This paper studies a diffusion-based framework to address the low-light image enhancement problem. To harness the capabilities of diffusion models, we delve into this intricate process and advocate for the regularization of its inherent ODE-trajectory. To be specific, inspired by the recent research that low curvature ODE-trajectory results in a stable and effective diffusion process, we formulate a curvature regularization term anchored in the intrinsic non-local structures of image data, i.e., global structure-aware regularization, which gradually facilitates the preservation of complicated details and the augmentation of contrast during the diffusion process. This incorporation mitigates the adverse effects of noise and artifacts resulting from the diffusion process, leading to a more precise and flexible enhancement. To additionally promote learning in challenging regions, we introduce an uncertainty-guided regularization technique, which wisely relaxes constraints on the most extreme regions of the image. Experimental evaluations reveal that the proposed diffusion-based framework, complemented by rank-informed regularization, attains distinguished performance in low-light enhancement. The outcomes indicate substantial advancements in image quality, noise suppression, and contrast amplification in comparison with state-of-the-art methods. We believe this innovative approach will stimulate further exploration and advancement in low-light image processing, with potential implications for other applications of diffusion models. The code is publicly available at https://github.com/jinnh/GSAD.
Paper
Similar papers
Peer review
Summary
This work introduces a low-light image enhancement based on diffusion model. By incorporating global structure-aware regularization, this work achieves more promising performances than existing methods.
Strengths
By considering the global structure during the diffusion process, this work achieves better performance.
Weaknesses
Despite its better performance, I am very sure that this work cannot advance this field. It does not solve a problem in a new scenario and it just make incrimental revision on top of established framework. My concerns for this work can be concluded as follows: 1. The authors claim that a naive implementation of the diffusion model alone is inadequate for effectively resolving the low-light image enhancement, while it does not prove this claim. All baselines considered in the experiments are non-diffusion based approaches. Thus, it is difficult to say whether a naive implementation is enough for the problem solving. 2. In line 151, the autors claim that the rank of matrix is an effective measurement of the global structures. This is a very important assumptions in this place. This part needs more experiments and justifactions. Or, the authors should provide very solid references for this claim in this place. 3. Except the matrix ranking, several approachs have been proposed to better represent the global structures including some deep learning features. The authors should compare with them. 4. Why is it important to use global structure in this place? If U-Net is a substantial structure of this proposed framework, it is able to capture the global structure during its network propagation. Thus, why do this work design an additional module to capture global structure. 5. I am not sure about the effectiveness of uncertainty in this place. I cannot find Table 5 mentioned in this manuscript. 6. This work just puts several different image processing modules together without enough insights. It is easy to come up with such an idea. The non-local solutions have been discussed in traditional solutions. I donot know why they choose to combine it with diffusion model. Besides, weighted loss has been considered for a large number of methods. Using it here cannot be considered as a new contribution.
Questions
See my question above.
Rating
3: Reject: For instance, a paper with technical flaws, weak evaluation, inadequate reproducibility and incompletely addressed ethical considerations.
Confidence
5: You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
Soundness
3 good
Presentation
3 good
Contribution
2 fair
Limitations
See my questions above.
Summary
This paper presents an innovative and efficacious approach to regularization within the domain of diffusion models, with particular utility for low-light image reconstruction. The method ingeniously takes into account the latent global structure of images. To achieve this, an image is first divided into patches, and clustering algorithms are then employed to group analogous patches, where the matrix rank is further utilized to minimize the discrepancy between the intermediate reconstruction and the final output. Besides, uncertainty is used to boost performance.
Strengths
1) The novel perspective opens up a promising avenue for enhancing the performance of diffusion models, representing an important contribution to the field. Also, it may have great potential for other diffusion-based image/video processing tasks. 2) The introduction of this form of regularization is a groundbreaking shift for diffusion models, distinguishing this work from existing literature. 3) The experimental validations lend strong support to the efficacy of the proposed regularization term, and the proposed method achieves the current SOTA. 4) The paper is well-written, and the supplementary material includes the code.
Weaknesses
See the detailed questions.
