Survey on Data Hiding Based on Block Truncation Coding
Round 1
Reviewer 1 Report
This paper introduces a systematic survey of the researches based on block truncation coding adopting simulations of many representative methods that are effective through experiments expected to be directing future research within this data hiding field. The results show that the work is very much pioneering, but needs to be revised thoroughly, in order to be ready as publication, as noted within the following points that have to be fulfilled:
1- The abstract requires to one more phrase briefing overview of motivation behind this research. The Abstract intention is to have some indication of the practical benefit and contribution as well as specific originality trying to attract the reader to select the paper to read and refer to as well as get motivated toward the work to continue research in similar direction.
2- The reversible and non-reversible data hiding is described brief and to the point to be interesting. To be complete, the work needs to involve the following references:
== "Improving data hiding within colour images using hue component of HSV colour space", CAAI Transactions on Intelligence Technology, IET (IEE), in press (2021) https://doi.org/10.1049/cit2.12053
== "Novel Embedding Secrecy within Images Utilizing an Improved Interpolation-Based Reversible Data Hiding Scheme", Journal of King Saud University - Computer and Information Sciences, ISSN:1319-1578, in press (2020) http://doi.org/10.1016/j.jksuci.2020.07.008
3- The numerical examples within the paper should be revised in order to be more clear as represented and simplified. These numerical examples are to make the reader walk through and understand the basic research experimentation framework. The numerical example computation should be small, simple and easy to understand. It is to allow any reader and future researcher to build upon the original work and repeat all the tests, i.e. to recover the benefits and propose advancements.
4- The experimental study shown in Figure 8 and Figure 9 needs revision and elaboration highlighting the benefit to avoid any possible confusion. Please focus on analyzing the strange differences within the results presented.
5- The work comparison of Figure 10 needs to cover testing images: Baboon and Goldhill, to complete the study. These remarks elaboration on all images are to be revised proofing fairness in the results and comparison. It needs to show some indication with more justification of these results. Explain more the observations feedback of the proposed work vs. others, i.e. as showing different results of evaluation with other the results.
6-Conclusion needs reconsideration. It needs to highlight the research main contribution with some brief indications and numerical improvement percentages to keep with the reader. Also, the conclusion needs to present some more ideas of open research and future work for researchers to build upon for further advancements.
Author Response
Reviewer#1
We appreciate the reviewer’s valuable comments. The followings are our point-by-point responses:
1. The abstract requires one more phrase briefing overview of motivation behind this
Response: Abstract was rewritten as follows.
Data hiding technology has achieved many technological developments through continuous research over the past 20 years along with the development of Internet technology and is one of the research fields that are still receiving attention. At the beginning of the study, there was an intensive study on the digital copyright issue, and after that, interest in the field of secret communication is increasing. In addition, research on various security issues using this technology is being actively conducted. Research on data hiding is mainly based on images and videos, and there are many studies using JPEG and BMP in particular. This may be due to the use of redundant bits that are characteristic of data hiding techniques. On the other hand, block truncation coding-based images are relatively lacking in redundant bits useful for data hiding. For this reason, researchers began to pay more attention to data hiding based on block-cutting coding. As a result, many related papers have been published in recent years. Therefore, in this paper, the existing research on data hiding technology of images compressed by block-cut coding among compressed images is summarized to introduce the contents of research so far in this field. We simulate a representative methodology among existing studies to find out which methods are effective through experiments and present opinions on future research directions. In the future, it is expected that various data hiding techniques and practical applications based on modified forms of absolute moment block truncation coding will continue to develop.
2. To be complete, the work needs to involve the following references:
Response: We cited two papers the reviewer introduced.
http://doi.org/10.1016/j.jksuci.2020.07.008; https://doi.org/10.1002/cpe.5322
3. The numerical example computation should be small, simple and easy to understand. It is to allow any reader and future researcher to build upon the original work and repeat all the tests.
