Fusion of Bilateral 2DPCA Information for Image Reconstruction and Recognition

Round 1

Reviewer 1 Report

Note, that modern classification research has mostly focused on neural networks and their subtypes, older methods are considered obsolete due to their lack of flexibility. In my opinion, a new look at PCA and methods related to PCA may look interesting and useful in the means of their adaptation in some new areas where other methods are currently dominant.

Remarks:

1. It is suggested (if possible) to expand the introduction by adding at least one block of text so that the transition from PCA to 2DPCA is clear, at least with its own edges. In my opinion, it a good example of using this method for facial recognition. It would also be a good idea, just for unfamiliar readers, to write down the definition of 1D PCA as written in formulas 1-3 so that the more complex definition can be traced back to the simpler one (for 1D counterpart).

2. The sentence in lines 24-25 states that the images of rows and columns are so perceived differently. This can be supported by a link or a more detailed explanation (whichever is more appropriate in your case because it is easy to prove with example data). Also, the term symmetry is contextual because it can be confused with other means of symmetry (forward and reverse transformation of the matrix itself and some matrices against each other). This is a general view, but it is there. You decide for yourself how to formulate this specific and not very important part of the article.

3. Despite the fact that the proposed method is well explained by formulas, its properties may have a wider theoretical part. Let's explain. People familiar with conventional PCA know that the properties of the original and transformed matrix are related through linearly independent features (in your case, eigenimages or "faces"). Thus, the result of the transformation is clear, and its properties do not contradict each other. But there is a problem here. A row-only or column-only transformation has properties unique to each transformation. These are: orthogonality of scales, unit norm, etc. (and even symmetry, which is discussed in lines 277-283).

4. Since the experimental part is partially focused on how numerically enlarging or shrinking the image affects the response (in terms of classification speed), it would be a good idea to examine the energy of the response (this can give a clue about the difference between the 3 types of PCA transformation).

5. The article is saturated with long abbreviations of methods, which makes reading this work difficult. For example, lines 63-65, 142-147, 158-161, 200, 240, 299-301, etc.

6. Most of the literary sources are quite old (10-30 years). It is appropriate to cite more modern references.

7. There are undeciphered abbreviations, for example, XYNU (lines 320, 321)

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Fusion of bilateral 2DPCA information for image reconstruction and recognition

This paper proposes a new algorithm named direct bilateral 4 2DPCA (DB2DPCA) by fusing bilateral information from images directly. However, it needs major revision. The following comments would help to improve the quality of the paper.

 

 

1.      Results are not discussed in the abstract also how the proposed methodology is better than other works.

2.      What is the domain of this work any practical applications possible??

3.      Proper background is missing in the introduction section.

4.      Author should list all the significant contributions of the work at the end of the introduction section.

5.      Introduction section is very short. Please cite more recent papers and explore the problem statement in more detail with suitable recent references

6.      Please make a separate section for Literature Review and list all the recent works done on this problem and represent them in tabular format with suitable columns.

7.      Equations are not properly cited in the text. Please check it.

8.      A separate section for Proposed Methodology is needed and explains the workflow with a suitable figure. And also explain the proposed methodology in steps.

9.      Make another section, Results, and put all the results in that section.

10.   Figure 4, 5, 7, and 9 needs a table. Give all the parameters and values in the tabular format.

11.   Make another section, “Comparative Analysis” Show some previous work and your results. And make the discussion about the results achieved.

12.   Conclusion section is very, very weak. Talk about the problem handled and the methodology applied, and the results achieved in detail.

13.   Most of the references listed are very old. Add references from 2020, 2021, and 2022 at least 15 references more should be added to the paper.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

It looks authors have responded to my comments but could not highlight the changes. I need highlighted version.

 

Table 1 should be in section 2 (Related works). 

Moreover, Section 2 related work should be fully revised and proper contents should be put. Discuss the past research on this topic. What results did the researchers get? What was the targetted problem??

Comment number 7 should be checked again. I asked for equations.

Comment 9 is not highlighted.

Comment 10 is also not highlighted.

Comment 11: comparison needs a table with references to whom you are comparing with.

Comment 12 is also not highlighted.

Highlight the newly cited references. 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 3

Reviewer 2 Report

The authors did all the comments. Hence can accept the paper as is.