Questions
1) Are other advanced clustering algorithms more effective in enhancing reconstruction performance? This comparison could provide valuable insights into the optimal algorithm selection for this specific task. 2) For the ablation studies in Table 5 (Figure 5 should be Table 5), the authors are suggested to add two more settings, i.e., remove (b) and (c) and keep (a); remove (c) and keep (a) and (b), to directly validate the adaptive rank regularization, which is the main contributions. 3) In Figure 6, please add the input and GT images. It is better to use different symbols to distinguish the learnable closed-form samples and the input of the forward process. 4) Can other regularization terms (e.g., the L1 loss between X_0 and X_t) improve the original diffusion model? 5) There is potential for the proposed reconstruction algorithm to benefit diffusion models in tasks beyond the current scope of the study. It would be beneficial for the authors to explore this further, extending the validations to other potential applications. At least the authors can discuss this issue to improve the paper. 6) There are some grammar errors, e.g., line 163: a an
Rating
7: Accept: Technically solid paper, with high impact on at least one sub-area, or moderate-to-high impact on more than one areas, with good-to-excellent evaluation, resources, reproducibility, and no unaddressed ethical considerations.
Confidence
5: You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
Soundness
3 good
Presentation
3 good
Contribution
4 excellent
Limitations
Maybe a little additional computation consumption during the training phase.
Summary
Aiming at better low-light image enhancement performance with diffusion models, this paper proposes a global structure-aware regularization utilizing the intrinsic non-local structural constituents of image data. An uncertainty map is incorporated into the diffusion model to ease the strict constraints on indeterminate regions. Experimental results demonstrate the effectiveness of the proposed components.
Strengths
1. The paper incorporates global structure regularization via matrix rank modeling into the diffusion models, showing significant performance improvement compared with state-of-the-art methods. 2. The paper is well-written, and the presentation is pretty good.
Weaknesses
1. The contribution is not convincing enough. The generality of the proposed global structure-aware regularization based diffusion models is not verified, and the main idea of the adopted uncertainty is not original in this paper. 2. Additionally, the main concern about this paper is that there are many unclear details. For example, what train data is used in this paper? What is the baseline setting in Table.5? Is the inference time of all the methods also performed on RTX 3080 GPUs? 3. More perceptual metrics (e.g., LPIPS, MUSIQ) could be reported for better evaluation.
Questions
Please see the strengths and weakness section.
Rating
5: Borderline accept: Technically solid paper where reasons to accept outweigh reasons to reject, e.g., limited evaluation. Please use sparingly.
Confidence
5: You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
Soundness
3 good
Presentation
3 good
Contribution
2 fair
Limitations
Several challenging cases are provided. The limitations of the proposed approach sholud be discussed in more detail.
Summary
This work proposes a diffusion-based low-light image enhancement framework that exhibits good performance in different benchmarks. A rank-informed regularization term during training and uncertainty-weighted loss is proposed.
Strengths
- This work proposes a diffusion-based low-light image enhancement framework that seems to exhibit good performance in different benchmarks.
Weaknesses
- The motivation of the "Non-local patch-based matrix rank modeling" seems not clear. According to my understanding, the proposed "Non-local patch-based matrix rank modeling" is only used during training. How does it make the model have the capacity to be aware of the global structure during inference? Besides, during the training using paired training data, what's the necessity of using the proposed method to measure the structure consistency between Xt−1 and X0? The difference be directly calculated using metrics like L1, L2, SSIM, or feature-level similarities. - How does the proposed method increase the training overhead? - Some formulation is not self-contained. For example, where is the definition of "\alpha_1^2" in Eq. (4)? How to set the value of k_t? - It's better to illustrate which dataset/subset of the dataset is used to train the proposed method and other competitors. Since it greatly affects the performance of the model. And including a comparison of inference cost is better.
Questions
- Please refer to the weakness section. - While the weighting strategy of the training loss is reasonable for idea cases. But for real-captured high-resolution images where small misalignments may exist, will it lead to some side effects? - Could you explain why the constraint need only be added to the singular value, while no need for the singular vector?
Rating
4: Borderline reject: Technically solid paper where reasons to reject, e.g., limited evaluation, outweigh reasons to accept, e.g., good evaluation. Please use sparingly.
