Hyperspectral Image Denoising Based on Deep and Total Variation Priors
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors* Literature survey needs to be expanded.
* Some missing references and typos needs to be carefully corrected. (L153)
* Conclusion needs to be expanded also.
* Figure 4-5-6 is not clear.
Comments on the Quality of English LanguageThis papers should be re-evaluated after corrections.
* Literature survey needs to be expanded.
* Some missing references and typos needs to be carefully corrected. (L153)
* Conclusion needs to be expanded also.
* Figure 4-5-6 is not clear.
Author Response
Dear Reviewer,
We greatly appreciate your time on our paper and clear instructions for revision.
We have carefully considered your comments and suggestion and have provided item-by-item response. We are sorry to make some small mistakes and have corrected them now. The whole paper has been revised accordingly and the revised parts are marked in green. We hope the new manuscript will meet your standard.
In the end, we want to convey our earnest thanks to you. Your valuable comments have greatly improved this manuscript.
Best wishes
The authors
Reviewer 2 Report
Comments and Suggestions for AuthorsHere are several suggestions to improve the manuscript:
1. The first paragraph of the introduction mentions that HSI inherently contains noise, but lacks citations. Please include references to survey papers or studies that discuss the origins and types of noise in HSI to strengthen your claims.
2. It is incorrect to represent mathematical symbols as figure on line 69. This should be formatted using LaTeX as $L_2$ or in equation mode in MS Word to maintain the manuscript’s professionalism and clarity. The same problem happens across the full manuscript. Several symbols are wrongly presented.
3. In Section 2, where TV is discussed, it would be beneficial to cite works that utilize variants of TV in HSI processing, such as those found at doi.org/10.3390/rs14163998 and doi.org/10.1109/WHISPERS56178.2022.9955140 to provide a comprehensive review of its applications.
4. Please review and correct lines 153-154 of the manuscript. Any formatting or content issues there should be addressed to ensure the manuscript meets the submission standards.
5. The format and content of algorithm table on line 394 needs to be improved. The algorithm should be structured in a way that clearly outlines the steps of the algorithm, allowing readers to grasp the overall method quickly. Ensure that all variables and steps are clearly defined and logically presented, contributing to an intuitive understanding of how the algorithm functions.
6. For Figures 1-3, ensure better alignment and presentation by utilizing a full line mode, and have 4 images on one line for example.
7. Considering that the two datasets currently used are city scenes, I recommend incorporating a dataset with a forest setting to enhance diversity and applicability of the algorithm to different environments. Datasets like Salinas or Indian Pines, which contain significant amounts of noise, could provide a more challenging and varied test scenario.
Comments on the Quality of English LanguageModerate edits are needed.
Author Response
Dear Reviewer,
We greatly appreciate your time on our paper and clear instructions for revision.
We have carefully considered your comments and suggestion and have provided item-by-item response. We are sorry to make some small mistakes and have corrected them now. The whole paper has been revised accordingly and the revised parts are marked in green. We hope the new manuscript will meet your standard. Please see the attachment.
In the end, we want to convey our earnest thanks to you. Your valuable comments have greatly improved this manuscript.
Best wishes
The authors
Author Response File: 
Author Response.doc
Reviewer 3 Report
Comments and Suggestions for AuthorsReviewer’s Report on the manuscript entitled:
Hyperspectral Image Denoising Based on Both Deep and Total Variation Priors
The authors proposed a framework for denoising hyperspectral images (HSI). Their method process the original HSI using a randomly binarized sparse observation matrix and use FFDNet and total variation simultaneously. They compared their denoising method with other methods. Though the methods and results are interesting and promising, I found the manuscript too abstract and technical. The presentation, figures, and literature must be improved. Please see below my comments.
Main comments:
Abstract must be re-written. Please start by giving one or two sentences about motivation and why HSI is important. Then state the objectives, methods, and results, and add one sentence at the end for recommendation/application/usefulness.
Similarly, the Conclusion section is too technical, and the general audience cannot follow. Please provide general and non-technical descriptions and motivations. Conclusions should start by mentioning the objective and why this research is important in plain language. In addition, please mention the limitations and future directions at the end of the conclusions.
Literature review must be improved.
Line 39. Adaptive weighting rank-reduction method using multichannel singular spectrum analysis algorithm is also used for denoising that can be added here (Doi: 10.3390/s23020577). Also, random noise attenuation in images using the antileakage least-squares spectral analysis (ALLSSA) can also be added here (doi: 10.1190/geo2017-0284.1).
Figures 1,2,3. Please add the residual images with a fixed value range for all the methods to aid comparison. The residual image is the original band, i.e., panel (a) minus the denoised image.
Figures 4,5,6 have several issues. The bluish colors are hard to distinguish between. For example,
“Degraded” and “Ours” have similar colors (hard to distinguish). Please use other colors. The legends should not overlap the bars. Please place them above the plot horizontally. You do not need one legend per panel. Since the legends of panel (a) and (b) are the same, simply use one legend horizontally above the entire figure. Please do this for all the three figures. The font size must be enlarged. The recommended front size of the texts and numbers in the figures should be the same size as the font size of figure caption.
