Optical Remote Sensing Image Registration Using Spatial-Consistency and Average Regional Information Divergence Minimization via Quantum-Behaved Particle Swarm Optimization
Round 1
Reviewer 1 Report
This paper addresses image registration problem by developing a novel joint area-feature based method. The presented approach relies on intensity similarity/discrepancy metrics (SM/DM) to fine tune registration parameters by using an optimization process of registration parameters, according to a Quantum-behaved Particle Swarm Optimization strategy.
The approach is interesting; however, I would recommend this paper for publication only after it has been improved as suggested below:
1) Methods for remote sensing image registration depend on the type of image considered (e.g. Optical or microwave images, and pertinent required registration accuracy). Speckle affected images, for instance, pose distinctive difficulties in the SAR registration process. This distinction has not been emphasized in the manuscript, and the approach and the results in the paper seem essentially targeted to optical (e.g. multispectral, hyperspectral) image registration.
Therefore, the specific emphasis on “optical” image registration given in the manuscript should be reflected in the manuscript and in its title.
2) A detailed computational complexity of the proposed technique should be provided in details, also with respect the other considered methods.
3) The List of Acronyms in the paper do include both adopted symbols and Acronyms. It could be appropriate to consider two different lists.
Author Response
Responses to Reviewer 1: Comments
This paper addresses image registration problem by developing a novel joint area-feature based method. The presented approach relies on intensity similarity/discrepancy metrics (SM/DM) to fine tune registration parameters by using an optimization process of registration parameters, according to a Quantum-behaved Particle Swarm Optimization strategy.
The approach is interesting; however, I would recommend this paper for publication only after it has been improved as suggested below:
1) Methods for remote sensing image registration depend on the type of image considered (e.g. Optical or microwave images, and pertinent required registration accuracy). Speckle affected images, for instance, pose distinctive difficulties in the SAR registration process. This distinction has not been emphasized in the manuscript, and the approach and the results in the paper seem essentially targeted to optical (e.g. multispectral, hyperspectral) image registration.
Therefore, the specific emphasis on “optical” image registration given in the manuscript should be reflected in the manuscript and in its title.
Response:
Many thanks for excellent suggestions. We have added “optical” to and made modification on the paper title to reflect nature of the work presented in this paper. In addition, we also revised the sued images to optical images whenever it is appropriate in the text
2) A detailed computational complexity of the proposed technique should be provided in details, also with respect the other considered methods.
Response:
At suggested, we have included computational complexity analysis in a new section, Section 7: Discussions in this revised paper.
3) The List of Acronyms in the paper do include both adopted symbols and Acronyms. It could be appropriate to consider two different lists.
Response:
Many thanks for reviewer’s comments. We have added two lists to separate acronyms and symbols used in the paper.
Reviewer 2 Report
This work’s contribution for the scientific community lies in providing a novel image registration method named “Spatial-Consistency and Average Regional Information Divergence Minimization via Quantum-Behaved Particle Swarm Optimization”.
The full name is important as, reading the manuscript’s title, one could remind of other existing approaches. What sets this method apart from others already employed in remote sensing images registration, beyond the employment of Q-PSO, are the shown results’ accuracy and robustness.
Considering this, my first comment is directed to the manuscript title, which should be more accurate, even if necessarily longer.
The theme is current and interesting and well in line with this journal goals.
The research problem is fairly well defined. Even if there is not a profuse discussion on the matter, it is quite straightforward to be understood by the reader.
Also, there is not a deep literature review. However, such task is directed to other article by the same Authors. I like such approach, as there is no need to flood the scientific literature with repetition, but I believe that reading such article would be important for accurately reviewing this one. That has not been the case, since the Early access did not work, at least for me when I tried.
By the way, I cheer the Authors for their enthusiasm but I am afraid to say that the affirmation “[42] can be considered as a companion paper to this paper to form a nice pair of works” may not be very adequate in science writing and better left for the readers to judge.