Confidence
5: You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
Soundness
2 fair
Presentation
2 fair
Contribution
2 fair
Limitations
N/A
Summary
This paper presents a diffusion-based framework to enhance low-light images. They propose a global structure regularization, which leverages the intrinsic non-local structural constituents of image data, besides, they introduce an uncertainty-guided regularization technique, which relaxes constraints on the most extreme portions of the image. The result outperforms some SOTA methods in some low-light image enhancement datasets.
Strengths
The paper proposes a global structure-aware regularization scheme, which capitalizes on image intrinsic non-local structures and gradually adjusts the regularization strength according to the sampling steps. The method achieves good results in the LOL dataset.
Weaknesses
My main concern is about the novelty of the method, the uncertainty-guided regularization seems from[4] Although the author has done comparison experiments in some datasets. I think the number of these images is quite limited, I hope the paper can provide a comparison experiment in the Adobe-FiveK dataset[1], following the setting of[2] The method needs to resample 500 steps. it will cost a long time to infer an image. The paper should cite and compare with[3]. [1] Learning Photographic Global Tonal Adjustment with a Database of Input / Output Image Pairs [2] Underexposed Photo Enhancement Using Deep Illumination Estimation [3]Pyramid Diffusion Models For Low-light Image Enhancement [4]. Uncertainty-driven loss for single image super-resolution.
Questions
Why does the method choose DDPM? How about DDIM?
Rating
5: Borderline accept: Technically solid paper where reasons to accept outweigh reasons to reject, e.g., limited evaluation. Please use sparingly.
Confidence
5: You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
Soundness
3 good
Presentation
3 good
Contribution
2 fair
Limitations
See in weakness.
Thanks for the rebuttal.
Thanks for the rebuttal. The AC and reviewers will take this into account for the final rating.
Global structure
Thanks for the rebuttal. If the global structure is an important contribution, have you ever done more experiments to validate the effectiveness of your proposed framework with other established ones.
Comparison with other diffusion-based low-light image enhancement methods
We are pleased to hear back from the reviewer. Regarding the comparison with “other established ones,” we assume the reviewer is referring to the comparisons with other diffusion-based low-light image enhancement methods. Actually, we indeed made it. We refer the reviewer to the $4^{th}$ response to **Reviewer ts6H**, where we compared our method with the pyramid diffusion [C3], the latest diffusion-based method for low-light enhancement published in IJCAI’23. For your convenience, we have pasted some results here. | Methods | Architecture | Loss term | PSNR | SSIM | LPIPS | |------------|---------------|------------------------------------------|------|-----|----| | PyDiff [C3] | Pyramid | Vanilla with Multi-scale L1 | 27.090 | 0.879 | 0.100 | | Ours | Vanilla (U-Net) | Rank-based modeling with basic KMeans clustering | 27.336 | 0.874 | 0.097 | | Ours++ | Vanilla (U-Net) | Rank-based modeling with advanced Hierarchical clustering | **27.697** | **0.880** | **0.092** | Additionally, when comparing other augmentation strategies or loss terms, we've evaluated our method's performance against L1, L2, and Perceptual losses. Please see our 2nd response to **Reviewer gyFW** for further comparison details. |Method| PSNR| SSIM| LPIPS| |--|--|--|--| | Baseline| 26.02| 0.859| 0.123 | | L1 Reg. | 26.45| 0.868| 0.101 | | L2 Reg. | 26.53| 0.869| 0.102 | | SSIM Reg.| 26.23| 0.870| 0.101| | Perceptual Reg.| 26.42| 0.863| 0.099| | Rank Reg. (Ours) | **27.02** | **0.872** | **0.097** | Last but not least, we also want to note that regarding our main contributions, the motivation/findings of introducing regularization between $X_{t-1}$ and $X_{0}$ is also essential , as clarified in the $2^{nd}$ response to **Reviewer gyFW** [C3] Zhou, Dewei, et al. Pyramid Diffusion Models For Low-light Image Enhancement, IJCAI'23.
