A Constrained Convex Optimization Approach to Hyperspectral Image Restoration with Hybrid Spatio-Spectral Regularization
Round 1
Reviewer 1 Report
Although the work is significant, there does not seem to be anything new and significant with respect to the two publications by the same authors in both ICASSP and ICIP.
In fact they are two very reputable conferences.
Reviewing both works, I honestly think that there is no remarkable novelty. Anyway, the authors should have included a much more detailed report on this aspect.
I have marked "low" in all items because the work submitted has already been presented and published and also in duplicate
Author Response
Dear. reviewer 1
Thank you for your careful reading of our manuscript and for giving valuable comments. We have carefully studied all the comments given by you and made revisions, which we hope will meet your approval. Please kindly see the attachment.
Best Regards,
Saori Takeyama
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Well done. The work can be improved by adding some information about the computational methods used by the authors. Have the authors written a library in some language or they have used an existent one? Please give details.
Author Response
Dear. reviewer 2
Thank you for your careful reading of our manuscript and for giving valuable comments. We have carefully studied all the comments given by you and made revisions, which we hope will meet your approval. Please kindly see the attachment.
Best Regards,
Saori Takeyama
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper proposed a novel hybrid spatio-spectral total variation (HSSTV) method for hyperspectral image restoration. The hybrid SSTV consists of two parts: direct local spatial differences and local spatio-spectral differences of an HSI image. The idea is interesting and somewhat novel. The manuscript is organized and written well, the literature review is insufficient, many low rank or tensor low rank methods related to hyperspectral image restoration are not mentioned in the introduction part. The technical issues are all correct, and extensive experiments also provide convincing results. However, some major problems should be solved before publication.
1. Some low rank or tensor low rank works should be reviewed in the introduction. For instances,
[1] Low Rank Component Induced Spatial-spectral Kernel Method for Hyperspectral Image Classification, IEEE Transactions on Circuits and Systems for Video Technology, 2019, online, DOI:10.1109/TCSVT.2019.2946723.
[2] Weighted Nonlocal Low-Rank Tensor Decomposition Method for Sparse Unmixing of Hyperspectral Images, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 13, no. 1,pp. 1174-1188, 2020.
[3] SLRL4D: Joint Restoration of Subspace Low-Rank Learning and Non-Local 4-D Transform Filtering for Hyperspectral Image, remote sensing, 12(18) 2979, 2020.
2. More state-of-the-art restoration methods should be compared with the proposed method. In the experiment, most of the comparison methods are TV based ones, however, these days many HSI denoising methods are proposed and their codes are released, it is suggested that the author can employ more state-of-the-art denoising methods, especially those methods published within recent two years, for comparison.
3. In Table 3, the results of PaviaRight are highlighted wrongly.
4. Figures 3 7 8 are wrong during the conversion.
Author Response
Dear. reviewer 3
Thank you for your careful reading of our manuscript and for giving valuable comments. We have carefully studied all the comments given by you and made revisions, which we hope will meet your approval. Please kindly see the attachment.
Best Regards,
Saori Takeyama
Author Response File: Author Response.pdf
Reviewer 4 Report
In this paper, authors propose a novel constrained optimization approach, named Hybrid Spatio-Spectral Total Variation (HSSTV), to hyperspectral image restoration. HSSTV leverages both spatio and spectral informations and has a strong ablity of noise and artifact removal while avioding oversmoothing and spectral distorations. In addition, authors also propose an efficient algorithm to solve the optimization problem. In the experimental part, the proposed method gets a desirable results. I have several concerns here:
1. The experimental settings look strange. For ADMM is proposed for HSSTV, it is not suitable to use ADMM for other baseline methods. It is better to use original methods for baselines and use ADMM for HSSTV. Then, to demonstrate the effectiveness of ADMM, authors could make an ablation study on ADMM. Otherwise, authors need to make experiments to prove that ADMM will not influence the results of baseline methods.
2. The definition of PSNR in the bottom of page 9 should be 10log(MAX^2_I/MSE).
3. Several typos need to be checked. For instance, it should be Section 5.1 instead of Section V.A in the bottom of page 13.
4. Several important works are not referred. For instance,
Y. Gou, B. Li, Z. Liu, S. Yang, and X. Peng, “CLEARER: Multi-Scale Neural Architecture Search for Image Restoration,” presented at the Neural Information Processing Systems, Vancouver, Canada, 2020, pp. 1–11.
Author Response
Dear. reviewer 4
Thank you for your careful reading of our manuscript and for giving valuable comments. We have carefully studied all the comments given by you and made revisions, which we hope will meet your approval. Please kindly see the attachment.
Best Regards,
Saori Takeyama
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Analyzing the response of the authors, unfortunately I consider that the novelty introduced in the article regarding the commented publications is not enough to generate a new publication. In addition, two aspects must be taken into account, the first is that the authors have responded but have not introduced this information in the article (perhaps because it detracts from the article) and secondly, as I said, the conferences are of sufficient quality and impact not to have to generate a new publication with almost the same content.
Author Response
Dear. reviewer 1
Thank you again for your careful reading of our manuscript and for giving valuable comments. We have studied all the comments given by you carefully and made revisions, which we hope will meet your approval. Please kindly see the attachment.
Best Regards,
Saori Takeyama
Author Response File: Author Response.pdf
Reviewer 4 Report
I have no more suggestions.
Author Response
Dear. reviewer 4
Thank you again for your careful reading of our manuscript and for giving useful comments. We have checked and improved the English language and style of the revised paper by using a professional English proofreading service, which we hope will meet your approval.
Best Regards,
Saori Takeyama
