Polarimetric Synthetic Aperture Radar Speckle Filter Based on Joint Similarity Measurement Criterion

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Speckle filtering is an essential step in SAR data processing. Currently, commonly used filtering algorithms are mainly based on spatial domain methods. However, these algorithms have shown limitations as SAR systems continue to evolve. To address these limitations, this study proposes a polarimetric speckle filter. This method utilizes image segmentation principles to determine the filtering window and introduces a joint similarity measurement parameter for identifying homogeneous pixels.

This paper is interesting and well written.

Some suggestions:

1）The summary in the introduction section should be more concise. A brief description of early-stage algorithms would suffice.

2）Figure 5b can be replaced with a pseudocolor image of the data.

Comments on the Quality of English Language

This paper is well written.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

1. In the abstract part, the author says that ‘With the rapid development of PolSAR systems, traditional techniques can no longer satisfy the urgent demands for speckle suppression in maintaining structure and polarization characteristics.’ This is a little strict. Do the authors mean that there is no good noise filtering for PolSAR data currently?

2. In the introduction part, some references are from more than ten years ago. Reference 28, appearing in the experiment results part, is from a professor from UPC, Carlos Lopez Martinez. Please also introduce his work in introduction part. Here is his personal page, and you can find his early work about PolSAR noise filtering. https://futur.upc.edu/CarlosLopezMartinez/p/2

3. The proposed method applies an irregular window. I guess the intention is that a better result can be obtained with more sophisticated windows. But the problem is, you can’t know the accurate homogeneity before the filtering due to the noise. One of the consequences is that the filtering result could be unsatisfactory on the edge of the window. Besides, the moving window is not working anymore.

4. Another disadvantage of the irregular window is, for the application with more than one image, such as change detection and interferometry, this irregular filtering could bring more difference because it is not sure that every image will have the same filtering window. This is a system error introduced by the filtering.

5. I don’t understand why the waterbody in Fig.11(b) has a stronger span than the ground. It is clear in Fig 5 that the water body is black. And please use some legends, in Fig 5 and 11, to indicate the coordinates, direction, and distance scale.

6. When comparing with other filtering, the window size may be a trick. From Fig.3 it can be seen that most size of the irregular windows are under 40, while the comparison group has a window of 7x7, which means the window size is 49. So for sure, the proposed method would have a result with more details.

In conclusion, I think it is necessary for the authors to discuss the question I proposed and put them in the discussion part. And some little improvements need to be done to the experiments. Besides, the language is good, clear, and understandable, but can be improved, such as in line 40: the filter filters. a bit repetitive.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This article presents a PolSAR image despeckling method using the joint similarity measure criterion with the adaptive filtering window. The experiment results on the airborne and spaceborne real PolSAR data show good performance. However, there are some issues that need to be addressed:

1. The adaptive filtering window selection method based on a multi-directional ratio edge detector is interesting. From figure 2(c), it seems that the adaptive windows are similar to the results of super-pixels segmentation by SLIC (simple linear iterative clustering). The authors should compare the window construction differences between these two methods.

2. On page 5 Line 190, the authors claim that the average window size is about 47 pixels. But it is super strange that most of the windows have less than 40 pixels from figure 3. Also, I am curious if the PolSAR images in different areas (homogenous place and un) have the same average window size or not.

3. The authors used the Refined Lee filter (proposed in 1980), SMB filter (proposed in 2006), IDAN filter (proposed in 2006), and NLM filter (proposed in 2014) to compare with the proposed filters. The present state-of-the-art (SOTA) filters are not well-researched. The authors should at least compare the proposed filter with at least three SOTA filters.

Comments on the Quality of English Language

Extensive editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

I have no further comments