FASwinNet: Frequency-Aware Swin Transformer for Remote Sensing Image Super-Resolution via Enhanced High-Similarity-Pass Attention and Octave Residual Blocks

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

It was my pleasure to read and review this comprehensive manuscript, where the authors introduce a novel frequency-aware super-resolution network for remote sensing images aimed at solving the resolution increase problem. The paper is well-written and possesses some interest for readers. However, I formulated several questions and recommendations to improve the manuscript.

1. Is Figure 1 original (prepared for this manuscript exclusively) or taken from somewhere? If it is original, it will be beneficial to reveal how it was plotted. If not, please provide a direct reference in the figure's caption. In addition, the Figure 1 is mentioned in the text much later than it appears.

2. The problem statement lacks sources, e.g. in "Nevertheless, most existing methods primarily emphasize global context modeling while paying insufficient attention to high-frequency textures, edge structures, and frequency-aware feature fusion". Please, make it clear that mentioned problem really has not been addressed yet.

3. The model limitations are to be discussed further and in more detail.

4. I recommend paying more attention to computational complexity analysis. The authors discuss parameters and FLOPs in table 4, but it does not cover real execution times and memory usage. The comparison with other models should be given taking into account the fact, that the suggested model possesses more parameters than competing architecture.

5. It is hard to acquire the quality of each tested method from images given in Fig. 3 by the naked eye. Using quantitative metrics or calculating picture differences may help address this issue. Otherwise, the statement that it is "showing that FASwinNet achieves clearer texture and edge restoration" is a bit speculative.

6. Did you use any AI text generation\editing in this manuscript? Some terms, like "high-definition images," appear to be a term misuse typical for LLM generators.

Nevertheless, I believe the paper possesses necessary value for a separate peer-reviewed publication and can be accepted after necessary revisions.

Comments on the Quality of English Language

"width of the image respectively, , " - extra comma

" to process highlow and low- frequency feature" - possible typo?

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper presents a valuable contribution to the artificial intelligence-based concepts for high-resolution image analyses for remote sensing. The proposed idea of an enhanced frequency-aware super-resolution network is somewhat novel and has great technical importance.

However, the paper is written in a chaotic manner. Abbreviations should be clearly defined (e.g., PSNR) even if they are widely accepted by the specialists. Moreover, Figure 1 appears in the introduction, whereas it is cited in the text of Section 3.1, “Network architecture.” As a result, the manuscript has to be reorganized to have a clear structure, which is required for a scientific paper.

Moreover, the training method must be clearly specified. It is unclear why only 2 images were selected for testing. Please explain if such a split is sufficient and why.

The paper presents a valuable and comprehensive comparison of the results against the state of the art. This is one of the manuscript's most significant strengths. However, resenting parameters like PSNR or MSE with four-digit accuracy seems to be obviously unjustified.

Figure 3 presents a comparison, which is valuable during the analyses. However, considering the printing quality or quality of the archived PDF file in the Journal, I highly recommend reorganizing this figure. Please select the most important subplots and enlarge them to highlight important differences.

Finally, the conclusions should be rewritten and reorganized. Please clearly state the most important outputs of the paper, preferably in the form of bullet points and in quantitative form. Indicate the most important directions of further research and development.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors! Thank you very much for revising your manuscript following my suggestions. I am completely satisfied with the revised paper and point-by-point reply letter. I see no reason for delaying the publication of this paper and wish the authors good luck in their future studies.

Sincerely,

Reviewer

 

Reviewer 2 Report

Comments and Suggestions for Authors

The paper was corrected and can be accepted in the present state.