Combination of Fast Finite Shear Wave Transform and Optimized Deep Convolutional Neural Network: A Better Method for Noise Reduction of Wetland Test Images
Round 1
Reviewer 1 Report
Dear Authors!
The direction of research you have chosen is relevant and very useful, has a theoretical and practical background, and also represents practical interest for specialists in the field of developing artificial intelligence engineering applications, as well as for researchers in the field of environmental management and engineering ecology in particular.
But despite the general pleasant impression, there are a number of nuances that I would like to draw your attention to:
1. The title of the work "Combination Of Fast Finite Shear Wave Transform And Optimized Deep Convolutional Neural Network: A Better Method For Image Noise Reduction" is of a general (broad) nature. It would be nice to change the title of the work to reveal the object of study - wetlands. By the way, this is quite acceptable for image processing on peat deposits. Please think...
2. The paper states that "the network model has powerful feature extraction capabilities". It would be nice to describe the entire spectrum of features, describe the procedure for selecting the most informative (in terms of the problem being solved), and separately show how these features correlate with each other in order to improve the quality of the final result.
For an example, look in the literature: https://doi.org/10.1051/e3sconf/202017402009 (see tables 1 and 2 on pages 7 and 8)
3. Please pay attention to the need to detail the structure and justify the correctness of the neural network parameterization.
As an example, no more, I propose to familiarize yourself with the work:
For an example, look in the literature:
https://doi.org/10.1051/e3sconf/202017402020 (I am sure you will benefit from these 10 tables)
4. To what extent can your approach be considered acceptable for processing images from different hographic swamps, taking into account the variation and severity of climatic conditions?
As an explanation, I will give an example of different peat deposits: Latvia and the Northwestern lands of Russia (Irrigated and drained deposits). The image processing approach, in your opinion, should take into account both "characteristics of water mineralization in peat bogs" and "properties of peat deposits under technogenic impact"? Or the result will not depend on it? Consider this moment to avoid unnecessary questions.
5. In the formulas (for example 7) there are notations that require deciphering. I understand that a trifle, but it is necessary to reveal everything.
6.Section "Results" should contain methodological results.
7. The Discussion section should evaluate the approaches described in the Introduction section.
Good luck with revisions!
Before promoting a manuscript, I recommend proofreading it by a native speaker.
There are lexical errors throughout the text, and also pay attention to the conjugation of sentences.
Author Response
Dear reviewer:
Thank you for your decision and constructive comments on my manuscript. We have carefully considered the suggestion of Reviewer and make some changes. We have tried our best to improve and made some changes in the manuscript.
The red part that has been revised according to your comments. Please refer to the attachment for instructions on peer-to-peer revisions.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
While the proposed study seems promising and may present a novel approach to addressing the noise problem in wetland experimental images, some concerns and potential limitations should be considered:
1. The study doesn’t provide details on how the method was validated against other existing methods. Understanding the specific benchmarks and metrics used is vital in evaluating the effectiveness of the proposed method.
2. It's unclear how well the method might generalize to other types of noisy images. Can this approach be used in other environmental contexts, or is it specific to wetland images?
3. What type of training data was used? Is there enough diversity in the dataset to ensure that the model is robust to different types of noise, lighting conditions, or weather patterns?
4. The FFST and deep convolutional neural networks could be computationally expensive. Is the proposed method efficient enough to be used in real-time or near-real-time applications? What are the hardware requirements?
5. The use of a deep network model can lead to overfitting if not managed properly. Was any method applied to prevent overfitting, such as dropout or regularization?
6. Deep learning models are often considered "black boxes," meaning that it can be challenging to understand how they are arriving at a particular decision. How interpretable is this model? Understanding why a model is making a particular decision can be important in scientific contexts.
7. How dependent is the method on the specific use of FFST? Would other transformation methods be equally effective, or is there something specific about FFST that is crucial for this approach?
8. How sensitive is the method to hyperparameter tuning? Deep learning models can sometimes be very sensitive to the initial settings of the model parameters, and this could affect the repeatability of the method.
9. Deep learning is well-known and has been used in previous studies i.e., PMID: 36642410, PMID: 37519050. Therefore, the authors are suggested to refer to more works in this description to attract a broader readership.
10. While the study emphasizes the significance of the method, it might be worthwhile to further explore the broader implications and how it may actually impact field studies and environmental monitoring.
11. The paper claims superiority over traditional denoising methods but does not provide specifics. A detailed comparison, including quantitative evaluations, would strengthen the claim. Also, statistical tests should be conducted when comparing to see the significant differences.
12. Denoising can sometimes introduce artifacts that may distort the original features of the images. Was this aspect considered in the study?
13. Uncertainties of models should be reported.
14. More discussions should be added.
English writing should be minor checked.
Author Response
Dear reviewer:
Thank you for your decision and constructive comments on my manuscript. We have carefully considered the suggestion of Reviewer and make some changes. We have tried our best to improve and made some changes in the manuscript.
The red part that has been revised according to your comments. Please refer to the attachment for instructions on peer-to-peer revisions.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear Authors!
Thank you for providing answers to my questions.
Basically, everything is taken into account. I personally have no more questions about the current version of the article.
And I'm ready to support you!
I wish you scientific success.
Reviewer 2 Report
My previous comments have been addressed.