My feedback
Thanks for the feedback. Diffusion has become mainstream for low-level vision. Does this mean low-light image enhancement based on diffusion is an easy idea to come up with? For any new papers on low-light image enhancement, we should set very high standards. In my opinion, this work is too generic. Such a method can be easily employed for other low-level vision problems like underwater image enhancement, image dehazing, or even denoising. We may be unable to find enough specific low-light image enhancement methods in this situation. It is better to say that this is a framework designed for general low-level vision problems, since global structure is also useful for other tasks. In this regard, the authors are suggested to conduct more experiments on aforementioned areas to investigate whether this proposed method is applicable.
Thanks again for your feedback. Based on the comments, it appears that we have addressed all technical issues you posed previously. Here's our detailed rebuttal to your queries regarding our contributions: We agree with you that introducing diffusion-based methods into low-light enhancement is **somewhat** easy to come up with. Our selection of this subject stems from our ongoing research project's needs and our sustained efforts in this area. Several **concurrent** works on diffusion-based low-light image enhancement have been proposed, such as Diffusion-prior in CVPR’23 [C1] and Pyramid diffusion in IJCAI’23 [C2]. They primarily emphasize on crafting novel architectures, yet don’t yield impressive performance. Thus, **how to adapt the diffusion process effectively and efficiently to low-light image enhancement remains a tough issue**. **By contrast**, we innovatively model the latent global structure in a low curvature trajectory reverse process (see the detailed clarification in the second response to **Reviewer gyFW**) that is strongly recognized by **Reviewer U5ii**, instead of simply introducing a vanilla diffusion model. As experimentally demonstrated, our method with the advanced patch grouping algorithm improves the SOTA work (SNR-aware) with 1 dB, and **significantly** suppresses the concurrent work Pyramid diffusion [C2] IJCAI’ 23 (see the 4th response to **Reviewer ts6H**). Second, we recognize that there are two distinct research guidelines: one that offers diverse yet suboptimal performance, and another that focuses on a specific domain yielding admirable results. **Both paradigms are meaningful and necessary**. We believe **it's unfair to underestimate the significance of our novel technique and notable improvement (by over 1 dB) just because its potential applications in other areas haven't been explored**. Moreover, we have fully exploited the potential of diffusion models for low-light enhancement using additional plug-and-play regularization term. **This, in itself, stands as a complete work.** Last, as we delved deeper into this work, we became increasingly aware that our theoretical insights and empirical findings of the plug-and-play regularization to the diffusion process may benefit various diffusion-based low-level image restoration/reconstruction tasks. However, if we were to achieve the same magnitude of performance improvement in other areas as we did in low-light enhancement, we believe the workload would surpass that of a single conference submission. But, as evidence of the generality of our method, we have **indeed** conducted super-resolution experiments in the first response to **Reviewer 9ew6**. We also copy the results here for your convenience. | Methods | PSNR | SSIM | |------------------------------------------------|-------|-------| | IDM [C3] | 24.01 | 0.710 | | Ours w/o global structure-aware regularization | 23.94 | 0.713 | | Ours | 24.52 | 0.728 | [C1]B. Fei, et al., Generative Diffusion Prior for Unified Image Restoration and Enhancement, CVPR'23. [C2]D., Zhou, et al. Pyramid Diffusion Models For Low-light Image Enhancement, IJCAI'23. [C3]S. Gao, et al., Implicit diffusion models for continuous super-resolution, CVPR'23.
My feedback
Please do not make any assumptions on my behalf. I am bringing up this concern because I believe it is more significant as it pertains to the motivation behind this work. This does not imply that my other concerns have been resolved. Second, it would be unjust to claim that other works lack promising performance. As a highly competitive research field, I do not believe that their papers would be accepted by reviewers if they did not demonstrate promising performance. Since some ICCV papers has been released on arxiv, we can find more low-ligh image enhancement papers on diffusion (ignore this sentence, since ICCV is still on its way.) Third, it is important to acknowledge the previous works that have shown remarkable and promising performance in utilizing diffusion for low-light image enhancement. It is unclear whether you have reviewed these papers during the preparation. Discovering the potential of established structures in new areas is crucial for the advancement of a field. If you are publishing a paper on top of these works, it is essential to have significant results to demonstrate. Otherwise, what would be the purpose of publishing the article? Upon reviewing your response to other reviewers regarding the non-local patch, I have a question. I recall that the concept of utilizing non-local priors has been extensively applied in various image restoration problems, even in the use of convex optimization and sparse coding. What are the primary advantages of your approach in comparison to previous methods in this area? At that time, non-local prior is also a regularization term in covex optimization equation, just like the loss functions in this place. Moreover, I do not make any judgements about whether your architeture is novel or not. Besides the novelty, I just say that I feel this structure is rather generic.. At last, let us say something beyond this paper. From the usage of the large-scale models, it seems that "one that offers diverse yet suboptimal performance" may be a mainstream in the future.
