Improving the False Alarm Capability of the Extended Maximum Average Correlation Height Filter

Round 1

Reviewer 1 Report

The article has theoretical analysis and experimental verification, It  has certain innovation. It is recommended to supplement robustness experiments

The article has theoretical analysis and experimental verification, It  has certain innovation. It is recommended to supplement robustness experiments

Author Response

Comments and Suggestions for Authors:

1. The article has theoretical analysis and experimental verification, It has certain innovations. It is recommended to supplement robustness experiments.

Thanks very much for reading the manuscript and encouraging remarks.

Comments on the Quality of English Language

1. The article has theoretical analysis and experimental verification, It has certain innovations. It is recommended to supplement robustness experiments.

Thanks very much for reading the manuscript and encouraging remarks.

Reviewer 2 Report

- The paper tests the performance of a wavelet-modified EMACH filter and shows the effectiveness compared to MACH and EMACH. The main contribution of the paper is to apply a wavelet-modified EMACH filter and perform an experiment. The paper shows the effectiveness of EMACH in the numerical experiments, but the experimental results can be improved with additional experiments.

- The paper mentioned that "the decomposition of the signal into dilated and translated wavelets is done by 78 wavelet transform due to its amazing multiresolution, denoising, and feature extraction 79 capabilities." However, it is not clear to me how this decomposition is connected to Eq. (7). The explanation of (7) should be elaborated further.

- The authors should explicitly summarize the contribution in the paper.

line no. 275: is enhanced by the incorporating. 

- the should be removed

Author Response

Authors’ Response to Reviewer’s Comments

Reviewer # 2

Comments and Suggestions for Authors:

1. The paper tests the performance of a wavelet-modified EMACH filter and shows its effectiveness compared to MACH and EMACH. The main contribution of the paper is to apply a wavelet-modified EMACH filter and perform an experiment. The paper shows the effectiveness of EMACH in the numerical experiments, but the experimental results can be improved with additional experiments.

Thanks very much for reading the manuscript and encouraging remarks. We have included more results and studied performance measure parameters in the revised manuscript.  

2. The paper mentioned that "the decomposition of the signal into dilated and translated wavelets is done by 78 wavelet transform due to its amazing multiresolution, denoising, and feature extraction 79 capabilities." However, it is not clear to me how this decomposition is connected to Eq. (7). The explanation of (7) should be elaborated further.

Thanks very much for reading the manuscript carefully and providing encouraging remarks. We would like to thank the reviewer for pointing out the issue related to the decomposition of the signal into its connected dilation and translation properties. The statement is generic in nature based on the features of wavelet. In the revised manuscript, we have added the following - "Mathematically, ........... set of wavelets" in sub-section 2.3.

In this study, we have taken the Mexican hat function for the decomposition of the signal into daughter wavelets. We have explicitly added the dilation and translation properties of a generic function h.

3. The authors should explicitly summarize the contribution in the paper

Thanks very much for the kind comment. In the revised manuscript, we have summarized the contribution.

Comments on the Quality of English Language

1. Line no. 275 is enhanced by the incorporating.

~ the should be removed.

Thank you for pointing out the mistakes. The corrections have been incorporated in the revised manuscript.

Author Response File: Author Response.docx

Reviewer 3 Report

The submitted manuscript “Improving the False Alarm Capability of the Extended Maximum Average Correlation Height Filter” presents a wavelet-modified EMACH filter for image processing and computer vision applications. The authors claim the proposed wavelet-modified EMACH filter shows improved accuracy and reduced false alarms. After carefully reviewing the results presented, I believe key evidences are missing to support above claims. Therefore, despite the manuscript is well structured and easy to follow, I recommend rejection for this manuscript because it doesn’t meet the standards and requirements of Photonics. My detailed comments are as below:

Detailed comments

1.       My main concerns are lack of MACH/EMACH results for false class images (car2 and car3). For example, Figure 5 provides no additional information on top of Figure 4. I believe to claim the proposed wavelet-modified EMACH filter performs better than MACH/EMACH filters, key metrics such as CPI vs angle of rotation for false class images with MACH/EMACH must be provided.

2.       On page 1 line 29-32, I find the English of this sentence hard to understand.

3.       On page 3 line 102 and 116, the symbol for the complex conjugate transpose doesn’t agree with the symbol used in the equations.

4.       In Figure 4, the text in the legend is a little too small. Also, different markers can be used to improve readability.

1.       There are a few long sentences used with “respectively”, which I believe might be easier to read to break them into shorter sentences so readers don’t have to search and correspond the terms.

Author Response

Authors’ Response to Reviewer’s Comments

Reviewer # 3

Comments and Suggestions for Authors:

The submitted manuscript, “Improving the False Alarm Capability of the Extended Maximum Average Correlation Height Filter,” presents a wavelet-modified EMACH filter for image processing and computer vision applications. The authors claim the proposed wavelet-modified EMACH filter shows improved accuracy and reduced false alarms. After carefully reviewing the results presented, I believe key evidences are missing to support the above claims. Therefore, despite the manuscript is well structured and easy to follow, I recommend rejection for this manuscript because it doesn’t meet the standards and requirements of Photonics. My detailed comments are as below:

Thank you very much for carefully reading the manuscript and providing valuable comments. We have gone through the comments and tried our best to respond them.

Detailed comments

1. My main concerns are lack of MACH/EMACH results for false class images (car2 and car3). For example, Figure 5 provides no additional information on top of Figure 4. I believe to claim the proposed wavelet-modified EMACH filter performs better than MACH/EMACH filters, key metrics such as CPI vs angle of rotation for false class images with MACH/EMACH must be provided.

