Face Image Inpainting of Tang Dynasty Female Terracotta Figurines Based on an Improved Global and Local Consistency Image Completion Algorithm

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this study, the GLCIC algorithm is used to model how the original appearance of Tang Dynasty female terracotta figurines would look, aiming to provide a tool for the restoration processes of the figurines, which, according to the authors, have significant Heritage value. Through this work, the authors optimize the algorithm through data augmentation, guided filtering, and local enhancement techniques. The experimental results allow to propose an effective strategy for image inpainting where data scarcity is an issue.

The work is an exciting application of deep learning in CNNs, which, together with the Heritage background of the studied objects, could generate significant interest and impact of the paper.

Please address the following issues prior to publish:

 

INTRODUCTION

1.     LINES 39-51. Provide a more detailed review of “inpainting techniques” and add relevant references. This subject is of paramount importance to the aim of this work, and the reader should be able to clearly understand the significance and contribution of mathematical algorithms for systemizing such techniques. This, in turn, will lead to a proper restoration of the figurines or any other cultural heritage object analyzed through the GLCIC algorithm.  

2.     Consider adding a detailed discussion of alternate methods, such as hyperspectral imaging analysis, which has also provided interesting results for completing missing sections in paintings, codexes, and painted sculptures.

 

RELATED LITERATURE

3.     Consider renaming this section “Historical Background.”

4.     Figures 1, 2, 4, etc. Add a proper scale so the reader understands the size of the figurines.

5.     Information in sections 2.2 and 2.3 could be better placed in the introduction.

 

MATERIALS AND METHODS

6.     LINES 291-294. In my opinion, the study aims to provide a tool for modeling the missing parts of the figurines. The actual restoration is a different process. Correct such difference along the text.

 

RESULTS AND DISCUSSION

7.     LINES 476-488. Describe in better detail the process of assigning weights for facial deficiency evaluation. In particular, try to elaborate on the “Based on the ratings from five experts”, which are the criteria from such experts? Are the different criteria being uniformized following any systematic procedure? It might be convenient to summarize the relevant aspects and parameters considered in a table. It is essential to resolve this issue since the reader might be misled to consider that the results obey a non-standardized weighting method.

 

REFERENCES

8.     DOI’s are not working. Please check if any parameter is missing in the PDF document or if the format is incorrect.

9.     Instead of citing a thesis, provide relevant publications associated with it.  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors provided an interesting manuscript focused on implementing artificial intelligence (AI) algorithms for the restoration of historical relicts. It is quite perspective area for AI, but in this approach I see some weaknesses.

1. In their manuscript, authors are very focused on their specific use case in my opinion the introduction and overview of similar research must be much wider showing the full spectrum of possibilities to implement the AI for the restoration of missing/lost information. 

2. Research method and used algorithms should be explained, providing more technical details.

3. The reliability of the research results has no fundamental reference, ground troth is defined by the 5 experts, but it does not prove that restoration is correct. In my opinion to claim this method as valid and reliable, it is necessary to validate it in similar use case with existing or artificially generated known samples and evaluate the accuracy of inpainting by quantitative indisputable criteria.

Minor issues:

1. low quality figures, and missing explanations for separate parts of the figures;

2.  Conclusions are not supported by obtained quantitative results;

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The paper is in better shape now, and it is ready to be published. 

Author Response

Thank you for your review and positive feedback. I am glad to hear that the changes meet your expectations.

Reviewer 2 Report

Comments and Suggestions for Authors

Authors provided an improved version of their manuscript, the majority of my remarks were taken into account. Now the quality of manuscript is much higher. However, I noticed some technical issues:

1. Not all a, b, c parts of figures have their titles  (figures 11-12), not all parts of figures are specified (figures 4, 8).

2. Conclusions do not contain any numeric values, it would be nice if authors could provide some quantitative evaluation of their achievements.

Author Response

"Please see the attachment."

Author Response File: Author Response.pdf