2, 3, 7
2024
-
[10]
Depthlab: Real-time 3d in- teraction with depth maps for mobile augmented reality
Ruofei Du, Eric Turner, Maksym Dzitsiuk, Luca Prasso, Ivo Duarte, Jason Dourgarian, Joao Afonso, Jose Pascoal, Josh Gladstone, Nuno Cruces, et al. Depthlab: Real-time 3d in- teraction with depth maps for mobile augmented reality. In Proceedings of the 33rd Annual ACM Symposium...
2020
-
[11]
De- noising of continuous-wave time-of-flight depth images us- ing confidence measures
Mario Frank, Matthias Plaue, and Fred A Hamprecht. De- noising of continuous-wave time-of-flight depth images us- ing confidence measures. Optical Engineering , 48(7): 077003–077003, 2009. 1
2009
-
[12]
Theoretical and ex- perimental error analysis of continuous-wave time-of-flight range cameras
Mario Frank, Matthias Plaue, Holger Rapp, Ullrich K ¨othe, Bernd J ¨ahne, and Fred A Hamprecht. Theoretical and ex- perimental error analysis of continuous-wave time-of-flight range cameras. Optical Engineering, 48(1):013602–013602,
-
[13]
Time-of-flight range measurement in low-sensing environ- ment: Noise analysis and complex-domain non-local denois- ing
Mihail Georgiev, Robert Bregovi ´c, and Atanas Gotchev. Time-of-flight range measurement in low-sensing environ- ment: Noise analysis and complex-domain non-local denois- ing. IEEE Transactions on Image Processing , 27(6):2911– 2926, 2018. 2, 4, 1
2018
-
[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 Proceedings of the European Conference on Computer Vision (ECCV), pages 368–383, 2018. 2, 6
2018
-
[15]
itof2dtof: A robust and flexible representation for data-driven time-of-flight imag- ing
Felipe Gutierrez-Barragan, Huaijin Chen, Mohit Gupta, An- dreas Velten, and Jinwei Gu. itof2dtof: A robust and flexible representation for data-driven time-of-flight imag- ing. IEEE Transactions on Computational Imaging, 7:1205– 1214, 2021. 1
2021
-
[16]
Depth map denoising using graph-based transform and group sparsity
Wei Hu, Xin Li, Gene Cheung, and Oscar Au. Depth map denoising using graph-based transform and group sparsity. In 2013 IEEE 15th international workshop on multimedia signal processing (MMSP), pages 001–006. IEEE, 2013. 1, 2, 3
2013
-
[17]
Deep unrolled graph laplacian regularization for robust time-of-flight depth denoising
Jingwei Jia, Changyong He, Jianhui Wang, Gene Cheung, and Jin Zeng. Deep unrolled graph laplacian regularization for robust time-of-flight depth denoising. IEEE Signal Pro- cessing Letters, 32:821–825, 2025. 1, 3, 4, 5, 6, 7
2025
-
[18]
Gradient flow evolution for 3d fusion from a single depth sensor
Jiwoo Kang, Seongmin Lee, Mingyu Jang, and Sanghoon Lee. Gradient flow evolution for 3d fusion from a single depth sensor. IEEE Transactions on Circuits and Systems for Video Technology, 32(4):2211–2225, 2021. 1
2021
-
[19]
Evaluating the accuracy of the Azure Kinect and Kinect v2
Gregorij Kurillo, Evan Hemingway, Mu-Lin Cheng, and Louis Cheng. Evaluating the accuracy of the Azure Kinect and Kinect v2. Sensors, 22(7):2469, 2022. 7
2022
-
[20]
Temporally consistent online depth estimation in dynamic scenes
Zhaoshuo Li, Wei Ye, Dilin Wang, Francis X Creighton, Russell H Taylor, Ganesh Venkatesh, and Mathias Unberath. Temporally consistent online depth estimation in dynamic scenes. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 3018–3027...
2023
-
[21]
Deeptof: off-the-shelf real-time correction of multipath in- terference in time-of-flight imaging
Julio Marco, Quercus Hernandez, Adolfo Munoz, 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 (ToG), 36(6):1–12, 2017. 2, 7
2017
-
[22]
Learning robust perceptive locomotion for quadrupedal robots in the wild
Takahiro Miki, Joonho Lee, Jemin Hwangbo, Lorenz Well- hausen, Vladlen Koltun, and Marco Hutter. Learning robust perceptive locomotion for quadrupedal robots in the wild. Science Robotics, 7(62):eabk2822, 2022. 1
2022
-
[23]
Depth video en- hancement based on weighted mode filtering
Dongbo Min, Jiangbo Lu, and Minh N Do. Depth video en- hancement based on weighted mode filtering. IEEE Trans- actions on Image Processing, 21(3):1176–1190, 2011. 2, 3, 6, 7
2011
-
[24]
Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing
Vishal Monga, Yuelong Li, and Yonina C Eldar. Algorithm unrolling: Interpretable, efficient deep learning for signal and image processing. IEEE Signal Processing Magazine , 38(2):18–44, 2021. 2
2021
-
[25]
Graph signal processing: Overview, challenges, and applications
Antonio Ortega, Pascal Frossard, Jelena Kova vcevi´c, Jos´e MF Moura, and Pierre Vandergheynst. Graph signal processing: Overview, challenges, and applications. Pro- ceedings of the IEEE, 106(5):808–828, 2018. 2
2018
-
[26]
Graph laplacian regulariza- tion for image denoising: Analysis in the continuous domain
Jiahao Pang and Gene Cheung. Graph laplacian regulariza- tion for image denoising: Analysis in the continuous domain. IEEE Transactions on Image Processing, 26(4):1770–1785,
-
[27]
Graph spectral image restoration
Jiahao Pang and Jin Zeng. Graph spectral image restoration. Graph Spectral Image Processing, 133, 2021. 3
2021
-
[28]
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. 6
2017
-
[29]