Nevertheless, the introductory chapter is of good quality.
There are some self-citation, by the amount is not excessive and, more importantly, those look pertinent in the context.
Research methods are adequate but not deeply explained and frame. Section 5 could be much more developed.
The research beneath the manuscript is substantial and replicable to a certain degree On the other hand, one shall note that the experimental results are not as abundant as could be. More and, mostly, more heterogeneous datasets would strengthen the results. This would be important, since what Authors bring is not a completely new endeavour but an innovative method whose value lies in attaining better results.
Conclusions are significant but supported by the experimental findings.
I have no doubts that the approach/method is innovative, but would like to discuss the following point from section 2:
“The second novelty is multiple roles of ARID plays in SC-ARID-QPSO. Since SM/DM usually have many local extrema, maximum SM or minimum DM, which is a non-convex optimization process, may obtain worse results if it is solely based on similarity maximization or discrepancy minimization.”
It is highly arguable that the multiple roles of ARID can be regarded as a novelty. A distinct feature, for sure but, from my point of view, not an added value to be regarded as an innovation. Furthermore, I must stress out that mitigation optimization functions local extrema is an effective strategy for attaining better results, but it does not necessary lead to it. You may be able to demonstrate it for certain cases, but an encompassing demonstration which sustains your broad claim may be difficult to provide.
The document is generally well written. There are some redaction issues, such as “Another method is [35] which developed”, which should be revised, but those are minor problems.
Concerning its organization, I believe that “2. Novelties of SC-ARID-QPSO” would make more sense after current section 4, since the reader needs have read about the methods details to be able to critically assess the novelty claims.
All things considered, I believe that this manuscript is a good scientific contribute and shall be published. My previous comments should be attended to enhance the manuscript, but are only minor issues. Therefore I am recommending acceptance after minor revisions.
Author Response
Responses to Reviewer 2: Comments
This work’s contribution for the scientific community lies in providing a novel image registration method named “Spatial-Consistency and Average Regional Information Divergence Minimization via Quantum-Behaved Particle Swarm Optimization”.
The full name is important as, reading the manuscript’s title, one could remind of other existing approaches. What sets this method apart from others already employed in remote sensing images registration, beyond the employment of Q-PSO, are the shown results’ accuracy and robustness.
Considering this, my first comment is directed to the manuscript title, which should be more accurate, even if necessarily longer.
Response:
Many thanks for reviewer’s suggestion. We have slightly modified our paper title as “Optical Remote Sensing Image Registration Using Spatial-Consistency and Average Regional Information Divergence Minimization via Quantum-Behaved Particle Swarm Optimization”.
The theme is current and interesting and well in line with this journal goals.
Response:
Thanks for your positive comments.
The research problem is fairly well defined. Even if there is not a profuse discussion on the matter, it is quite straightforward to be understood by the reader.
Response:
Thanks for your positive comments.
Also, there is not a deep literature review. However, such task is directed to other article by the same Authors. I like such approach, as there is no need to flood the scientific literature with repetition, but I believe that reading such article would be important for accurately reviewing this one. That has not been the case, since the Early access did not work, at least for me when I tried.
Response:
Thanks for the comments. As a matter of fact, the reference [42] which detailed related works is indeed available on early access now. In addition, we have already sent the pdf file of reference [42] to the editor for information.
By the way, I cheer the Authors for their enthusiasm but I am afraid to say that the affirmation “[42] can be considered as a companion paper to this paper to form a nice pair of works” may not be very adequate in science writing and better left for the readers to judge.
Response:
Thanks for reviewer’s excellent opinions. We have removed this statement.
Nevertheless, the introductory chapter is of good quality.
There are some self-citation, by the amount is not excessive and, more importantly, those look pertinent in the context.
Response:
Thanks for reviewer’s comments.
Research methods are adequate but not deeply explained and frame. Section 5 could be much more developed.
