A SAR Image Despeckling Method Based on an Extended Adaptive Wiener Filter and Extended Guided Filter
Round 1
Reviewer 1 Report
This paper proposes a SAR image despeckling method, consisting of an extended adaptive wiener filter, an extended guided filter and a wls filter. From the result, the denoising effect of the method is very good. However, I think this manuscript cannot be accepted before major revision. The reasons are as follows:
The method is too mixed. There must be one part of the method that is the most important and some improvements are unnecessary or make little impact on the final filtering results. The ablation experiments should be added in the manuscript. Why is the computational complexity of the proposed method lower than that of the Lee filter? Based on the computation of the proposed method, the computational complexity of the extended adaptive wiener filter is equal to that of the Lee filter. Not to mention that the proposed method has three filters in series. Equation (17). Will the size of the local window affect the result of PEAW? The EGF is applied twice. Please give a detailed explanation. The noise level in Table 1. What the definition of the noise level is? Please give the reference or the mathematical definition. All abbreviations and acronyms should be defined where they are first used, once in the Abstract and once in the main text. There are some format problems, such as “Because of The multiresolution filters” on page 3. And the manuscript should be polished by the native speaker.Author Response
Response to Reviewer 1 Comments
Dear reviewers, Thank you for your careful review and your comment
Point 1: The method is too mixed. There must be one part of the method that is the most important and some improvements are unnecessary or make little impact on the final filtering results. 

Response 1: In order to increase the clarity of the paper, the abstract was rewritten to clarify the necessity of applying each filter. In addition, in order to avoid too much attention to detail and reduce confusion, a number of images and tables as well as related material have been deleted (Table1, some part of figure 5, Figure 12 and Figure 14).
Point 2: The ablation experiments should be added in the manuscript.
Response 2: Several SAR images and experiments have been added to update paper to examine the proposed method more accurately. SAR image 2 is added to the update paper (Figure 14). In addition, the ENL and STD of this image are compared with the other existing filters. Next, the SAR image 3 is added to the updated article (Figure 15). Noise with different variants was added to the image (0.04, 0.06, 0.08, and 0.1) and denoised image evaluated by the PSNR, SNR, SSIM, and MAE. In addition, the runtime of all filters (EAWF, EGF and WLS Filter) in the proposed method have been added to the updated paper (Table 10, 12, 14, and Table 16).
Point 3: Why is the computational complexity of the proposed method lower than that of the Lee filter? Based on the computation of the proposed method, the computational complexity of the extended adaptive wiener filter is equal to that of the Lee filter.
Response 3: There were a number of mistakes about run time comparation in the paper - Section 6 that are corrected in updated paper. But, the propose method is faster than Lee filter. According to table 9 in updated filter. The average run time of the proposed method with fourteen images is 1.92. But the average run time of Lee filter with fourteen images is 6.10. Therefore, proposed method is faster than Lee filter. In addition, according to table 10 (The runtime of all filters in the proposed method) the average run time of the EAWF with fourteen images is 0.03. Therefore, EAWF is faster than Lee filter.
Point 4: Equation (17). Will the size of the local window affect the result of PEAW?
Response 4: Yes, it affects. So we've mentioned window size on page 7 (Added to updated paper). We also explain at the top of page 8 the effect of window size on the proposed filter. (Added to updated paper)
Point 5: The EGF is applied twice. Please give a detailed explanation.
Response 5: in the second paragraph-page 9, I have mentioned (highlighted) that why The EGF is applied twice (Added to updated paper).
Point 6: The noise level in Table 1. What the definition of the noise level is? Please give the reference or the mathematical definition. .
Response 6: The level of noise is the intensity of the variance of the noise that we add to image. Table 1 is omitted in the updated paper, but in Figure 3 and Table 8 I have replaced the noise level with the noise variance. At the end of the second paragraph of page 19 is explained.
Point 7: All abbreviations and acronyms should be defined where they are first used, once in the Abstract and once in the main text.
Response 7: Corrected according to format.
Point 8: There are some format problems, such as “Because of The multiresolution filters” on page 3. And the manuscript should be polished by the native speaker.
Response 8: I got the opportunity from the editor to edit the paper after the reviewers review my paper.