Thanks for the additional feedback, we will try to resolve all your concerns. **Performance of SOTA diffusion methods**. We want to claim that there are still potential for applying diffusion-based network to low-light enhancement. At their submission stage, their performance was indeed promising and admirable. However, with time passing, the requirement for high performance is gradually increasing. Especially, we have found that applying global structure regularization could greatly improve diffusion model performance. Moreover, our method is **the best method among all the compared methods**, we acknowledged, upon the most widely adopted datasets. **Other SOTA methods**. We have made our best to find SOTA methods, e.g., LLFlow AAAI'22, LLFormer AAAI'23, and SNR-Aware CVPR' 22, which can representatively show the state-of-the-art methods. There may be some papers that we have overlooked, but it is impractical for a single paper to review all published related work. **Advantage of using global structure-regularization**. Our contribution/ advantage mainly lies in introducing the global-structure modeling process into diffusion models. We have developed a method to **seamlessly integrate non-local priors into the diffusion model**, which results in superior performance. While there are indeed various methods to embed non-local priors, such as the optimization-based methods you discussed, these cannot be adaptively incorporated into the diffusion model, or at least, **no existing literature indicates this**. Moreover, in the realm of image reconstruction, there are currently no works attempting to merge classic image priors into diffusion models. **A generic method**. Yes, our method could be generally applied to other image reconstruction problems. We have validated their superior performance in low-light enhancement and super-resolution. We firmly believe that our contribution is substantial enough for a conference submission, which also is strongly recognized by **Reviewer U5ii**. Furthermore, we will also validate their performance in other related fields. **Large-scale models**. We agree that large-scale models represent a promising approach. However, other specialized models also hold significant value in the computer vision community. Their practical usefulness in real-world applications cannot be overlooked. **Last but not least, we are pleased to address your remaining unsolved technical concerns to enhance the quality of the paper.**
Comments
Thanks for your paper. I may need to consider my final rating at a later stage. You mention that your global structure-regularization is "seamlessly integrate," but it's just a loss function. Your specific design can just be found in Line 157. However, it only considers the step-by-step generation of the diffusion process. It is not an appropriate idea to call it "seamlessly". Lines 146 to 156, however, are not new to me. This part is actually very similar to a lot of classical papers published about eight years ago. Such methods have been attacked many times when deep learning dominates the society. Besides, please cite more papers in Line 146 to Line 156, since you are not the first to come up with such ideas. You cite those paper in Line 113, which may make readers feel that you propose those ideas alone..
Rebuttal by Authors
Thanks for the further feedback. **Seamless integration**. We wish to emphasize that integrating regularization terms into the training phase of diffusion models presents inherent challenges. During the training phase, the network's output is **just the latent noise map** instead of a clean image. This demands **careful consideration of multiple factors** for effective regularization. For instance, optimal placement of regularization between variables (whether between $\hat{X_0}$ and $X_0$ or between $\hat{X}_{t-1}$ with $X_0$, et al., where hat means the reconstructed one), the potential necessity of introducing noise into the closed-form samples analogous to the reverse process, and the dynamic coordination of regularization weights. Although our findings are presented in a concise manner, they consist of our extensive research efforts. We've appended the relevant experimental results for your review. |Constraint items| PSNR | SSIM| |:-|-|-| | Baseline: W/O regularization | 26.02 | 0.859 | | + Regularization between $\hat{X}_{0}$ and $X_0$ | 26.53 | 0.865 | | + Reg. btw. $\hat{X}_{t}$ and $X_0$ | 26.39 | 0.862 | | + Reg. btw. $\hat{X}_{t-1}$ without noise and $X_0$ | 26.61 | 0.867 | | ... | ... | ... | | + Ours: $\hat{X}_{t-1}$ and $X_0$ | 27.02 | 0.872 | In addition, we've provided a theoretical analysis of our regularization term in response to **Reviewer gyFW**. For ease of reference, we replicate our response below. ".......Analysis of applying regularization between $$\hat{X}_{t-1}$ and $X_0$, from two distinct perspectives. **a) Trajectory Curvature.