Response: Thank you for your advice. The example included in the paper explains how to hide data using a simple formula. The reason for including "Examples" in the paper is to help readers understand the data hiding formula described in the paper. If the reader needs a more thorough explanation of this thesis, I think of course using the references. Since this paper is a survey paper, it is a limitation of this paper that it cannot completely explain the contents of all papers. However, according to the reviewer's advice, some errors in formulas or lack of explanation in the text were rewritten.
4. The experimental study shown in Figure 8 and Figure 9 needs revision and elaboration highlighting the benefit to avoid any possible confusion. Please focus on analyzing the strange differences within the results presented.
Response: Thanks for the good advice. We reorganized Figures 8 and 9 with deep insight again. The rewritten paragraph reads:
Figure 8 shows the comparison of the EC of the representative DH methods (i.e., Ou & Sun[30], Huang et al.[40], Hong[34], Kim et al.[42], and Horng et al.[44]) using six images ([55]). Here, EC was measured in case that the image quality was above 30dB. The reason is that the measured EC when less than 30 dB may not be suitable for use in real systems. Although the other research ignores this, in Figure 8, EC was measured with this criteria. Of course, even if the experimental result maintains 30dB, it cannot be guaranteed that the image is not absolutely broken, but for the objective of the experiment, the lower limit of the image is set to 30dB. The methods in Figure 8 are used the DBS method ([30]) commonly, which is to substitute bits of the same size in a 4×4 bitmap. The merit of DBS is essential for hiding many bits (1 bpp), and the image quality does not deteriorate significantly even after performing DBS, because on average 50% of the original pixel values of the block are maintained after DBS is performed. The methods of Huang et al.([40]) and Horng et al.([44]) respectively proposed a DH technique that exploits the difference between two pixels to sufficiently hide the data at the quantization level. Huang et al.’s method ([40]) and Horng et al.’s method ([44]) have average bpp of 1.3 and 1.28, respectively, and Huang et al.’s method is slightly superior. Since Ou & Sun’s method ([30]) hides data only with the DBS method, the data hiding rate corresponds to 1 bpp. The methods of Hong ([34]) and Kim et al. ([42]) is 1.18 bpp and 1.16 bpp, respectively, which is superior to DBS, but it is measured to be slightly inferior in performance compared to the Huang et al.’s method. AMBTC-based DH does not have enough redundant bits compared to standard grayscale images, so achieving an average of 1.2 bpp while maintaining about 30 dB is a very valuable achievement.
Figure 9 compares and evaluates the performance of a typical RDH method. The methods Lin & Liu([47]), Shie et al.([48]), and Chang et al.([51]) are AMBTC-based RDH methods that were proposed relatively early. Lin & Liu([47]) and Shie et al.([48]) are RDH methods using histograms, and Chang et al.’s([51]) method is a method using HC(7,4). The data hiding performance is 0.08bpp, 0.13bpp, and 0.24bpp, respectively. These methods show the potential of early BTC-based RDH and can be considered valuable. On the other hand, the relatively recent RDHs (Kim et al.([49]), Li et al.([52]) and Lin et al.([53])) ranges from about 0.57bpp to about 1.3bpp. In particular, Lin et al. method ([53]) utilizes multiple histograms. So it shows the best performance. Like the existing gray image-based RDH, it can be seen that the HS method is usefully used in the BTC-based RDH system. It is evaluated that methods capable of higher EC will be proposed through the development of a modified method of HS in the future. Meanwhile, sufficient bitmap resources are required to hide a large amount of data. For this purpose, it is a recent trend to use a method of increasing the number of pixels in the bitmap, but this is only expedient. Since the performance is improved by distorting the original method of AMBTC, the compression performance is of course badly affected.
5. The work comparison of Figure 10 needs to cover testing images: Baboon and Goldhill, to complete the study. These remarks elaboration on all images is to be revised proofing fairness in the results and comparison.
Response: Thanks for the good advice. We added an experiment for two images in Figure 10. In addition, we have further improved the description of the sentences described here to make it easier for readers to read and understand.
