A General Deep Learning Point–Surface Fusion Framework for RGB Image Super-Resolution

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors proposed a novel method, GRSS-Net, for RGB image spectral super-resolution that addresses limitations in current spectral super-resolution methods, utilizes hyperspectral point data, incorporates compressed sensing theory, and combines deep learning with traditional models. The proposed method is experimentally validated and provides a flexible framework for diverse application scenarios.
 
 Comments:

 1) Only two sources are cited in the Related Work section. This type of section should be more comprehensive.
 
 2) The ablation test shows that the created model is superior to the pure physical and deep learning models. However, the qualitative assessment is not outstanding compared to four deep learning-based methods. It would be valuable to present suggestions for enhancing the spectral super-resolution effectiveness of the proposed method in the Conclusions section.

 

Author Response

Thank you for taking the time to review our manuscript. Thank you for your valuable and constructive comments. Following your suggestions, we have made the necessary revisions, and we believe that these changes have considerably strengthened the manuscript.

Below, we provide a detailed response to each of comments:

1) Only two sources are cited in the Related Work section. This type of section should be more comprehensive. 

R: Thank you for your useful suggestion. We have added content related to gradient descent method, as well as related reference. So far, all basic methods related to the proposed method have been introduced in this section.

It is worth noting that, due to content restructuring, original section 2 (related work, now is subsection 2.1) and section 3 are integrated into current section 2.

2）The ablation test shows that the created model is superior to the pure physical and deep learning models. However, the qualitative assessment is not outstanding compared to four deep learning-based methods. It would be valuable to present suggestions for enhancing the spectral super-resolution effectiveness of the proposed method in the Conclusions section.

R: Thank you for your insightful suggestion. We have added suggestions for enhancing the effectiveness of the proposed method in the Conclusions section. The supplementary content is as follows.

In practical application scenarios, the optimization method, the basic image observation model and deep learning-based module in GRSS-Net can be replace flexibly according to different application task to improve the actual super-resolution effectiveness.

Moreover, we have added additional discussion section to further verify the adaptability of the GRSS-Net framework to various application scenarios, which is more general than other deep learning-based methods. More specific results can be seen in Section 4 with the associated experiments.

Reviewer 2 Report

Comments and Suggestions for Authors

Even though I am not a developer specialized in that area of expertise, I think the paper is well written and the authors presented a good analysis. I have no questions about the proposed method and experiment validation, but some minor problems have to be solved before publication.

1.     All the images are not clear enough especially the words inside.

2.     There are some small mistakes, such as there should be a space between 4) and “More”.

3.     If you used “,” after a formula, I think the “Where” in the next line should be “where”.

4.     Section 3 and subsection 3.3 share the same name.

5.     Figure 5 and Figure 7 can be further integrated.

Overall the topic is very interesting.

 

Author Response

Thank you for taking the time to review our manuscript. We sincerely appreciate your insightful suggestions, which have undoubtedly enhanced the quality of our manuscript. We have fully considered each comment and the point-by-point, item-by-item responses are as follows: 

1. All the images are not clear enough especially the words inside.

R: Thank you for your useful suggestion. We have improved the quality of all images in this manuscript. We ensure that every image now can clearly distinguish the words inside.

2. There are some small mistakes, such as there should be a space between 4) and “More”.

R: Thank you for your kind suggestion. We have corrected it. To avoid these minor mistakes from happening again, we have conducted multiple checks of our manuscript.

3.If you used “,” after a formula, I think the “Where” in the next line should be “where”.

R: Thank you for your detailed suggestion. We have rectified it.

4. Section 3 and subsection 3.3 share the same name.

R: Thank you for your constructive suggestion. We have changed the title of subsection 3.3 to ‘The architecture of GRSS-Net’, which is more consistent with the content of this subsection.

It is worth noting that, due to content restructuring, original subsection 3.3 is now subsection 2.4, and the original section 2 and section 3 are now integrated into section 2.

5. Figure 5 and Figure 7 can be further integrated.

R: Thank you for your suggestion. We have improved images.

Reviewer 3 Report

Comments and Suggestions for Authors

In this paper, the authors propose a new RGB spectral super-resolution reconstruction method, named GRSS-Net, which utilizes hyperspectral image points instead of hyperspectral images as auxiliary data to provide spectral reference information. Utilizing compressed sensing theory as the foundational mechanism, the proposed method expands the traditional hyperspectral image reconstruction optimization problem into a deep network to obtain high spatial resolution hyperspectral images. Generally speaking, this manuscript is well-motivated and the approach is innovative. However, there are some problems to be solved. 

1. The EXPERIMENT section currently relies solely on simulated RGB  data and hyperspectral point data extracted from hyperspectral data for comparative analysis. It is recommended to incorporate actual hyperspectral point data and real RGB remote sensing images for experimental validation.

2. The concluding sentences of both the abstract and conclusion sections suggest the adaptability of the GRSS-Net framework to various application scenarios. This aspect warrants further exploration and discussion within the paper.

3. The structure of the paper could be more aligned with the journal’s standards. Specifically, the RELATED WORK section could be condensed or integrated into Section 3 for greater coherence. Additionally, a separate DISCUSSION section is advised to provide a comprehensive analysis of the proposed method’s performance.

4. In the Introduction section (page 1, line 39), the authors note that snapshot hyperspectral imaging technology is still evolving. However, this should not be misconstrued as a flaw or problem with hyperspectral imaging itself. Please revise the content accordingly.

5. In the Abstract and Introduction section, the background of RGB spectral super-resolution needs clearer and more accurate elaboration. For instance, the authors have not mentioned the limitations of the hyperspectral imaging platform, a key factor necessitating spectral super-resolution techniques. 

Comments on the Quality of English Language

 There are several grammatical  errors throughout the manuscript. Please thoroughly proofread to correct these issues. For instance, "by scanning-based way" on page 1, line 10, should be revised to "in a scanning-based way".", "to provides" should be "to provide"( page 1, line 18), "recover the spectral..." should be "recovering the spectral..." ( page 1, line 42)...

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for your detailed response to the questions and suggestions provided in my previous comments. I have noticed a marked improvement in the quality of the manuscript. I don’t have more questions except for some minor grammar problems.

Comments on the Quality of English Language

There are still some minor word errors and grammatical issues that require attention. For instance:

1.      In lines 66, 72, and 74, "et al" should be punctuated as "et al."

2.      Please ensure the proper use of the definite article "the" throughout the manuscript

3.      On line 320, the word "finial" should be corrected to "final."

4.      On line 409, the phrase "simple than" should be amended to "simpler than."

5.      On line 566, "be replace" should be revised to "be replaced."

Author Response

We sincerely thank reviewer for patiently and carefully reviewing our manuscript again. We have made revisions to each question you raised and indicated it in red font.

The detailed, point-by-point responses are as follows.

There are still some minor word errors and grammatical issues that require attention. For instance:

1. In lines 66, 72, and 74, "et al" should be punctuated as "et al."

R: Thank you for your suggestion. We have corrected it.

2. Please ensure the proper use of the definite article "the" throughout the manuscript

R: Thank you for your suggestion. We have also checked full manuscript and made additions and deletions to ensure the proper use of the definite article 'the'.

3. On line 320, the word "finial" should be corrected to "final."

R: Thank you for your suggestion. We have corrected it.

4. On line 409, the phrase "simple than" should be amended to "simpler than."

R: Thank you for your suggestion. We have corrected it.

5. On line 566, "be replace" should be revised to "be replaced."

R: Thank you for your suggestion. We have corrected it.

We have checked our manuscript several times and made every effort to correct any minor errors that could be found.