Pith. sign in

REVIEW 3 major objections 7 minor 58 references

DiffStereo: High-Frequency Aware Diffusion Model for Stereo Image Restoration

T0 review · 3 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read DiffStereo is the first diffusion model for stereo image restoration.

desk verdict First stereo diffusion restoration paper, but the key ablation confuses oracle guidance with what the diffusion model actually delivers. read the letter →

arxiv 2501.10325 v1 pith:OCL45WXU submitted 2025-01-17 cs.CV

classification cs.CV
keywords DiffStereoimagerestorationdiffusionmodellatenthigh-frequencyrepresentationtransformersuper-resolutiondeblurringlow-lightenhancement
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

DiffStereo sets out to bring diffusion models to stereo image restoration, a setting where they have not been used, by running the diffusion process on a compact latent space dedicated to high-frequency detail rather than on full images or semantic latents. The model first distills ground-truth stereo pairs into a single-channel latent high-frequency representation (LHFR) at full resolution, then trains a diffusion model to predict that map from the degraded left-right input, and finally fuses the predicted map into a transformer-based restoration network with a depth-dependent position encoding. The paper reports that this design improves both pixel accuracy (PSNR/SSIM) and perceptual quality (FID/LPIPS) over state-of-the-art CNN and transformer methods on stereo super-resolution, deblurring, and low-light enhancement. If the claim holds, DiffStereo is the first effective diffusion-based approach to stereo restoration, and it offers a division of labour where the diffusion model supplies realistic texture guidance while the regression network keeps fidelity.

What carries the argument

The central object is the latent high-frequency representation (LHFR): a single-channel map at the same resolution as the input stereo images, produced by LREN from the HQ pair and estimated by the diffusion model from the LQ pair. It carries the high-frequency texture and edge information that ordinary latent diffusion discards during compression, and its full spatial resolution preserves the texture of the input. The other load-bearing pieces are the channel interaction block (CIB), a transformer unit that computes a shared channel attention map over concatenated left-right features, and the position encoding scheme that modulates CIB features with the LHFR differently according to the block's depth. Together they let the diffusion prior steer fine detail while the transformer enforces cross-view consistency and fidelity.

What would settle it

A decisive check is to ablate the diffusion-estimated LHFR on the same benchmark by replacing them with a constant map after training; if PSNR/SSIM stay within the roughly 0.07 dB margin seen in Table 5, the high-frequency guidance is not load-bearing, while a large drop would confirm the mechanism.

Watch

Extended reading notes

Core claim

The central claim is that the right role for a diffusion model in stereo restoration is not to generate the output image but to generate a compressed, full-resolution high-frequency map of the target stereo pair. The latent representation extraction network (LREN) is trained, jointly with the stereo image restoration network (SIRN), to distill ground-truth stereo pairs into a single-channel map that retains edge and texture structure. A diffusion model is then trained in that space to predict the map from the degraded stereo input, and the predicted map is fused into each channel interaction block (CIB) of SIRN with a position-encoding scheme that lets the same map guide different network depths differently. The paper argues this keeps the generative power of diffusion for distribution-level realism while the regression-based transformer keeps pixel fidelity, avoiding the accuracy drop typical of image-space diffusion restoration. Experiments across three tasks are presented as evidence that this division of labour improves both PSNR/SSIM and FID/LPIPS over previous methods.

Load-bearing premise

The pipeline rests on the assumption that a one-channel map distilled from the ground-truth stereo pair retains the specific high-frequency detail that matters for restoration, and that the diffusion model can predict that map from the degraded input; if the predicted map adds only about 0.07 dB on the main super-resolution table, the claimed mechanism is not what carries the results.

Editorial extensions

If this is right

  • Diffusion models can be applied to stereo restoration without the usual compute blow-up, because the diffusion process runs on a one-channel map at full resolution instead of on two full-color images.
  • Accuracy and perceptual quality do not have to be traded off: on the reported tasks DiffStereo improves PSNR/SSIM and FID/LPIPS at the same time, unlike image-space diffusion methods that typically sacrifice one.
  • Because the LHFR is estimated from the degraded input and then fused into a regression network, the diffusion model does not directly synthesize pixels, which the paper credits with suppressing hallucinated details.
  • The position-encoding fusion means the same high-frequency guidance can be reused at different network depths, giving the restoration network a way to exploit the prior at multiple scales of abstraction.
  • The framework transfers across degradation types: super-resolution, deblurring, and low-light enhancement are all handled by the same architecture, suggesting the LHFR prior is not task-specific.

