AMSMC-UGAN: Adaptive Multi-Scale Multi-Color Space Underwater Image Enhancement with GAN-Physics Fusion

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

"Figure 2. Flowchart of the physical model-adaptive decision matching deep learning network" - This is not a flowchart; flowchart should follow certain rules.

"Both are divided into 80%, 15%, and 5% =?. The remaining data was subdivided into Test-EP1000 (paired synthetic with real images), Test-UP100 (paired real with real images), and Test-C60 (unpaired real images, challenging scenarios)." need to revise.

The results in Table 3 indicate that our=? algorithm achieves state-of-the-art check other place in paper

Figures 8 and 9.- must be Figures 8 and Figure 9. check other place in paper table number also.

"Our approach consistently yields visually superior and more natural results  on test images, thus strongly indicating the robust generalization capabilities of AMSMC-UGAN in real-world applications." - why it is possible need to justify. check other place in paper

- what are the units in Figure 10?

"4.4 Adaptive control verification experiment

This study validates" - what is the "This study" means? check other place in paper

Abstract and Conclusion should be revised because there are no and finding data shown.

Comments on the Quality of English Language

Please follow the instructions as above written.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

­­The proposed approach has enough novelty in methodology. But, revision in terms of technical detail and paper organization is needed. In this respect, some comments are suggested to improve scientific presentation and technical details.

1. What is the meaning of the phrase “name” in the Table 5 & 6?

2. Did you re-implement all of the compared methods in the Table 3? If no, add related references. If answer is yes, discuss more about hyper-parameters of the compared methods. 

3. The main claim of the proposed method is stand on “underwater condition”. Exactly, which part of the proposed method is designed for this condition? In the other words, is it possible to use your proposed model for natural images (not underwater)?

4. Describe the figure 6 with more details about input and output size of each layer. For example, how do you concatenate the output of CBAM and convolution layers?  

5. Describe the figure 4 with more details. How do you prove the multi scale definition in the figure 4? Difference in size of convolution filters don't make different output feature maps. It is related to input size, batch size, pooling, number of filters, etc.

6. What is the meaning of labels (poor, average, fair, good) in the figure 2? How do you define it?

7. Your proposed approach can be used for under-water and natural image retrieval as pre-processing step to improve performance. For example, I find a paper titled “Innovative local texture descriptor in joint of human-based color features for content-based image retrieval”, which has enough relation and can accept it as pre-process step. Cite this paper and discuss about potential applications and future work ideas.    

Author Response

请参阅附件。

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This article proposes Adaptive Multi-Scale Multi-Color Space Underwater Image Enhancement with GAN-Physics Fusion. While the article addresses an interesting and practical topic, the following points are recommended for improvement: 

1.Enrich it by presenting the best numerical results in the abstract.

2.The contribution of the paper is clearly addressed in the introduction, however, it is necessary to clearly state the motivation and novelty.

3. After stating the challenges in Table 1, the reason for using the proposed method should be justified.

 

4. It is necessary to provide different parts of the proposed method using pseudo code

 

5. In Figure 10-a, remove the border of the graphs.

 

6. Please add software and hardware specifications for experimental results to the article.

 

7. The conclusion needs revision. In the introduction, it should be added for future research direction.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Most of comments have been considered by authors in the revised version. This version is better than original submission in terms of technical details.

Reviewer 3 Report

Comments and Suggestions for Authors

The comments have been addressed and the article is acceptable