Performance Evaluation of Interest Point Detectors for Heterologous Image Matching
Round 1
Reviewer 1 Report
The authors present a comparison among the five state-of-the-art SAR-Harris, UND-Harris, Har-DoG, Harris-Laplace and DoG interest point detectors. results are compared in terms of scale difference adaptability, nonlinear intensity difference adaptability, distribution uniformity and detection efficiency.
The paper is well written and sufficiently clear to expert readers.
The comparison tends to evidence the higher performance of one of the selected algorithms with respect to the others. This information might be useful for researchers operating in the filed to select the best performing algorithm in their initial stacking of images with different resolution and taken by different sensors to construct a multidimensional dataset to be used for scientific purposes.
I'm not convinced that the chosen metrics might allow a definitive decision about the method, so my suggestion is:
1. use more metrics than the selected (for example coherence images should be shown and scored to assess the good matching of image alignments)
2. use more cases to detect the overall performance of each algorithm
3. show the performances of the co-registration of images with different scales and different sensors, acquired in different times as the ground response might change with the specific meteorological and seasonal conditions
Select a definitive way to define the best algorithm, maybe basing on objective quantities.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper deals with image matching which is one of typical operations in modern remote sensing where data from different sensors have to be combined and processed jointly. Interest point detection is one of the main approaches used. In this sense, the paper can be of interest to readers. A positive feature of the paper is that the authors compare different methods using several criteria and give conclusions that can be useful in practice. However, there are several ways to improve the paper (if accepted):
1) First of all, I dislike Introduction. The authors have to go from general to details. They have to mention possible methods of matching and explain why do they consider methods based on interest point detection. Then, requirements to such methods have to be mentioned. Only after this, surveying the literature is possible.
2) Second, I do not agree with the statement in page 2: "no one has comprehensively studied the performance of interest point detectors". for example, there is https://hal.inria.fr/inria-00548302/document with more than 2000 citations (and the authors mention this paper) and https://www.researchgate.net/publication/269034691_Improving_the_Performance_of_Interest_Point_Detectors_with_Contrast_Stretching_Functions
3) there are also minor errors as Ruby, PDK, Ruby in page 1 and Melo, W.A.L.J. et al in page 2.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The new version of the paper addressed my hints
Reviewer 2 Report
I am satisfied by the changes done according to my comments. It seems to me that the corrections done according to comments of other reviewers improved the paper sufficiently. So, it can be accepted.