Reading between the lines

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

  • A cautious inference from the ablations is that the diffusion-estimated maps are not yet the main source of the accuracy gain: on the super-resolution benchmark they add about 0.07 dB, while the oracle-map comparisons show gains of several dB, so the transformer backbone and training-time information are likely doing much of the work.
  • A testable extension beyond the paper is to condition the diffusion model on disparity or epipolar geometry from the stereo pair, in addition to the degraded views; if high-frequency prediction is the bottleneck, cross-view geometric cues should close part of the gap.
  • The same LHFR-as-guidance design could transfer to other multi-view settings such as light-field restoration or multi-frame super-resolution, where one shared full-resolution high-frequency map could be estimated by a single diffusion pass and then guide a regression network.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper proposes DiffStereo, a two-stage framework for stereo image restoration that combines a diffusion model with a transformer-based restoration network. In Stage 1, a latent representation extraction network (LREN) is trained jointly with a stereo image restoration network (SIRN) to compress high-frequency details of HQ stereo images into single-channel latent high-frequency representations (LHFR), which are fused into the SIRN via a depth-position encoding scheme. In Stage 2, a diffusion model is trained to estimate these LHFR from degraded stereo images and the estimates are used as guidance at inference. Experiments on ×4 super-resolution, deblurring, and low-light enhancement report favorable PSNR/SSIM, FID, and LPIPS against CNN/transformer baselines. The central claim is that this combination yields both higher reconstruction accuracy and better perceptual quality than state-of-the-art methods.

Significance. If substantiated, the work would be the first diffusion-based stereo image restoration method and introduces a plausible design choice: running the diffusion process in a spatially preserved, channel-compressed latent space rather than in pixel space, and using the generated representation as auxiliary guidance rather than as the direct output. This is a sensible way to avoid the typical PSNR drop of generative restoration models. The experimental coverage of three tasks and the inclusion of perceptual metrics (FID/LPIPS) are strengths. However, the paper's own ablations do not currently confirm that the diffusion-estimated LHFR are responsible for the reported gains: Table 4 evaluates oracle (ground-truth) LHFR at Stage 1, while Table 5 shows only marginal DM-based gains. The 'high-frequency' identity of the learned representation is also asserted largely from visualizations rather than from quantitative analysis. The central mechanism is therefore plausible but unverified, and the SOTA margins on SR are small.