** Recent works [C1, C2] show that an efficient and effective ODE-based generative model, e.g., diffusion models, should have a *lower curvature (straight) trajectory*, which could help them stably converge to high-quality solutions. Our regularization term could indeed minimize the trajectory curvature. We further illustrate such properties by corresponding experimental results of the reverse trajectory curve in the additional pdf file. **b) Accumulated Error.** During inference, each step is established on the previous prediction. Thus, accumulated errors inevitably exist in the intermediate reconstruction. Through regularizing $$\hat{X}_{t-1}$ towards the ground truth, we expect that each reconstruction step would consistently move towards GT images, which may also potentially minimize the influence of such accumulated error. **2) Analysis of choosing rank-based global-structure aware regularization.** While the current approaches in image processing often involve pixel-wise treatments, inherent global or latent structures - typically represented by non-local similar patches - can be overlooked by existing models to some extent. Acknowledging and modeling such global structures potentially augments the various reconstruction algorithms' performance [C3]. Additionally, conventional pixel-wise regularization terms, such as L1, L2, and SSIM, do not adequately encapsulate an image's nonlocal structures or patterns. Although regularization within the feature domain potentially aids in modeling a given image's structure, such structure is usually confined within a local region due to the kernel perceptual regions. This type of regularization, moreover, cannot explicitly characterize the properties of the structure and is lacking in theoretical guidance. Another significant concern is that regularization features could result in significant fluctuations in the back-propagated gradients, thereby impeding network training. [C1] S. Lee, et al., Minimizing trajectory curvature of ode-based generative models, ICML'23. [C2] X. Liu, et al., Flow straight and fast: Learning to generate and transfer data with rectified flow, ICLR'23. [C3] S. Gu, et al., Weighted nuclear norm minimization with application to image denoising, CVPR'14. ......." **Global-structural modeling V.S. deep learning techniques**. The authors agree that rank-based structural modeling was prevalent before the deep learning era. With the emergence of deep learning techniques, those methods become less popular. However, each method has its uniqueness and strength. It's more plausible for us to make use of each potential to build a robust and effective method. How to make use of past mature knowledge/ techniques to direct or boost current work also show very significant importance for the research. **Citation** We will ensure further citations are incorporated where indicated. Our deep respect for preceding research is evinced by our extensive list of references, spanning three pages. Furthermore, we consistently strive to find relevant work, regardless of whether it's concurrent or from earlier studies. **Finally, we hope our work could give somewhat inspirations to the latent structure in the diffusion process and offer potential avenues for designing regularization terms. We appreciate your further consideration. THANKS.**
Response to the rebuttal
I greatly appreciate the authors for their exceptional efforts during the rebuttal phase. I am pleased to note that all of my concerns have been thoroughly addressed. The remarkable achievement of significantly enhancing performance through the utilization of advanced clustering algorithms for grouping image patches in adaptive rank-based regularization is truly impressive. I am also impressed by the authors' clear and well-justified explanations regarding the motivation behind the proposed method, as highlighted in their second response to **Reviewer gyFW**, as well as the superior performance demonstrated compared to state-of-the-art methods. Additionally, after carefully reviewing the comments from my fellow reviewers, I find that the authors' responses effectively address all the questions raised. Drawing from my five years of experience in this field, I hold great confidence in the novelty of the proposed method and believe that its elevated performance level will significantly contribute to the advancement of this research area. Therefore, I believe the paper quality sufficiently meets NeurIPS's standard although I also would like to see the other reviewers’ responses. Furthermore, as previously mentioned, I am convinced that the idea presented in this paper holds potential for other diffusion-based low-level image/video processing tasks, which is somewhat validated by the additional experiments conducted on image super-resolution, as detailed in the authors' response to **Reviewer 9ew6**. In preparing the final version, I highly recommend the authors incorporate the necessary responses as discussed during the rebuttal phase.