We would like to thank the reviewer for raising this important issue. In the revised manuscript, we have shown the false class (car2 and car3) CPI versus angle of rotation for MACH/ EMACH in Fig. 4 and also show the average CPI values after applying the MACH, EMACH, and WEMACH on car1, car2, and car3 in Table 1.

For better clarity of the results obtained from the proposed WEMACH filter, we have explicitly shown the true and false class images in Fig. 5. A threshold line is also drawn on the same plot, which clearly separates the aforementioned two classes.

2. On page 1, lines 29-32, I find the English of this sentence hard to understand.

Thanks very much for pointing this out. We have tried our best to revise the manuscript correcting the English language.

1. On page 3, lines 102 and 116, the symbol for the complex conjugate transpose doesn’t agree with the symbol used in the equations.

Thanks very much for pointing out the mistake. The corrections have been incorporated in the revised manuscript.

2. In Figure 4, the text in the legend is a little too small. Also, different markers can be used to improve readability.

Thanks very much for pointing this out. The corrections have been incorporated in the revised manuscript.

Comments on the Quality of English Language

1. There are a few long sentences used with “respectively,” which I believe might be easier to read to break them into shorter sentences so readers don’t have to search and correspond the terms.

Thanks very much for pointing this out. Considering the structure of the statements for describing the results, long sentences have been written. We have tried to revise them in the revised manuscript.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The authors have responded comprehensively and positively to my concerns with details and clarity. The revised manuscript is improved overall. I believe the presented work is solid and claims are well supported now. A few further comments I would like to discuss with the authors are listed below:

Detailed comments

1.       I thank the authors to update Figure 4 with MACH/EMACH results for false class images (car2 and car3). I believe the claims are solid with the added data.

2.       In Figure 4, by comparing car3 WEMACH vs MACH/EMACH results, it’s clear WEMACH’s behavior deviates significantly for angle of rotation near 28°, which is not observed in car1 and car2. I am wondering whether the authors have any insights to discuss or explain this behavior. As illustrated in Figure 5 and Figure 10, this behavior is what limits the separation/robustness for the proposed algorithm.

3.       In the Abstract, “incorporating the benefits” should capitalize the first word.

4.       In Figure 11, spatial light modulator can be added into the caption as well.

5.       In Figure 12, I would suggest use small caps roman numerals to refer column i-v to be consistent with the main text.

Author Response

Author's Response to Reviewer’s Comments

Reviewer # 3 (Round 2)

Comments and Suggestions for Authors:

The authors have responded comprehensively and positively to my concerns with details and clarity. The revised manuscript is improved overall. I believe the presented work is solid and claims are well supported now. A few further comments I would like to discuss with the authors are listed below:

Thank you very much for again carefully reading the manuscript and providing valuable comments. We have gone through the comments and tried our best to respond them.

Detailed comments

1. I thank the authors to update Figure 4 with MACH/EMACH results for false class images (car2 and car3). I believe the claims are solid with the added data.

Thanks very much for reading the manuscript and encouraging remarks.

2. In Figure 4, by comparing car3 WEMACH vs MACH/EMACH results, it’s clear WEMACH’s behavior deviates significantly for angle of rotation near 28°, which is not observed in car1 and car2. I am wondering whether the authors have any insights to discuss or explain this behavior. As illustrated in Figure 5 and Figure 10, this behavior is what limits the separation/robustness for the proposed algorithm.

We would like to thank the reviewer for correctly pointing out the different behaviour at 28⁰. It is true that in the simulation, we have got the maximum value of CPI for the false class of car3 at 28⁰ angle of rotation in case of WEMACH filter (Fig. 5). We have taken the maximum (at 28⁰ of rotation for car3) CPI as the threshold value. It is true that at a particular angle of rotation there is abnormal behaviour but in most of the cases, we find much difference in true class and false class objects’ CPI and DR. This may be due to a situation that at this particular angle we find better match in terms of intensity. This can be considered as an aberration not a general rule. However, we need to carry out simulation study on much larger database and check the effectiveness. We shall take it up as a futuristic problem.

Although the CPI for false class (car3) maximum can change by changing the training data set, robustness is not affected by this fact. On the other hand, a key concept for robustness is the discrimination ratio (DR). Equation (13) describes the DR and the subsequent paragraph discusses its characteristics. The denominator of the DR contains the CPI value of the false class (car3), which has a minimum at 28⁰ (Fig. 10) in the case of WEMACH. This behaviour is consistent with Figure 5, i.e., the maximum CPI value (car3) leads to the minimum value for DR. A quantitative measure of robustness is the average DR value, which is approximately two times more in case of WEMACH compared to MACH and EMACH. This clearly establishes the advantage of our WEMACH filter over the MACH and EMACH. Also, more than one DR value is regarded as a robust case [for reference [2] (Chap. 8.) Yu, F. T. S.; Jutamulia, S. Optical Pattern Recognition. Cambridge Univ. Press, 1998].

3. In the Abstract, “incorporating the benefits” should capitalize the first word.

Thanks very much for pointing out the mistake. The correction has been incorporated in the revised manuscript.

4. In Figure 11, spatial light modulator can be added into the caption as well.

Thank you for the suggestion. Now, ‘spatial light modulator’ is added in the revised manuscript.

5. In Figure 12, I would suggest use small caps roman numerals to refer column i-v to be consistent with the main text.

We again thank the reviewer for pointing this out. The correction has been incorporated in the revised manuscript.

Author Response File: Author Response.pdf