M2GF: Multi-Scale and Multi-Directional Gabor Filters for Image Edge Detection
Round 1
Reviewer 1 Report
Dear Authors,
Appreciate your efforts in preparing manuscript titled "M2GF: Mutli-scale and Multi-directional Gabor Filters for Image Edge Detection" and manuscript written and organized very well. Considering the objective and scope of proposed method, i would suggest some modification to current version manuscript to improve further. Please find the file attachment.
Comments for author File: 
Comments.pdf
Minor english editing recquired
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The paper shows a very good research and design of an image edge detection algorithm using multi-directional and multi-dimensional Gabor filters.
The article is well written and structured. The authors use relevant references to support their study.
To improve the quality of the paper I suggest:
1. The overall scientific contribution of the article should be more emphasized in the abstract - it should contain the main findings and the novelty of the results obtained by experiments
2. The use of a large number of references in brackets is not appropriate - p. 2 [20,26,30,31,35,37,42,43], p. 9 [20,26,30,31,35,37] - I suggest breaking down the text so that max. 3 references in parentheses.
3. The descriptions in the pictures are illegible: figures 2, 3, 9 – the font must be enlarged
The contribution is original and interesting, it solves a current issue.
After the above modifications, I propose to publish the article.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript applsci-2507579 by Yunhong Li et al. presents the authors' work on a multi-step protocol for Image Edge Detection (IED) by means of multiple Gabor filtering processes, each with a tuned scale and direction and finally combined to enhance the IED with respect to the results of the more canonical Gabor filtering procedure.
The authors demonstrate to have mastered the topic here under discussion and to have a deep knowledge on the specific functions and methods required to set all the processing stages here presented. Despite this, the presented manuscript lacks of a certain originality and, most important, it does not feature some important characters, which are required for any scientific article intended for public dissemination.
For this reason, I will consider this paper worthy of publication in the Applied Sciences scientific journal only after having ultimated it with the strong modifications I suggest in the following list.
1) First of all, all the mathematical/numerical items included in the main formulation/schematics should be clearly defined before use; for instance, see the "x" and "y" coordinates cited in (4) (between lines 113-114 of the proof), which quite obviously refer to the spatial coordinates into an image frame, but that should be nevertheless specified.
Then again, in eq.(8) and later on, we see the appearance of a "z" variable whose meaning has not been clarified at all.
2) As a second point, the statements explaining the numerical steps of the proposed protocol should be accompanied by clarifying schematics; for instance, the statements in proof lines 188-190 are not accompanied with a depiction of the described frame with indications addressing the single technical items reported (this could help to better understand the meaning of an internal square window "Q").
In addition, a longer and clearer to obtain expression (8) is strongly recommended.
Also the single original pictures show in fig.(1) (each for any column) should be labeled with the names used later in the main text.
3) As a third point, I recommend some simplifications in the mathematical expressions; we see the 3 different "g" functions in (4) and simply related by rules related to the classical complex numbers theory, so they could be reduced; the same could happen in (6): by using a symbolic representation of the same function in (4), the authors could reduce the use of function definitions in the text; this last expressions should also redirect the reader to a specific reference for further readings.
Please, check also the "σ" used at the leftmost side of expr.(6): is it an index to be related with the frequency "f", or it is just another parameter (as stated in line 128)?
Few later, the authors cited "the s-th scale", but this scale is not specified in its precise meaning; maybe, according to the 2nd point of this review, it should be better to add a schematic representation of this peculiar operation.
The caption of fig.(2) could be reduced, since subsets (a) and (b) are related to the same parametric values.
4) As a fourth point, I strongly recommend to perform extensive fixings on the English all around the text; there are too many incomplete/deceiving sentences and typos which make the comprehension of the manuscript harder to achieve.
For instance, there are many ". And" repeated along the text which must be fixed; the statement "The edge algorithm .... gradient-based method" (lines 27-28) is incomplete, as "Robinson.." (lines 36-37) and again in line 42.Please, check the terms "inextenso" (line 106) and "liner" (line 110), "centra" (line 116); in "The total, real and virtual (?)" (line 113) maybe you need to add a term like "component".
There are some articles ("the") with no correct use in the text.
What is the meaning of the phrase in line 247?
These ones are just a small portion of the text corrections to complete, and I suggest a full inspection from the very beginning of the draft to ultimate it.
5) Finally, I add some minor points referring to the imprecisions/uncertainties in the data analysis; we see from fig.(3n) that a Gabor filter loses reliability when increasing the number of angular sectors with different luminance values (for instance, it seems that the contrast between domains T1 and T5 does not induce a polar lobe with an orientation consistent with the corresponding angle of T1-T5 edge; the authors should add some lines to analyze this point.
Also stating that there is a certain "superiority of the multi-scale filter" should be addressed with cautions: we kwon that a procedure refined with the addiction of further processing steps can be more accurate than its basic version, but still it requires more data processing, and consequently increase the need of resources in terms of hardware power and process time.
As explained in the main review text, the English requires some important fixings.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
