A Statistical Analysis for Intensity Wavelength-Resolution SAR Difference Images
, Dimas Irion Alves 2,*,†, CrÃstian Müller 1,†, Renato Machado 2,†, Lucas Pedroso Ramos 2,†, Viet Thuy Vu 3 and Mats I. Pettersson 3Round 1
Reviewer 1 Report
This study presented a complete statistical hypothesis test for wavelength-resolution incoherent SAR change detection. Based on statistical analysis for intensity wavelength-resolution SAR difference images, a change detection method was proposed in this paper. Compared with other methods, the work reported higher detection accuracy. The whole paper has been well structured. However, I still have some critical comments as follows.
My first concern is about the contributions and novelty of this work. This paper presented a statistical analysis for intensity wavelength-resolution SAR difference images. By goodness-of-fit tests, the authors’ claimed the conclusion that “the Gamma distribution is a good fit for the background of the tested SAR images”. However, this work has been reported by authors’ previous works (e.g. ref [29] Vu, V.T.; Gomes, N.R.; Pettersson, M.I.; Dammert, P.; Hellsten, H. Bivariate Gamma Distribution for Wavelength-Resolution SAR 407 Change Detection. IEEE Transactions on Geoscience and Remote Sensing 2019, 57, 473–481.) The new contribution of this work should be addressed.
The second concern is about the experimental design. The results cannot fully support the conclusions. Firstly, though 24 images in the CARABAS-II data set were employed for validation, no visual inspection was employed. The local patch of typical images, the difference between predicted and reference may increase the impacts of research. Secondly, more state-of-the-art methods should be considered for comparison, and both quantitative (by table or figure) and visual assessments should be presented.
The third concern is about the discussion part. A more comprehensive discussion should be presented. As a reader, I want to know the pros and cons of the presented method. For instance, except for CARABAS-II, is the Gamma distribution suitable for modeling other SAR images? How about the time complexity of the proposed method? Moreover, the VHF band has strong penetration. Suppose the acquired time interval of two SAR data is relatively long. In that case, the surface electric dielectric constant of typical objects (such as soil moisture, etc.) may change, resulting in a large change in the intensity value of SAR data and the introduction of noise. How about the result in this situation?
Here are some other minor comments：
1. Figures, Tables, and related captions should provide enough information that a reader can understand the data presented without referring to the main text. Therefore, the meaning of “M2P1”, etc., in Table 1 should be added. In Figure 5 and Figure 6, the symbol “-” in “… method - …..” may be confused with the minus symbol. Figure 5 and Figure 6 can be merged and the captions in Figure 5 and Figure 6 addressed nothing.
2. Experiment 18 in Table 1 should be discussed, why is FAR as high as 15.33？
3. The introduction part should be refined by including more related studies.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This manuscript finds that the Gamma distribution is a better distribution than the exponential distribution in modeling the background of intensity wavelength-resolution SAR difference images. However, this should be immediately obvious since the exponential distribution is known to be a special case of the gamma distribution.
It is also the case that the proposed method is tested on a small dataset (only 24 images). Testing on a larger dataset is desirable.
On line 210, there is a slight error in notation - k_(x_i) should be k_i and theta_(x_i) should be theta_i.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper presents a statistical analysis for intensity wavelength-resolution SAR difference images. And taking advantage of the results of this statistical analysis, a change detection application for the detection of concealed targets is presented. Finally, the effect of the application is verified by the CARABAS-II data.
However, there are still some problems with the description of the article :
1. There are two abbreviations "SAR" and "FAR" in the abstract, but their specific meanings are not explained. Secondly, keywords that are not in the abstract appear in “Keywords”.
2. From the results in Table 1, the current threshold has a better detection effect. Is there any basis for choosing this threshold?
3. At the end of the paper, it is said that the gama distribution is more in line with the statistical characteristics of the current SAR images. Can the image be numerically analyzed to verify this conclusion instead of just starting from the detection point of view.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Thanks for the detailed rebuttal letter. All my concerns have been properly addressed. I have no more comments.