major comments (3)
  1. [§4.3.1, Tables 4 and 5] The text states that 'the latent representations estimated by DM benefit the quality of reconstruction' and cites Table 4, but all rows of Table 4 are Stage 1 results, in which the LHFR are extracted from ground-truth HQ images by the pretrained LREN and the diffusion model is not used. Row 3 versus row 1 shows a 3.45 dB gain on Flickr1024, which is an oracle-guidance result, not a DM result. The actual inference-time contribution of the diffusion-estimated LHFR is shown in Table 5: on SR the gains are 0.07, 0.02, 0.05, and 0.11 dB on Flickr1024, KITTI2012, KITTI2015, and Middlebury, respectively, and 0.06 dB on low-light enhancement; several SSIM values decrease or stay flat (e.g., deblur Flickr1024 SSIM 0.9790 vs 0.9790, KITTI2012 SSIM 0.9725 vs 0.9724). These magnitudes are close to noise level and do not support the claim that the DM-estimated LHFR are the mechanism behind the reported improvements. The paper should present a Stage-2 ablation with and without DM-estimated LHFR as the main evidence, and clearly separate oracle and estimated settings in Table 4.
  2. [§3.3 and Eq. (15)] The diffusion-related loss is not derived from the standard DDPM objective in Eq. (14). Instead, the network is trained by backpropagating the L1 distance between the final deterministic output after T iterations and the LREN-extracted LHFR, with the random term dropped in Eq. (11). This is a legitimate objective for a deterministic iterative denoiser, but it is not the variational bound of a diffusion model, and the paper does not explain why the full-chain loss is preferable to the conventional noise-prediction loss. The choice of T=4 is justified only by a sentence in the Fig. 6 caption ('more iterations do not provide further improvement') without a supporting ablation. Either derive the objective from a principled surrogate, or reframe the module as an iterative refinement network and provide the missing T-ablation.
  3. [§3.2.1 and Fig. 6] The claim that the learned latent representations are 'high-frequency' is not quantitatively established. The LHFR are defined as the output of LREN, which is co-trained with the reconstruction loss of the final restoration; the visualization in Fig. 6 shows edge-like patterns, but no spectral or fidelity analysis is provided to show that these representations capture high-frequency content of the HQ image beyond what a generic compressed feature would capture. Since the method is named and motivated by high-frequency awareness, the paper should include a quantitative characterization (e.g., comparing the Fourier spectrum of the LHFR with that of the input LQ/HQ images, or a detail-recovery metric) and, ideally, an ablation with a similarly sized latent that is not trained with the 'high-frequency' objective.
minor comments (7)
  1. [§4.3.1, Table 4] The 'Stage' column should be renamed or annotated so that rows 3 and 4 clearly indicate that LHFR are extracted from the ground-truth HQ images (oracle) and not estimated by the diffusion model; the current layout invites the misinterpretation that the text makes.
  2. [§3.4, Eq. (15)] The notation for the extracted ground-truth LHFR and the estimated LHFR is confusing because both are denoted Z in different equations; use e.g. Z^gt and Z^est consistently.
  3. [Fig. 6 caption] 'interations' should be 'iterations', and the claim about T=4 should be moved to the main text with an ablation table.
  4. [§4.2.2, Table 2] The column header 'Left (Left + Right) / 2' is ambiguous; it should be split into two settings, e.g., 'Left' and '(Left+Right)/2'.
  5. [Various] There are several typos and unresolved references, including 'denosing' in §3.3, 'generalizaiton' in §4.2.2, 'Visiual' in Fig. 4 caption, and a placeholder '[31?]' in §2.2.
  6. [§3.2.3] What is called 'position encoding' is a depth-index concatenation learned via convolution; the name might be confused with spatial position encoding used in transformers. Consider renaming to 'depth conditioning' or justify the terminology.
  7. [Abstract and Table 1] The abstract's claim of 'higher reconstruction accuracy' is not universally supported: on KITTI2012 DiffStereo's PSNR (26.67) is lower than SSRDE-FNet (26.69). The text should state the per-dataset comparison precisely, e.g., as 'comparable on KITTI2012/KITTI2015 and better on Flickr1024/Middlebury'.

Circularity Check

2 steps flagged · score 4.0 of 10

LHFR are self-defined through the co-trained LREN, and Table 4's large ablation gains come from oracle Stage-1 features rather than DM predictions; the externally benchmarked SOTA comparisons keep the paper from full circularity.

  1. self definitional [Sections 3.2 and 3.3, Eq. (15), Table 5]
    "In the first stage, we train the latent representation extraction network (LREN) and stereo image restoration network (SIRN) together to enforce the LREN to learn high-frequency representations conducive to restoration. ... The diffusion model then learns to estimate the LHFR given LQ images, approximating the output of the LREN."

    The 'high-frequency' target is not defined by any external measurement. It is defined as the output of LREN trained jointly with SIRN under the reconstruction loss L_stage1 = L_rec + lambda1 * L_para, i.e., the same restoration objective used to evaluate the final network. The DM is then trained to reproduce this self-defined feature (Eq. 15: L_diff = ||Z_hat_L - Z_L||_1 + ||Z_hat_R - Z_R||_1, with Z_L, Z_R extracted by LREN). Demonstrating that injecting this feature improves restoration is therefore partly a statement that injecting a feature optimized to improve restoration improves restoration.