Don’t forget the past: Recurrent depth estimation from monocular video
Vaishakh Patil, Wouter Van Gansbeke, Dengxin Dai, and Luc Van Gool. Don’t forget the past: Recurrent depth estimation from monocular video. IEEE Robotics and Automation Let- ters, 5(4):6813–6820, 2020. 2, 7
2020
-
[30]
Depth restoration in under- display time-of-flight imaging
Xin Qiao, Chenyang Ge, Pengchao Deng, Hao Wei, Matteo Poggi, and Stefano Mattoccia. Depth restoration in under- display time-of-flight imaging. IEEE Transactions on Pat- tern Analysis and Machine Intelligence , 45(5):5668–5683,
-
[31]
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 International Conference on Medical Image Com- puting and Computer-Assisted Intervention, pages 234–241. Springer, 2015. 6
2015
-
[32]
Joint graph-based depth refinement and nor- mal estimation
Mattia Rossi, Mireille El Gheche, Andreas Kuhn, and Pas- cal Frossard. Joint graph-based depth refinement and nor- mal estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12154– 12163, 2020. 1, 2
2020
-
[33]
Radu: Ray-aligned depth update convolutions for tof data denoising
Michael Schelling, Pedro Hermosilla, and Timo Ropinski. Radu: Ray-aligned depth update convolutions for tof data denoising. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 671–680,
-
[34]
Deep end-to-end time-of-flight imaging
Shuochen Su, Felix Heide, Gordon Wetzstein, and Wolfgang Heidrich. Deep end-to-end time-of-flight imaging. In Pro- ceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 6383–6392, 2018. 1, 2, 6, 7
2018
-
[35]
Consistent direct time-of-flight video depth super-resolution
Zhanghao Sun, Wei Ye, Jinhui Xiong, Gyeongmin Choe, Jialiang Wang, Shuochen Su, and Rakesh Ranjan. Consistent direct time-of-flight video depth super-resolution. In Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5075–5085, 2023. 1, 2...
2023
-
[36]
libfreenect2: Re- lease 0.2, 2016
Lingzhu Xiang, Florian Echtler, Christian Kerl, Thiemo Wiedemeyer, R Gordon, and F Facioni. libfreenect2: Re- lease 0.2, 2016. 1, 2, 6, 7
2016
-
[37]
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 Proceedings of the European conference on com- puter vision (ECCV), pages 151–167, 2018. 2
2018
-
[38]
Graph-based depth video denoising and event detection for sleep monitoring
Cheng Yang, Yu Mao, Gene Cheung, Vladimir Stankovic, and Kevin Chan. Graph-based depth video denoising and event detection for sleep monitoring. In 2014 IEEE 16th international workshop on multimedia signal processing (MMSP), pages 1–6. IEEE, 2014. 2, 3
2014
-
[39]
Time-of-flight and structured light depth cameras
Pietro Zanuttigh, Giulio Marin, Carlo Dal Mutto, Fabio Dominio, Ludovico Minto, Guido Maria Cortelazzo, et al. Time-of-flight and structured light depth cameras. Technol- ogy and Applications, 978(3), 2016. 3
2016
-
[40]
Deep graph laplacian regularization for robust denoising of real images
Jin Zeng, Jiahao Pang, Wenxiu Sun, and Gene Cheung. Deep graph laplacian regularization for robust denoising of real images. In Proceedings of the ieee/cvf conference on com- puter vision and pattern recognition workshops , pages 0–0,
-
[41]
Graph-based depth denoising & dequantization for point cloud enhance- ment
Xue Zhang, Gene Cheung, Jiahao Pang, Yash Sanghvi, Ab- hiram Gnanasambandam, and Stanley H Chan. Graph-based depth denoising & dequantization for point cloud enhance- ment. IEEE Transactions on Image Processing , 31:6863– 6878, 2022. 1, 2 Consistent Time-of-Flight Depth Denois...
2022
-
[42]
As assumed in Sec
Data Fidelity Term in MAP Problem In this section, we derive the data fidelity term based on ToF depth noise distribution. As assumed in Sec. 4.3, xi and xq are corrupted by additive white Gaussian noise (AWGN) [12, 13], and the pixels in yt i, yt q are independent and iden- t...
-
[43]
For small time steps ∆t = 1, 2, the performances are similar
Analysis of Frame Time Step Following [9], to investigate the effect of time step between reference and current frames, we test GIGA-ToF on DVToF dataset with different time steps. For small time steps ∆t = 1, 2, the performances are similar. When time steps become larger, ∆t ...
-
[44]
Algorithm Summary Based on the algorithm unrolling of graph Laplacian regu- larization, we obtain the solution to (14), which is summa- rized in Algorithm 1. Algorithm 1 Unrolling of Cross-frame Graph Fusion based ToF Depth Denoising Algorithm Require: Noisy ToF raw data yt i,...
-
[45]
In particular, we demon- strate results on synthetic DVToF dataset in Fig
More Visualization We provide more results for the qualitative comparison of ToF depth denoising methods. In particular, we demon- strate results on synthetic DVToF dataset in Fig. 11 and Fig. 12, and DVToF dataset with noise augmentation in Fig. 13. To further demonstrate the...
-
[46]
Please kindly refer to the supple- mentary video for better temporal visualizations
Note that we directly apply the model trained on orig- inal DVToF dataset to the noise-augmented DVToF dataset and Kinectv2 dataset without fine-tuning, which validates its generalization ability. Please kindly refer to the supple- mentary video for better temporal visualizati...
Reviewed August 6, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.