Response:
Thanks for reviewer’s comments. As suggested, we have expanded Section 5 by including experimental design and data generation and selection in this revised revision.
The research beneath the manuscript is substantial and replicable to a certain degree. On the other hand, one shall note that the experimental results are not as abundant as could be. More and, mostly, more heterogeneous datasets would strengthen the results. This would be important, since what Authors bring is not a completely new endeavour but an innovative method whose value lies in attaining better results.
Response:
Thanks for reviewer’s valuable comments. As a matter of fact, the data sets used in our experiments are indeed heterogeneous and they include airborne, satellite data sets. As we explained in Section 5.1, these selected datasets are acquired from different bands, different resolutions, the same/different sensors and the same/different times. These are indeed heterogeneous and very diverse. The images used in experiments have various affine or projection transforms involving a wide range of distortions, such as scale differences, rotation differences, translation differences, and perspective differences. In addition, because the data contains different sensors at different times, data sets also have grayscale differences.
Since this paper develops a novel criterion ARID and a registration method based on the ARID and spatial consistency, the experiments are particularly designed based on consists two parts. On the one hand, the experiments are used to verify the better accuracy of ARID than existing criteria. The evidence in accuracy and robustness of ARID is provided by the consistency feature set selection in Fig. 3 (Fig. 6) and the iteration curve in Fig. 4 (Fig. 7) in Section 6. On the other hand, the experiments are also used to verify the accuracy of SC-ARID-QPSO. Its superiority can be seen through the final registration errors listed in the Table 2, Table 3 and the comparative error curves in Fig. 9.
Response:
Thanks for Reviewer’s positive comments.
Conclusions are significant but supported by the experimental findings.
I have no doubts that the approach/method is innovative, but would like to discuss the following point from section 2:
“The second novelty is multiple roles of ARID plays in SC-ARID-QPSO. Since SM/DM usually have many local extrema, maximum SM or minimum DM, which is a non-convex optimization process, may obtain worse results if it is solely based on similarity maximization or discrepancy minimization.”
It is highly arguable that the multiple roles of ARID can be regarded as a novelty. A distinct feature, for sure but, from my point of view, not an added value to be regarded as an innovation. Furthermore, I must stress out that mitigation optimization functions local extrema is an effective strategy for attaining better results, but it does not necessary lead to it. You may be able to demonstrate it for certain cases, but an encompassing demonstration which sustains your broad claim may be difficult to provide.
Response:
Thanks for reviewer’s comments. The reviewer’s suggestions are well-taken. We have removed the second novelty.
The document is generally well written. There are some redaction issues, such as “Another method is [35] which developed”, which should be revised, but those are minor problems.
Response:
Thanks for reviewer’s comments. We have fixed it.
Concerning its organization, I believe that “2. Novelties of SC-ARID-QPSO” would make more sense after current section 4, since the reader needs have read about the methods details to be able to critically assess the novelty claims.
Response:
Thanks for reviewer’s comments. As suggested, we have moved Section 2 after Section 4.
All things considered, I believe that this manuscript is a good scientific contribute and shall be published. My previous comments should be attended to enhance the manuscript, but are only minor issues. Therefore I am recommending acceptance after minor revisions.
Response:
Thanks for reviewer’s comments.
Reviewer 3 Report
This paper presents a novel remote sensing image registration approach using spatial consistency and average regional information divergence. The proposed idea is novel and good results have been verified. This paper is also well presented. It is suggested to accept this paper for publication.
Author Response
Responses to Reviewer 3: Comments
This paper presents a novel remote sensing image registration approach using spatial consistency and average regional information divergence. The proposed idea is novel and good results have been verified. This paper is also well presented. It is suggested to accept this paper for publication.
Response:
Thanks for reviewer’s comments. We really appreciate it.
Round 2
Reviewer 1 Report
The manuscript deserves to be published in the present form.
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.