  2. fitted input called prediction [Section 4.3.1, Tables 4 and 5]
    "Comparing (1) with (2) and (3), it is obvious that the latent representations estimated by DM benefit the quality of reconstruction in terms of PSNR and SSIM. ... Furthermore, we assess the role of LHFR generated by DM on SR, deblurring and low-light enhancement."

    The large improvements attributed to 'LHFR estimated by DM' are taken from Table 4, where every ablation row is a Stage-1 result. In Stage 1 the LHFR are extracted by LREN directly from ground-truth HQ images, not estimated by the diffusion model; row 3 versus row 1 gives about 3.45 dB on Flickr1024. Table 5, which actually evaluates the inference-time setting with DM-estimated LHFR, shows only 0.07, 0.02, 0.05 and 0.11 dB SR gains on the four datasets, and 0.06 dB on low-light enhancement. Thus the load-bearing evidence that DM-estimated LHFR provide beneficial high-frequency guidance reduces to oracle ground-truth-derived features, and the claimed 'prediction' benefit is not what the cited table measures.

full rationale

No load-bearing self-citation chain or imported uniqueness argument was found; the comparisons against EDSR, PASSRnet, iPASSR, SSRDE-FNet, and SIR-Former use external benchmark metrics and are not circular. The two substantive circularity concerns are: (1) the LHFR are internally constructed by LREN under the same restoration loss that SIRN is evaluated on, so the label 'high-frequency' and the claim that injecting LHFR helps restoration are partly self-definitional; and (2) Section 4.3.1 presents Stage-1 oracle-LHFR gains from Table 4 as if they came from DM-estimated LHFR, while Table 5 shows the actual DM-estimated contribution is near noise level on the main SR benchmark. Because the final PSNR/SSIM/FID results are independent and the framework is not a pure self-referential loop, the paper is only partially circular; the central mechanism, however, remains largely unverified as a genuine diffusion-based high-frequency prediction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central mechanism depends on the learned LHFR target, which is defined internally by the co-trained LREN and SIRN. The free parameters (lambda_1, lambda_2, T) are tuned by hand, and the key assumption that LHFR capture high frequencies is only visually supported.

free parameters (3)
  • lambda_1 (parallax loss weight) = 0.1
    Set by hand in Section 3.4.
  • lambda_2 (diffusion loss weight) = 0.25
    Set by hand in Section 3.4.
  • T (number of diffusion steps) = 4
    Chosen because more iterations did not improve results (Figure 6 caption).
assumptions (4)
  • standard math Standard DDPM forward and reverse process equations
    Used in Section 3.3 without proof, following Ho et al.
  • domain assumption The learned single-channel LHFR map represents high-frequency information
    Only supported by visualization in Figure 6; no quantitative frequency analysis is provided.
  • ad hoc to paper The full-chain L1 loss in Eq. (15) is a valid training objective for the diffusion model
    The paper does not derive Eq. (15) from the standard noise-prediction objective; it is a custom training choice.
  • domain assumption Left and right views share the same channel covariance
    Assumed in Section 3.2.2 to justify shared channel attention in CIB.
invented entities (1)
  • Latent high-frequency representations (LHFR)
    purpose: Guide the transformer restoration network with texture information
    The paper claims LHFR encode high-frequency details, but provides only visualizations and no external or quantitative verification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DiffStereo: High-Frequency Aware Diffusion Model for Stereo Image Restoration." pith.science (2026). https://pith.science/paper/OCL45WXU

@misc{pith2026250110325,
  author       = {Pith},
  title        = {Pith review of: DiffStereo: High-Frequency Aware Diffusion Model for Stereo Image Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OCL45WXU}},
  note         = {Machine review of arXiv:2501.10325}
}
read the original abstract