Kind regards,
Hadi Salehi
Author Response File: 
Author Response.docx
Reviewer 2 Report
In this paper the authors have proposed a SAR filter based on a proposed Extended Adaptive Wiener Filter, Extended Guided Filter and wls Filter. The presented results show that the proposed algorithm has a better performance in comparison with the other existing filters. In my opinion the paper addresses an interesting topic with some innovation. However, it is not ready for publication.
Minor comments:
--- Please see the attached pdf file for minor comments.
Major comments:
--- As the authors mentioned in the manuscript as well, the topic of SAR image speckle filter is not new and it is a very well treated field. Accordingly, I suggest the authors to conduct more in-depth literature review, and explicitly describe the differences between the published methods in the literature and that being presented in the paper. For instance, one of the most famous filters (Weibull Multiplicative Model) which has a similar characteristic to the presented filter has been omitted in this paper.
--- Overall, the manuscript would also benefit from language editing and spell-checking. Some parts of the text are quite difficult to follow and some statements are not entirely clear due to issues in phrasing.
--- The abstract of the manuscript should be rewritten. It does not express the contact of the manuscript properly.
--- Most of the examples in this manuscript are non-RS imagery. I suggest the authors to reduce not relevant images and add more SAR data in order to make it more understandable.
--- Section 5 (Experiments on Real SAR Images) needs to be improved substantially. I suggest the authors to add more information about the SAR data they have used (e.g. sensor, resolution, pre-processing applied to the data etc…). I also recommend the authors to add more examples to this section.
--- References does not follow MDPI RS style.
Comments for author File: 
Comments.pdf
Author Response
Response to Reviewer 2 Comments
Dear reviewers, Thank you for your careful review and your comment
Major comments:
Point 1: As the authors mentioned in the manuscript as well, the topic of SAR image speckle filter is not new and it is a very well treated field. Accordingly, I suggest the authors to conduct more in-depth literature review, and explicitly describe the differences between the published methods in the literature and that being presented in the paper. For instance, one of the most famous filters (Weibull Multiplicative Model) which has a similar characteristic to the presented filter has been omitted in this paper. 

Response 1: I have tried to provide more detailed content in literature. On pages 1 and 2 the highlighted areas are added to the introduction as well as, the Weibull Multiplicative Model is described.
Point 2: Overall, the manuscript would also benefit from language editing and spell-checking. Some parts of the text are quite difficult to follow and some statements are not entirely clear due to issues in phrasing.
Response 2: I got the opportunity from the editor to edit the paper after the reviewers review my paper.
Point 3: The abstract of the manuscript should be rewritten. It does not express the contact of the manuscript properly.
Response 3: the abstract was rewritten to clarify the necessity of applying each filter.
Point 4: Most of the examples in this manuscript are non-RS imagery. I suggest the authors to reduce not relevant images and add more SAR data in order to make it more understandable.?
Response 4: Several SAR images and experiments have been added to update paper to examine the proposed method more accurately. SAR image 2 is added to the update paper (Figure 14). Next, the SAR image 3 is added to the updated article (Figure 15).
Point 5: Section 5 (Experiments on Real SAR Images) needs to be improved substantially. I suggest the authors to add more information about the SAR data they have used (e.g. sensor, resolution, pre-processing applied to the data etc…). I also recommend the authors to add more examples to this section.
Response 5: The ENL and STD of SAR image 2 (the new image that added to update paper) are compared with the other existing filters. In addition, the pixel values of three filters (Figure 14 k, n, m) are considered. Next, the SAR image 3 is added to the updated article (Figure 15). Noise with different variants was added to the image (0.04, 0.06, 0.08, and 0.1) and denoised image evaluated by the PSNR, SNR, SSIM, and MAE. In addition, the runtime of all filters (EAWF, EGF and WLS Filter) in the proposed method have been added to the updated paper (Table 10, 12, 14, and Table 16).
Point 6: -- References does not follow MDPI RS style.
Response 6: References changed to MDPI RS style.
Minor comments:
Response 1: Abstract rewritten
Response 2: All abbreviations corrected according to format
Response 3: All necessary references were provided.
Response 4: missing image corrected.
