Hyperspectral Image Super-Resolution Method Based on Spectral Smoothing Prior and Tensor Tubal Row-Sparse Representation
Round 1
Reviewer 1 Report
Dear authors, paper is interesting and solving a challenging problem: I have following recommendations:
- Decompose introduction into two sections i.e., Introduction and Related work.
- Add suitable citations in Section 2 from where you have adapted equations.
- How you have evaluated the hyper parameters of the model?
- If hit and trial approach was used then I suggest to add in future work regarding the optimization of hyperparameters. Refer remote sensing images using dynamic differential evolution
- Provide more details of Eqs. (13) and (14).
- Add more details about optimization algorithm in Section 3.
- Try to increase the text in images for better visibility.
Author Response
Please see the attached file for details.
Author Response File: Author Response.docx
Reviewer 2 Report
Since my comments are addressed, thus I recommend it for publication.
Author Response
Reviewer 2:
Since my comments are addressed, thus I recommend it for publication.
Response: Thanks for your positive comments.
Reviewer 3 Report
Dear authors,
thanks for interesting paper. I have only one recommendation. Please move description of used data from results into beginning of previous chapter, with some extension of this part.
It will help to read and understand better theoretical part
Author Response
Reviewer 3
Dear authors,
thanks for interesting paper. I have only one recommendation. Please move description of used data from results into beginning of previous chapter, with some extension of this part.
It will help to read and understand better theoretical part.
Response: Thanks for your positive comments. Indeed, the suggestions are of great help to our manuscript. But it should be noted that most super-resolution methods put the introduction of the dataset at the beginning of the experimental results. Thank you again for your sincere guidance on our paper.
Round 2
Reviewer 1 Report
Authors have made the necessary changes.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
In the manuscript, the authors have proposed a hyperspectral image super-resolution method based on fusion. I think this manuscript still has a lot of room for improvement.
- In the contribution part, for the tensor sparse representation, the difference or relationship between the manuscript with [1] should be clarified.
[1] Denoising and Completion of 3D Data via Multidimensional Dictionary Learning, IJCAI
- The convergence of the proposed algorithm should be clarified.
- Some state-of-the-art methods published in recent years should be compared in experiments. Besides, the authors have discussed some deep learning (DL)-based methods and I suggest comparing the proposed method with the DL-based methods, which could better verify the superiority of the proposed method.
- The Introduction section is mussy. The authors divided the existing methods into four kinds and I think the classification standard should be given. Besides, I think the authors should discuss the existing work following one stream.
- Many citations are inaccurate and all citations should be double-checked. For example, the references of CNN, and Tucker, CP, TT, and TR decompositions. To the best of my knowledge, the most related decomposition of the Tensor-tensor product decomposition should be tensor SVD proposed by Misha E, K. I suggest the authors read more references and cite the references with the origins of the results.
- Line 176: Mathematically, the caption is not appropriate.
- The reference of Lemma 1 should be given.
- Line 204, Line 208, Line 380: some mistakes. The whole manuscript should be double-checked.
Author Response
see the attached file
Author Response File: Author Response.docx
Reviewer 2 Report
Paper is interesting and authors come up with a good solution. I have following recommendations for the authors:
1. Explain tensor sparse representation. From present paper, it is difficult to obtain the information, uses and role of tensor sparse representation in the paper.
2. Improve the quality of the figures like text size in Figure 8 and Figure 10 should have more explanation.
3. To justify the proposed model, more comparative analyses are required. From present analyses difficult to evaluate the contributions of the authors.
4. Conclude paper with quantitative findings like evaluate the degree of improvement in terms of various performances metrics. Better if authors can use some statistical testing analyses.
5. Better to present a future work and also mention some suitable limitations of the proposed model.
6. Convergence speed analyses are not presented in the paper which is key to evaluate the performance of the proposed model.
7. How authors have selected the hyper-parameters of the proposed model and associated algorithms. I think parameters are selected purely on random or trial and error basis?
8. Authors should improve the introduction by removing the irrelevant literature and more focus on recently published literature.
9. Besides there are many grammatical and typo mistakes in the paper. Some equations symbols are also confusing.
Author Response
see the attached file
Author Response File: Author Response.docx
Reviewer 3 Report
I'm not the expert in this field (these methods). For me lack of complexity comparison of presented method and others, also nice to have some results comparison with DL based methods. Some discussion about other possible optimization techniques, future work is expected and some enhancements? .Some more citations needed (e.g. ADMM). Some improvements in language (native).
Author Response
see the attached file
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
1. The model (13) in the manuscript is similar to the proposed model (18) in Ref. [1]. It seems like this work is an increment work of Ref. [1]. I think Ref. [1] should be discussed and the novelty should be further clarified.
[1] Nonlocal Patch Tensor Sparse Representation for Hyperspectral Image Super-Resolution.
2. The notation of tensor variables and the notation of operations should be distinguished.
3. English language should be further improved.
Reviewer 2 Report
The authors have addressed all my concerns.