Diffusion models (DMs) have achieved promising performance in image restoration but haven't been explored for stereo images. The application of DM in stereo image restoration is confronted with a series of challenges. The need to reconstruct two images exacerbates DM's computational cost. Additionally, existing latent DMs usually focus on semantic information and remove high-frequency details as redundancy during latent compression, which is precisely what matters for image restoration. To address the above problems, we propose a high-frequency aware diffusion model, DiffStereo for stereo image restoration as the first attempt at DM in this domain. Specifically, DiffStereo first learns latent high-frequency representations (LHFR) of HQ images. DM is then trained in the learned space to estimate LHFR for stereo images, which are fused into a transformer-based stereo image restoration network providing beneficial high-frequency information of corresponding HQ images. The resolution of LHFR is kept the same as input images, which preserves the inherent texture from distortion. And the compression in channels alleviates the computational burden of DM. Furthermore, we devise a position encoding scheme when integrating the LHFR into the restoration network, enabling distinctive guidance in different depths of the restoration network. Comprehensive experiments verify that by combining generative DM and transformer, DiffStereo achieves both higher reconstruction accuracy and better perceptual quality on stereo super-resolution, deblurring, and low-light enhancement compared with state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2501.10325 by the authors.

Figure 1
Figure 1. An overview of DiffStereo in training stage one. The latent representation extraction network (LREN) learns a [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The architecture of our proposed: (a) channel interaction block (CIB), (b) Position Encoding scheme, (c) channel [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. An overview of DiffStereo in training stage two. The DM learns to estimate the LHFR extracted by pretrained LREN, [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Visiual comparisons for ×4 SR by different methods on Flickr1024 and Middlebury dataset [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visiual comparisons in Low-Light Enhancement by different methods. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Visualizations of LHFR in different training stages and different timesteps of denoising. The total timestep [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 45 canonical work pages

  1. [1]

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dockhorn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. 2023. Align your latents: High-resolution video synthesis with latent diffusion models. In CVPR

  2. [2]

    Canqiang Chen, Chunmei Qing, Xiangmin Xu, and Patrick Dickinson. 2021. Cross parallax attention network for stereo image super-resolution. IEEE TMM (2021)

  3. [3]

    Liangyu Chen, Xiaojie Chu, Xiangyu Zhang, and Jian Sun. 2022. Simple baselines for image restoration. In ECCV

  4. [4]

    Qinyan Dai, Juncheng Li, Qiaosi Yi, Faming Fang, and Guixu Zhang. 2021. Feed- back network for mutually boosted stereo image super-resolution and disparity estimation. In ACM MM

  5. [5]

    Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. NeurIPS (2021)

  6. [6]

    Patrick Esser, Johnathan Chiu, Parmida Atighehchian, Jonathan Granskog, and Anastasis Germanidis. 2023. Structure and content-guided video synthesis with diffusion models. In CVPR

  7. [7]

    Patrick Esser, Robin Rombach, and Bjorn Ommer. 2021. Taming transformers for high-resolution image synthesis. In CVPR

  8. [8]

    Sicheng Gao, Xuhui Liu, Bohan Zeng, Sheng Xu, Yanjing Li, Xiaoyan Luo, Jianzhuang Liu, Xiantong Zhen, and Baochang Zhang. 2023. Implicit diffusion models for continuous super-resolution. In CVPR

Show all 58 references
  1. [9]

    Andreas Geiger, Philip Lenz, and Raquel Urtasun. 2012. Are we ready for au- tonomous driving? the kitti vision benchmark suite. In CVPR

  2. [10]

    Hansheng Guo, Juncheng Li, Guangwei Gao, Zhi Li, and Tieyong Zeng. 2023. Pft-ssr: Parallax fusion transformer for stereo image super-resolution. InICASSP

  3. [11]

    Lanqing Guo, Chong Wang, Wenhan Yang, Siyu Huang, Yufei Wang, Hanspeter Pfister, and Bihan Wen. 2023. Shadowdiffusion: When degradation prior meets diffusion model for shadow removal. In CVPR

  4. [12]

    Zhi-min Guo et al. 2011. Research of hand positioning and gesture recognition based on binocular vision. In IEEE International Symposium on VR Innovation

  5. [13]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. NeurIPS (2020)

  6. [14]

    Jonathan Ho, Chitwan Saharia, William Chan, David J Fleet, Mohammad Norouzi, and Tim Salimans. 2022. Cascaded diffusion models for high fidelity image generation. JMLR (2022)

  7. [15]