Response 5: More information and examples were added
Response 6: The scale of the images became larger
Response 7: references changed to MDPI RS style
Kind regards,
Hadi Salehi
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
After the revision, the content of the article is obviously improved. However, some issues are still left open to be discussed:
The motivation of the EAWF should be described more clearly. At least, the denoising result of the EAWF+EGF+WLS, EGF+WLS and EAWF+WLS as well as the corresponding analysis should be given as the ablation experiment to verify the correctness and necessary of the designed components in the proposed method. After applying the EGF twice, the result seems smoother. Some unwanted spots are removed, but the details in the lower right corner of the Fig.7 are also missing. Therefore, the tradeoff between noise suppression and detail preservation seems not to be kept well. In addition, the parameters in EGF are set manually, and the robustness of the EGF is not verified well. The ability of point target maintenance requires additional discussion. Figures 7 to 10 should be placed together where Table 11 and 13 (or 12 and 14) should be placed together. In table 6, the Matlab code is not necessary and can be removed. The look number of the simulated and the real SAR images should be given. Dose the look number affect the denoising result? Please give a detailed explanation.Author Response
Response to Reviewer 1 Comments
Thanks for your Comments that has improved the quality of the paper.
Point 1: The motivation of the EAWF should be described more clearly. At least, the denoising result of the EAWF+EGF+WLS, EGF+WLS and EAWF+WLS as well as the corresponding analysis should be given as the ablation experiment to verify the correctness and necessary of the designed components in the proposed method. 

Response 1: the denoising result of the EAWF+EGF+WLS, EGF+WLS and EAWF+WLS as well as the corresponding analysis are applied. 6 images are studied (Figure 7) and highlighted texts on pages 1, 2 and 3 as well as Table 2 describe the motivation of the EAWF and EGF.
Point 2: After applying the EGF twice, the result seems smoother. Some unwanted spots are removed, but the details in the lower right corner of the Fig.7 are also missing. Therefore, the tradeoff between noise suppression and detail preservation seems not to be kept well. In addition, the parameters in EGF are set manually, and the robustness of the EGF is not verified well. The ability of point target maintenance requires additional discussion.
Response 2: This is true, areas that are not fully recognized by the DOV or afterwards the CV' as the edge or detail areas are slightly softened. This softening also reduces the PSNR. This decrease is less than 2% (According to table 2 that add to paper). In addition, it is not true for all images. So we've added two more images in Figure 8. On the other hand, we need to increase the ENL criteria and reduce the STD criteria in homogeneous areas. Therefore, the use of the second EGF and WLS are acceptable. (Full descriptions are on page 12, 13, Figure 8 and Table 2)
Point 3: Figures 7 to 10 should be placed together where Table 11 and 13 (or 12 and 14) should be placed together.
Response 3: It is done.
Point 4: In table 6, the Matlab code is not necessary and can be removed.
Response 4: the Matlab code is removed.
Point 5: The look number of the simulated and the real SAR images should be given. Dose the look number affect the denoising result? Please give a detailed explanation.
Response 5: In this study, the actual SAR images were obtained from the following image bank.
- Dataset of Standard 512X512 Grayscale Test Images. Available online: http://decsai.ugr.es/cvg/CG/base.htm (accessed on 30 December 2018).
- Dataset of Standard Test Images. Available online: https://www.sandia.gov/ RADAR/ imagery/ #xBand2.
Only the image is contained in these sources and doesn't exist additional information. Of course, no additional information is needed to evaluate the noise removal rate of the proposed method. currently other researchers can compare their own method with the proposed method.
also, in the simulated images, the speckle noise is added with the desired noise variance to the original image. the proposed method is applied to the noisy image. Finally, the denoised image is compared with the original image.
Author Response File: Author Response.docx
Reviewer 2 Report
Minor comments:
--- When you use an abbreviation define it first. E.g. Weighted Least Squares (WLS) not weighted least squares (WLS). The first letters should be in capital.
--- Please check the references again, still they are not in MDPI standard.
Author Response
Response to Reviewer 2 Comments
Thanks for your Comments that has improved the quality of the paper.
Point 1: When you use an abbreviation define it first. E.g. Weighted Least Squares (WLS) not weighted least squares (WLS). The first letters should be in capital.
--- Please check the references again, still they are not in MDPI standard.


Response 1: The paper was corrected and the corrected areas were highlighted.
Author Response File: Author Response.docx
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.
