Identifying Heterogeneity in SAR Data with New Test Statistics
Pablo Blanco
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsPage 2, line 81. When defining L, do you refer to the number of looks or to the equivalent number of looks?
Page 15, line 266. Could you please better define the intensity format? Does it refer to GRD? In that case, which acquisition mode and which final equivalent number of looks? How does this information relates with the L values in Figure 13 caption? It would also be useful to complement the description with the images acquisition date.
Page 15, Figure 13. In order to ease the interpretation of the results for the 3 sites, it would be useful to show an optical image (the closest's Sentinel-2 available for example) of the corresponding SAR images.
Page 15, Figure 13 caption. In reference to the L values 18, 36 and 36. Could you please provide more information about how you get those values?
Page 16, line 269. Could you comment on why you selected a 7x7 window?
As a suggestion, in order to conduct a more quantitative analysis of the results for each estimator, would you consider comparing the mean and standard deviation values of the different tests among some reference areas in the images? Ideally, a land cover map would serve. Alternatively, you may use some polygons with a clear land cover label (water, vegetation, urban). This would also serve to evaluate the sensitivity of the estimators for classification purposes.
Some general comments:
- in the case you chose an GRD image. Wouldn't it make more sense to apply the operators on an Single Look Complex image so in native SAR coordinates?
- could you comment what to expect if doing the same tests employing higher resolution images such as TerraSAR-X's?
- could you comment what would be the impact of your work in classification applications with SAR images?
Author Response
Please see the file attached
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper estimates homogeneous areas, and conversely heterogeneous areas, in SAR imagery via speckle statistics by exploiting variations in scattering at the pixel level. Teasing out information from this statistical behavior remains an intricate, delicate process; one which you appear to have accomplished.
I have only a few comments and suggestions for clarification that, hopefully, will enhance the readability & ultimate applicability of the ideas presented.
1. The 3rd to the last paragraph in the Introduction seems redundant. First entropy is discussed and then the 2 CV estimates, so why return again to entropy? (Perhaps some of this paragraph could be incorporated into the previous entropy discussion.)
2. In Eq. (8), the limit of alpha -> -inf is not transparent. The first 5 terms match eq. (7), and the ln(gamma) & digamma terms seem to cancel, however, that leaves the ln(-1-alpha) & L factors. A short discussion, or footnote, would help the reader.
3. Fig. 4 Q-Q plots: Are the dashed green lines fits or theoretical estimates of the distributions? What do the different scales imply? (Perhaps a larger variance in one of the distributions?) A short note in the caption or clarification in the text would help. (The later Q-Q plots all show an x=y line for reference. Why the change here?)
4. What does 'nominal levels of 1%, 5%, and 10%' mean? And, how are these 'achieved' in all cases? This paragraph is the only place where 'nominal levels' are mentioned, please clarify their meaning & significance.
The mathematics is, by necessity, a bit dense, therefore I've concentrated my comments on the flow of the paper so that the casual reader may readily follow. To fully follow, comprehend & replicate this work will require effort on the reader's part, nonetheless clarification of the above points will assist in understanding the gist of the paper.
Author Response
Please see the file attached
Author Response File: Author Response.pdf