    Yiwen Hua, Puneet Kohli, Pritish Uplavikar, Anand Ravi, Saravana Gunaseelan, Jason Orozco, and Edward Li. 2020. Holopix50k: A large-scale in-the-wild stereo image dataset. arXiv preprint arXiv:2003.11172 (2020)

  8. [16]

    Jie Huang, Xueyang Fu, Zeyu Xiao, Feng Zhao, and Zhiwei Xiong. 2022. Low-light stereo image enhancement. IEEE TMM (2022)

  9. [17]

    Rongjie Huang, Jiawei Huang, Dongchao Yang, Yi Ren, Luping Liu, Mingze Li, Zhenhui Ye, Jinglin Liu, Xiang Yin, and Zhou Zhao. 2023. Make-an-audio: Text-to-audio generation with prompt-enhanced diffusion models. In ICML

  10. [18]

    Daniel S Jeon, Seung-Hwan Baek, Inchang Choi, and Min H Kim. 2018. Enhancing the spatial resolution of stereo images using a parallax prior. In CVPR

  11. [19]

    Bingxi Jia, Jian Chen, and Kaixiang Zhang. 2016. Drivable road reconstruction for intelligent vehicles based on two-view geometry. IEEE Transactions on Industrial Electronics (2016)

  12. [20]

    Diederik P Kingma and Max Welling. 2013. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114 (2013)

  13. [21]

    Haoying Li, Yifan Yang, Meng Chang, Shiqi Chen, Huajun Feng, Zhihai Xu, Qi Li, and Yueting Chen. 2022. Srdiff: Single image super-resolution with diffusion probabilistic models. Neurocomputing (2022)

  14. [22]

    Jingyun Liang, Jiezhang Cao, Guolei Sun, Kai Zhang, Luc Van Gool, and Radu Timofte. 2021. Swinir: Image restoration using swin transformer. In CVPR

  15. [23]

    Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee. 2017. Enhanced deep residual networks for single image super-resolution. In CVPR workshops

  16. [24]

    Jianxin Lin, Lianying Yin, and Yijun Wang. 2023. Steformer: Efficient stereo image super-resolution with transformer. IEEE TMM (2023)

  17. [25]

    Shitong Luo and Wei Hu. 2021. Diffusion probabilistic models for 3d point cloud generation. In CVPR

  18. [26]

    Moritz Menze and Andreas Geiger. 2015. Object scene flow for autonomous vehicles. In CVPR

  19. [27]

    Axi Niu, Kang Zhang, Trung X Pham, Jinqiu Sun, Yu Zhu, In So Kweon, and Yanning Zhang. 2023. CDPMSR: Conditional Diffusion Probabilistic Models for Single Image Super-Resolution. arXiv preprint arXiv:2302.12831 (2023)

  20. [28]

    Taragay Oskiper, Mikhail Sizintsev, Vlad Branzoi, Supun Samarasekera, and Rakesh Kumar. 2015. Augmented reality binoculars. IEEE transactions on visual- ization and computer graphics (2015)

  21. [29]

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen

  22. [30]

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In CVPR

  23. [31]

    Chitwan Saharia, William Chan, Huiwen Chang, Chris Lee, Jonathan Ho, Tim Salimans, David Fleet, and Mohammad Norouzi. 2022. Palette: Image-to-image diffusion models. In SIGGRAPH

  24. [32]

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. 2022. Photorealistic text-to-image diffusion models with deep language understanding. NeurIPS (2022)

  25. [33]

    Chitwan Saharia, Jonathan Ho, William Chan, Tim Salimans, David J Fleet, and Mohammad Norouzi. 2022. Image super-resolution via iterative refinement. IEEE TPAMI (2022)

  26. [34]

    Daniel Scharstein, Heiko Hirschmüller, York Kitajima, Greg Krathwohl, Nera Nešić, Xi Wang, and Porter Westling. 2014. High-resolution stereo datasets with subpixel-accurate ground truth. In Pattern Recognition

  27. [35]

    Shuyao Shang, Zhengyang Shan, Guangxing Liu, and Jinglin Zhang. 2023. Resdiff: Combining cnn and diffusion model for image super-resolution. arXiv preprint arXiv:2303.08714 (2023)