Lines 502-503. Please move MPSNR, MSSIM, and SAM to the method section in a subsection called: 2.3. Evaluation Metrics. Then please mathematically describe them in more detail.
Line 540. Please have a separate full section for the discussion section not as a subsection. Then please update lines 94-95. Please also expand the discussion section by commenting on the results in the light of other studies and also mention the computational time of your method vs others listed in Table 3.
Eq. (19) Please comment on the use of L1 vs L2 norm in the discussion section. I think L2 is usually smoothing out while L1 maintain the sharp edges...
Also, what is the stopping criteria for denoising? Any thresholds and how they are selected. Please elaborate in the discussion section.
Other editorial comments:
Line 69. Style issue. Please fix the symbol L2.
Line 153. Error is the reference.
Line 165. I suggest replacing “remove” with “attenuate”.
Eq. (6). Please remove the comma.
The style and format of all the symbols and equations must be consistent and according to the MDPI guidelines.
Line 613. “mapping” not “map**”
I see more than 10 references are from “Remote Sensing” that is considered extensive. Please replace at least 5 of them with similar publications from other journals, such as
doi: 10.1016/j.jag.2023.103481 and from other publishers.
Thank you!
Regards,
Comments on the Quality of English LanguageThere are many typos/grammar/style issues in the manuscript that must be checked and corrected.
Author Response
Dear Reviewer,
We greatly appreciate your time on our paper and clear instructions for revision.
We have carefully considered your comments and suggestion and have provided item-by-item response. We are sorry to make some small mistakes and have corrected them now. The whole paper has been revised accordingly and the revised parts are marked in green. We hope the new manuscript will meet your standard. Please see the attachment.
In the end, we want to convey our earnest thanks to you. Your valuable comments have greatly improved this manuscript.
Best wishes
The authors
Author Response File: Author Response.doc
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper addressed most issues, but it can be further enhanced by enrich the introduction and related works.
Author Response
Dear Reviewer,
We greatly appreciate your time on our paper and clear instructions for revision.
We have carefully considered your comments and suggestion and have provided item-by-item response. The whole paper has been revised accordingly and the revised parts are marked in green. We hope the new manuscript will meet your standard.
In the end, we want to convey our earnest thanks to you. Your valuable comments have greatly improved this manuscript.
Best wishes
The authors
Author Response File: Author Response.doc
Reviewer 3 Report
Comments and Suggestions for AuthorsDear authors,
Thank you for addressing most of my comments satisfactorily and improving your manuscript. However, I still see several issues with writing, style, and correctness of figures and results that should be very carefully checked and corrected. Please see below:
Equations and symbols must be carefully checked. There are many many issues with them:
Line 283, 311, 326, 357, 404, Please fix the latex error.
Equation (6) has style issue,
Equation (9), Instead of saying "Vec" either use boldface or an arrow above the variable indicating Vec.
Equation (25). MPSNR should not be italic. It should be MPSNR. Please note that when a variable has multiple letters, you should not use italic font otherwise, it means the letters are variables multiplied together. Please do this for PSNR, MAX, MSE, SSIM, MSSIM, etc. Please check them very carefully.
Equations (30), (31), etc. What is "x" between P and P1. Is it the multiplication symbol? If so, please use latex command "\times" instead. There are many typo/format issues with the symbols and equations that must be very carefully checked and corrected.
Figures 1,2,3,4. Please remove "method", "technique", "strategy", "technique", "model", "algorithm"!!! LRMR, WSNM, LRTDTV, FGSLR, FFDNetm E3DTV are all methods by default so all these words are redundant.
Figure 4. Are we supposed to compare the result image of each method with panel (a) that is the benchmark image? If so, none of the methods are similar to (a). What is the ground truth panel?
Figs 6,7,8 are improved but please place the legend horizontally outside the panels. For example, the legend in Fig. 6 can be horizontally placed on top-outside of the panels such that it covers the full width of the figure, i.e., from left side of panel (a) to the right side of panel (b). Secondly both panels (a) and (b) should be nicely aligned. Please also enlarge the font size of the texts and numbers in these figures. Figures must be professional. I would like to see the new revisions.
References. Some of the references have missing information, such as page number, article number, doi, etc. Please provide full information for them.
Overall, the presentation, equations, and figures have many style/format issues that at present form are very confusing and must be corrected.
Thank you and regards,
Comments on the Quality of English LanguageThe presentation, equations, and figures have many style/format issues that at present form are very confusing and must be corrected.
Author Response
Dear Reviewer,
We greatly appreciate your time on our paper and clear instructions for revision.
We have carefully considered your comments and suggestion and have provided item-by-item response. We apologize for making some minor mistakes in the mathematical notation and figures, but we have now corrected them. The whole paper has been revised accordingly and the revised parts are marked in green. We hope the new manuscript will meet your standard.
In the end, we want to convey our earnest thanks to you. Your valuable comments have greatly improved this manuscript. Please see the attachment.
Best wishes
The authors
Author Response File: Author Response.doc