Figure 10 shows a chart where the correlation between EC and PSNR can be observed when six conventional methods are applied to six standard images. The existing five methods can see the data hiding ability through Figure 8. In Figure 10(a)∼(f), it is common that the increase in data deteriorates the image quality. Because it is accompanied by distortion of the image. Nevertheless, depending on the performance of the proposed method, the PSNR may be lowered gently or rapidly. We can observe such a basic point. In Figure 10, it can be visually judged that the method proposed by Horng et al. is the method that best meets the two evaluation criteria with the most data. It can be observed that Huang et al.’s([40]) method is the best when only the data hiding capacity is used as a standard. In the figure, if the data hiding capacity is limited to 0.2∼1.5, Horng et al.’s([44]) method may not be recommended. In this case, it can be said that all existing methods are excellent. Figure 10(f) is an image belonging to a high-frequency image, and since the quality of the cover image itself belongs to about 27dB, when data is hidden in this cover image, it is lowered to a maximum of about 18dB, so it is not suitable for data hiding by BTC compression.
6. Conclusion needs reconsideration. It needs to highlight the research main contribution with some brief indications and numerical improvement percentages to keep with the reader.
Response: Thanks for the advice. We have rewritten our conclusions to clarify to our readers the subject of our proposed paper. Here is the rewritten conclusion.
We briefly looked at the studies and analyzed their performance using simulations after classifying the various studies on AMBTC-based DH. The cover images used in the experiment are Lena, Boat, Baboon, Peppers, Goldhill, and Airplane. The PSNRs of these cover images are 33.65, 31.57, 26.97, 34.09, 32.83, and 32.03, respectively. Therefore, hiding an amount of data while maintaining more than 30dB can be a very difficult challenge. However, in the case of a high-frequency image like the Baboon image, it is difficult to use it as a cover image because the image quality does not reach 30 dB after compression with BTC. With DH and RDH, the maximum data hiding performance is 1.3bpp. However, if data is hidden by 1.3 bpp, it may not be suitable for the purpose of confidential communication or copyright protection. Existing methods can be a good alternative if existing methods are selected according to the purpose. Most of the methods proposed so far are actually modified methods used in gray images to apply the compression method. In other words, most of the methods verified in the grayscale image were applied to BTC and showed good performance. In the future, we intend to review the papers of various methods to consider future research directions.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
In this work, the authors provide an interesting survey on data hiding based on block truncation coding. The paper turns out to be well written and is well structured, and the methodology used to complete the survey is based on solid foundations. The authors have carried out a careful analysis of the literature by taking into consideration very recent works.
However, I think that the paper is still affected by some issues, which should be addressed:
- In some parts of the paper, the clarity and editorial quality of the paper weaken. As a consequence, such parts result to be quite difficult to read. Therefore, I would suggest carefully improving the prose of writing to make this paper easier to read.
- The set of references should be expanded by taking into account the following set of related works:
[1] https://doi.org/10.1002/cpe.5322
[2] https://doi.org/10.1080/02564602.2019.1698978
- An accurate proofreading is strongly required since the paper is affected by some typos and formatting issues.
Author Response
Reviewer#2
We appreciate the reviewer’s valuable comments. The followings are our point-by-point responses:
1. I would suggest carefully improving the prose of writing to make this paper easier to read.
Response: Thanks for the advice. We followed the reviewer's advice to make the articles easier to read. This section spans several parts of the paper, so we would like reviewers to read the paper. The corrected part is marked in blue.
2. The set of references should be expanded by taking into account the following set of related works:
Response: Thanks for the advice. We wrote the thesis using the thesis recommended by the reviewer's advice.
3. An accurate proofreading is strongly required since the paper is affected by some typos and formatting issues.
Response: According to the reviewer's advice, the typos were checked.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Satisfied with modification. Good work to be accepted.
Reviewer 2 Report
The authors addressed all the issues I pointed out by carefully revising the paper. The paper has been substantially improved with respect to the relative previous version. Due to the reasons given above, the paper is now ready to be accepted for publication.
