Analysis of Random Local Descriptors in Face Recognition

Round 1

Reviewer 1 Report

Dear Authors

The paper aims at development of a model to analyse the random Local descriptors in face recognition through image processing. The novelty of the work is weak; the image processing tasks are not clearly described. The paper length is excessive! There are plenty of similar works available in the literature dealing with this topic. However, due to the efforts that authors made to produce the results, I believe that the paper might have the possibilities for publication if the authors addressed the reviewer’s comments.

1. What is the novelty of the work? It was expected to find it within the abstract or introduction section! Please clearly stress the innovative point of the work and the main contribution to the state of the art.
2. Please simplify the introduction section, the authors stated too many information which is not essential for this paper. I recommend to put the attention on the main features of the work and explaining the main contribution of this study to the literature.
3. The paper lacks a table of nomenclature including all variables and acronyms.
4. Please refer the mathematical development to an already published article or book unless if you develop by yourself please state it clearly.
5. Section 2 “PCNC”? first of all, a main section should not have presented with an acronym. Second, it must be better organised. The content is vague.
6. LBP algorithms were used in this article, good, but how did you implement it? A flow chart is needed to describe all the necessary stages. The interface with the software must be described in detail.
7. How did you calculate the error? An equation must be indicated.
8. In the experimental procedure, I cannot find any detailed information regarding the camera specifications? And most importantly the camera calibration procedure.
9. How is the system accuracy? It is crucial to know that.
10. It is recommended to use passive tense instead of active tense starting with “we”. This is an academic manuscript!
11. Figures are not presented in good quality! I recommend the authors to reproduce all of them.
12. Too many figures (23 figures) presented in the paper. Please merge or simplify them.
13. Decimals! Please unify all decimals when presenting the number. This format used in the paper is not acceptable.
14. Conclusions are weak! It must be improved and discuss more detail, applications must be included.

Very Best

The Reviewer

Author Response

Please, see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents a model of face features detection by using local descriptors. The idea is interesting and show potential models for handcrafted image processing. Major revisions may help to improve this paper:

1. Explain novelty in contrast to well known extractors and key point search models like Zernike moments, surf, sift, etc. This part is missing in your manuscript.
2. Consider related research works: An image super-resolution reconstruction method with single frame character based on wavelet neural network in internet of things, An adaptive local descriptor embedding Zernike moments for image matching, Object detection and recognition via clustered features.
3. In fig. 1 we can see 3x3 dimensional filtering. Have you tested also other filters?
4. In fig. 2 we can see neural network but there is no presentation of your architecture. How are fig. 7 and fig. 8 related to your neural network? Make revisions to your presentation.
5. How was your neural network trained? Which algorithm was taken for your training?
6. How was your eq. (1) defined? Values of coefficients need explanations.
7. Directions of your filter motion presented in fig. 6 need some explanation.
8. (7) is not clear. How do you define ONF and ON features sets in this equation?
9. 21 is your app or some interface form existing models?
10. Your model needs results for comparisons. Compare it to other solutions from neural network domain.

Author Response

Please, see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear Authors

Good Job!

Very Best 

The Reviewer

Author Response

Dear Reviewer,

Thank you for your valuable work!

Best regards!

Reviewer 2 Report

Thank You for revisions, now paper is better but here are some concerns to be solved:

1. Related work and introduction section are presenting actually 27 paper which are over 10-20 years old papers, please revise it and reduce the number of old research to minimum at max. 5 otherwise your paper is just a reproduction of old research.
2. Why eq. (1) was excluded? Why instead of explaining what was done you exclude such equation? It makes now serious concern about your research since if parts of model can be excluded how about the resto of the model?
3. Your neural network is not defined. In your mentioned fig. 2, gi. 7 and fig. 8 there is nothing about your neural network. Are you sure there was neural network used in your research? It makes serious concern about it.
4. Training of your neural network is vogue. In response You wrote "The training process of neural network is realized between penultimate and ultimate layers by changing of weights of the connections. It was the Hebbian algorithms. We included this information to the article text." But there is no information how this was done. You present a standard sentence which does not answer any question. I am now even more confused about your neural network and its actual application.
5.  

Author Response

Please, see attached file.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Still many of related work and introduction section are presenting paper which are over 10-20 years old papers, please revise it and reduce the number of old research to minimum at max. 5 otherwise your paper is just a reproduction of old research.

Author Response

Dear Reviewer,

thank you very much for helping us improve our article.

 

 

Reviewer question

Still many of related work and introduction section are presenting paper which are over 10-20 years old papers, please revise it and reduce the number of old research to minimum at max. 5 otherwise your paper is just a reproduction of old research.

Our response

We have done hard work to reduce the number of “old” references. Now we have only 37 references instead of 52. All corrections are presented in new version of our article.

Thank you!