An Edge Detection Algorithm for SEM Images of Multilayer Thin Films

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes an edge extraction method for scanning electron microscope backscattered electron (SEM-BSE) images with unclear cross-sections of multilayer thin films. The method involves several steps: initial preprocessing using various gray-scale image processing techniques to eliminate noise and smooth the image, selection of a high-quality denoised image based on evaluation indexes, segmentation using a masking algorithm, enhancement through different image enhancement algorithms, and logical algorithm combination. Finally, Canny edge detection is performed.

I find the paper lacks potential for the following reasons:

1-      Heavy preprocessing, especially when followed by a standard edge detection technique like Canny, might be seen as a less innovative approach. While preprocessing steps are often necessary to enhance image quality, excessive manipulation can sometimes be counterproductive or unnecessary.

2-      It is not clear whether the primary improvement in edge extraction comes from the standard edge detection algorithm or the preprocessing methods applied, therefore it is advised to try several other edge detections after the proposed edge detection and compare it to Canny to justify its need if it is found the best, moreover, it would be nice to try the Canny edge detection without the preprocessing methods and see the results to justify the use of such a preprocess. Also, the order of the preprocess methods needs to be investigated.

3-      The time consumed by the preprocessing steps is not discussed, and if the time tradeoff aligns well with the gained edge accuracy.

4-      The proposed algorithm is tested only on one cross-section SEM-BSE image. You can't draw significant conclusions on such weak evidence.

5-      There is no literature review.

6-     A comparative analysis with other state-of-the-art methods and benchmarking on standard image databases could provide a clearer perspective on the proposed method's effectiveness, rather than testing the method on only one image.

7-      The captions of the tables and figures are not informative.

Comments on the Quality of English Language

 Extensive editing of English language is required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This paper discussed the edge detection for SEM-ESB image of multi-layer thin film using mathematical approach. The authors investigated many conventional methods to the cross-sectional images and showed the effects to detect the boundary of the layers. However, this paper must be improved much, because some points of this paper is unclear. Please, check the following comments.

1. The algorithm was written in section 2 using text, thus it is difficult to understand the procedure and effects. Figure 1 is unclear to describe the proposed method, thus it is better to add the images at each processing or explain the role of each step. It will be helpful to understand the proposed algorithm.

2. The many sentences in section 3 results describes the method, thus methods and results are mixed in the section. In my opinion, some text should be moved to section 2. NLM, transforms and others are not described in section 2, what are the roles of the methods in your paper?

3. The specification of the multi-layer sample must be shown in section 3.1. Solar cells usually have flat and uniform surfaces. However, the figures in the results seems to be a tissue or cell in biology.

4. The titles of figures 3-5 should be corrected. For example, Figure 4(a) seems to be an image after Gaussian filter, and Figure 4(b) is gray level distribution of the blue line. The tile of figure 5 is not insuitable. Furthermore, the blue lines in figures 3-5 should be positioned at the same position to compare filtering effects.

5. In section 3.7, masks 1-5 are written in the text, but they are not defined. You'd better to show them in a figure.

6. The authors compare the results of NLM in Table 2, the proposed algorithm and others. However, the proposed method (our algorithm) is not clear in section 3 and 4. The definition of the score is not shown in the text. The processing time should be shown.

7 Abbrivations, such as SEM BSE BSD ESB, SE2, FIB, NLM and so on,  should be defined at the first uage.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revised version of this paper has significantly improved.

Reviewer 2 Report

Comments and Suggestions for Authors

The authors answered the comments properly and my decision is accept. Great job.