Ghost-Free Multi-Exposure Image Fusion Technology Based on the Multi-Scale Block LBP Operator
Round 1
Reviewer 1 Report
This work presents a methodology to fuse sequences of images including both static and dynamic objects. The main challenge of this problem is the generation of 'natural' results regarding brightness, dark areas, saturation, blur, etc. Furthermore, dynamic scenes are also covered to avoid the ghosting effect. The overall method is well presented, readable and innovative. Finally, the proposed method is evaluated according to two metrics that endorse the subjective evaluation that was previously presented by depicting fused images.
First of all, I have attached the revised document with comments regarding the use of English, errors, and some doubts. For example, the abstract describes two different weighting methods, whereas the summary image of your work only shows a single workflow. Did you maybe refer to Gaussian and Laplacian fusion?
The introduction and related work are good enough for understanding the background of the problem. The use of multi-scale LBP is highlighted to distinguish this work from previous, though there are already some multi-resolution studies. With this in mind, I wonder how would your method look with LBP applied to neighbourhoods of fixed size instead of varying the radius (multi-scale LBP). That would be an appreciated comparison to be included in your evaluation (to the best of my understanding, there is no previous method with this architecture, otherwise omit this).
Regarding the methodology, it is well presented and can be easily followed. There are some missing explanations of nomenclature that I highly encourage you to address for a better understanding of the work. Also, some steps depend on values that are fixed. There is no further explanation on why these values are used, or evaluation endorsing them. Do they work similarly for sequences of images that are significantly different? Maybe they could be explained similarly to the free parameter set to 15 in your evaluation. Also, there is a significant number of terms included in your methodology. Maybe, the summary figure could include the nomenclature of such image information next to each step, as included in your description (L, W, I, etc.).
The evaluation aims to objectively assess the results using two metrics. As an improvement, there could be a brief description of the two metrics and how they are calculated, so the article is self-contained. Also, subjective evaluation always shows that your method outputs better images, with clearly defined edges, etc. However, the objective assessment does not present your method as the best for every sequence. This is not necessarily bad, as it is just a number trying to measure something which is challenging to automatize. However, some in-depth revision of why this occurred would highly improve your work. Maybe the metric formulas give some light on this.
Finally, I would highly encourage the authors to revise the writing. The manuscript is readable, though there are some sections harder to read than others (e.g., results). Comments for improving readability are attached in the PDF. Also, comments in subjective evaluation are too dense. Either cut it a bit, make it clearer with a list, or even a table with the main features that are likely sought in fused images: well-defined edges, correct brightness, non-overexposed or underexposed, etc.
Comments for author File: 
Comments.pdf
Author Response
Dear Reviewers,
Thank you very much for your time involved in reviewing the manuscript We
also appreciate your clear and detailed feedback and hope that the explanation has
fully addressed all of your concerns. Attached is my point-by-point response to your
comments, please see the attachment.
With Kind Regards
Prof. Zhengping Li.
Author Response File: 
Author Response.docx
Reviewer 2 Report
In this paper, the authors proposed two methods for ghost-free multi-exposure image fusion. The work is novel in nature and has good scientific contribution. It contains all components of a scientific paper. There are some minor changes which are to be done
1. Some of the abbreviations are missing. For example- HDR not abbreviated on its first occurrence.
2. In statement on page number 9, “This paper compares the method with 7 existing state-of-the-art techniques [15, 17- 23]” - 7 should be replaced by 8. As 15, 17,18,19,20,21,22,23 are 8 in number not 7.
3. In sub-section 3.2, “It can be seen from the table that in most cases,” which table is referenced here.
4. In Conclusion, “In this algorithm”, replace algorithm by model.
Author Response
Dear Reviewers,
Thank you very much for your time involved in reviewing the manuscript We
also appreciate your clear and detailed feedback and hope that the explanation has
fully addressed all of your concerns. Attached is my point-by-point response to your
comments, please see the attachment.
With Kind Regards
Prof. Zhengping Li.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Dear Authors,
Thanks for addressing the suggestions I made, as well as for solving my doubts. The manuscript has significantly improved and my questions were perfectly solved.
I have attached the revised manuscript with a few comments, though they are minor errors/improvements. Also, the tables comparing different methods could be summarized with positive/negative symbols or 'Yes/No', though it is up to the author to do that.
Thanks for including the suggested evaluations and explaining formulas as well as their numeric results, as they have greatly contributed to improving the understanding of the method.
Comments for author File: Comments.pdf
Author Response
Dear Reviewer,
Thank you very much again for taking the time to review the manuscript, we also appreciate your clear and detailed feedback and hope that the explanation has fully addressed all of your concerns. Attachment is my explain the details of the manuscript revision to you separately, please see the attachment.
With Kind Regards
Prof. Zhengping Li.
Author Response File: Author Response.pdf