We sincerely appreciate your recognition of our work and effort. We will ensure that the final version will include all the necessary information to make it comprehensive and complete. Thanks.
Looking forward to hearing back from you
Dear **Reviewer gyFW** Thank you for taking the time to review our submission and providing us with constructive comments. We would like to inquire if our responses have adequately addressed the concerns you raised earlier. Additionally, if you have any further concerns or suggestions, we would be more than happy to address and discuss them in order to enhance the quality of the paper. We eagerly await your response and look forward to hearing from you. Best regards, The authors
Looking forward to hearing from you
Dear **Reviewer 9ew6** Thank you for taking the time to review our submission and providing us with constructive comments and a favorable recommendation. We would like to know if our responses adequately addressed your earlier concerns. Additionally, if you have any further concerns or suggestions, we would be more than happy to address and discuss them to enhance the quality of the paper. We eagerly await your response and look forward to hearing from you. Best regards, The authors
Looking forward to hearing from you
Dear **Reviewer ts6H** Thank you for taking the time to review our submission and providing us with constructive comments and a favorable recommendation. We would like to know if our responses adequately addressed your earlier concerns. Additionally, if you have any further concerns or suggestions, we would be more than happy to address and discuss them to enhance the quality of the paper. We eagerly await your response and look forward to hearing from you. Best regards, The authors
Thanks for your rebuttal. I think you should add a comparison with MAXIM: Multi-Axis MLP for Image Processing(CVPR' 22), which can achieve 26+ PSNR on the AdobeFiveK. It seems to have a 2+ db gap.
We appreciate the feedback received. In light of the previous comments, we conducted a comparative analysis in alignment with the original configuration proposed by [2]. Our evaluation encompassed **all 500 test images at their native resolutions**. It's noteworthy that MAXIM compared **only 400 test images and subjected them to cropping and resizing to the resolution of $512 \times 512$**. Consequently, **drawing a fair comparison with their published results proves difficult**. Given our time constraints, our present model on the FiveK dataset has been adjusted in a preliminary manner. With a more meticulous calibration, there is potential for enhancing our results further. It is also worth mentioning that we are devising global structure-aware loss terms to explore the potential of the diffusion model. Thus, advanced backbones, e.g., the mentioned MAXIM, could be integrated into our framework to replace the vanilla U-Net for further performance improvement.
Thanks for the rebuttal. For W2-1, my concern is whether the proposed method is trained only on LOL-v1 or separately on LOL-v1 and LOL-v2 due to the ambiguity stated in Section4.1. If the latter is the case, then for fair comparison, the methods in Table1 that only provide pretrained weights on LOL-v1 should be pointed out separately. Also, will you release the source code to ensure reproducibility?
We appreciate the feedback received. In accordance with the settings outlined in **the very recent** works, i.e., SNR-Aware (CVPR'22), LLFlow (AAAI'22), and LLFormer (AAAI'23), we also trained our models on LOLv1 and LOLv2 datasets separately. Regarding the remaining methods, we derived the most promising results from relevant papers (e.g., LLFlow, LLFormer, and SNR-Aware), ensuring consistent settings for fair comparisons. We will make this clear in the final version. Besides, it is worth noting that **we have conscientiously included the source code within the supplementary material**. Additionally, we will make our code and pre-trained models publicly available, with the aim of facilitating others in reproducing our results.
Thanks for your reply. How did you test the LOLv2-real dataset since according to the attached code the model is not adapted for the image with the resolution 400*600? Besides, the name of the released checkpoint is "LOLv2-syn.pth". Does it mean you train two separate models for LOLv2-real and LOLv2-syn?
Official Comment by Authors
We appreciate the feedback received. For testing on LOLv2-real, we resized the input image (400×600) by padding it to the size of 416×608 and cropped the output back to its original dimensions. Besides, we indeed train two separate models for the LOLv2-real and LOLv2-syn datasets, consistently following the settings outlined in the very recent work, i.e., SNR-Aware (CVPR'22), which conducted extensive experiments on these datasets. Lastly, we will update and make our code and pre-trained models publicly accessible to facilitate the reproduction of our results by others. We sincerely thank you once again for the efforts to assist in improving the quality of our work.
Decision
Accept (poster)