  28. [36]

    Uriel Singer, Adam Polyak, Thomas Hayes, Xi Yin, Jie An, Songyang Zhang, Qiyuan Hu, Harry Yang, Oron Ashual, Oran Gafni, et al. 2022. Make-a-video: Text-to-video generation without text-video data. arXiv preprint arXiv:2209.14792 (2022)

  29. [37]

    Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli

  30. [38]

    Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020)

  31. [39]

    Wonil Song, Sungil Choi, Somi Jeong, and Kwanghoon Sohn. 2020. Stereoscopic image super-resolution with stereo consistent feature. In AAAI

  32. [40]

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. 2020. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456 (2020)

  33. [41]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023)

  34. [42]

    Aaron Van Den Oord, Oriol Vinyals, et al. 2017. Neural discrete representation learning. NeurIPS (2017)

  35. [43]

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. 2023. Exploiting diffusion prior for real-world image super- resolution. arXiv preprint arXiv:2305.07015 (2023)

  36. [44]

    Longguang Wang, Yulan Guo, Yingqian Wang, Zhengfa Liang, Zaiping Lin, Jun- gang Yang, and Wei An. 2020. Parallax attention for unsupervised stereo corre- spondence learning. IEEE TPAMI (2020)

  37. [45]

    Longguang Wang, Yingqian Wang, Zhengfa Liang, Zaiping Lin, Jungang Yang, Wei An, and Yulan Guo. 2019. Learning parallax attention for stereo image super-resolution. In CVPR

  38. [46]

    Yingqian Wang, Longguang Wang, Jungang Yang, Wei An, and Yulan Guo. 2019. Flickr1024: A large-scale dataset for stereo image super-resolution. In CVPR Workshops

  39. [47]

    Yingqian Wang, Xinyi Ying, Longguang Wang, Jungang Yang, Wei An, and Yulan Guo. 2021. Symmetric parallax attention for stereo image super-resolution. In CVPR

  40. [48]

    Bin Xia, Yulun Zhang, Shiyin Wang, Yitong Wang, Xinglong Wu, Yapeng Tian, Wenming Yang, and Luc Van Gool. 2023. Diffir: Efficient diffusion model for image restoration. arXiv preprint arXiv:2303.09472 (2023)

  41. [49]

    Qingyu Xu, Longguang Wang, Yingqian Wang, Weidong Sheng, and Xinpu Deng

  42. [50]

    Bo Yan, Chenxi Ma, Bahetiyaer Bare, Weimin Tan, and Steven CH Hoi. 2020. Disparity-aware domain adaptation in stereo image restoration. In CVPR

  43. [51]

    Zizheng Yang, Mingde Yao, Jie Huang, Man Zhou, and Feng Zhao. 2022. Sir- former: Stereo image restoration using transformer. In ACM MM

  44. [52]

    Xinyi Ying, Yingqian Wang, Longguang Wang, Weidong Sheng, Wei An, and Yulan Guo. 2020. A stereo attention module for stereo image super-resolution. IEEE SPL (2020)

  45. [53]

    Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, and Ming-Hsuan Yang. 2022. Restormer: Efficient transformer for high- resolution image restoration. In CVPR

  46. [54]

    Yulun Zhang, Kunpeng Li, Kai Li, Lichen Wang, Bineng Zhong, and Yun Fu. 2018. Image super-resolution using very deep residual channel attention networks. In ECCV

  47. [55]

    Yulun Zhang, Yapeng Tian, Yu Kong, Bineng Zhong, and Yun Fu. 2018. Residual dense network for image super-resolution. In CVPR

  48. [2015]

    Deep unsupervised learning using nonequilibrium thermodynamics. In ICML

  49. [2021]

    IEEE SPL (2021)

    Deep bilateral learning for stereo image super-resolution. IEEE SPL (2021)

  50. [2022]

    arXiv preprint arXiv:2204.06125 (2022)

    Hierarchical text-conditional image generation with clip latents. arXiv preprint arXiv:2204.06125 (2022)

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.