A Large Capacity Histogram-Based Watermarking Algorithm for Three Consecutive Bins
Round 1
Reviewer 1 Report
1.
In the abstract there are concepts like histogram watermarking and geometric attacks, which are mentioned without providing a brief description.
The abstract would be more readable if such concepts were briefly explained.
For instance, the lines 32-33 of the introduction could be re-used in the abstract as a definition of watermarking.
2.
In the introduction there are concepts like histogram
geometric attacks and stego-image, which are mentioned without providing a description.
The introduction would be more readable if every concept was explained.
3.
The are acronyms like HVS and MSE, which are mentioned without providing their meaning.
4.
The caption of figure 8 is rather long. The comments to the figure should be places in the main text, reducing the length of the caption.
5.
Table 3 could be placed in only one page.
6.
Conclusions are brief.
Can they be extended with more discussion about results, applications, and future work?
7.
English needs many corrections. They are annotated in the pdf file in attachment.
Comments for author File: 
Comments.pdf
Author Response
Response to Reviewer 1 Comments
We would like to express our gratitude to you for your time and effort in reviewing and processing our manuscript and especially providing constructive comments and valuable suggestions for significantly improving the manuscript. Following these comments and suggestions, we have made changes in the revised manuscript. A point-by-point reply to the reviewer’s comments is given below, where in each case we quote the referee’s comments and then explain how we have revised the paper to accommodate the revisions requested.
Point 1: In the abstract there are concepts like histogram watermarking and geometric attacks, which are mentioned without providing a brief description.
The abstract would be more readable if such concepts were briefly explained.
For instance, the lines 32-33 of the introduction could be re-used in the abstract as a definition of watermarking.
Response 1: Many thanks for your excellent suggestion for improving our manuscript. We have added some explanations to the revised version, as shown below:
In the abstract:
“Histogram watermark, which performs watermark embedding by slightly modifying the histogram of original image, has been a hot research topic in information hiding technology due to the superiority that its pixel modification during the watermark embedding process is independent of the pixel position. This property makes histogram-based watermark strong resistance to geometric attacks, such as cropping attack, crossed attack, rotation attack, etc.”
Point 2: In the introduction there are concepts like histogram geometric attacks and stego-image, which are mentioned without providing a description.
The introduction would be more readable if every concept was explained.
Response 2: Thanks for your helpful and valuable comment. We have mentioned these concepts in the revised manuscript.
In the first paragraph of the section Introduction:
“Geometric attacks primarily introduce some synchronization errors between the encoder and the decoder. Under geometric attacks, the watermark is still present, whereas the detector is no longer able to extract it.”
“Meanwhile, the image quality of stego-image, which is generated by modifying a part of pixel values of original image for carrying watermark information, together with the capability to resist some geometric attacks, can also be ensured.”
Point 3: There are acronyms like HVS and MSE, which are mentioned without providing their meaning.
Response 3: We are sorry for some careless mistakes. We have added some explanations to the revised version, as shown below:
In the first paragraph of this section 3.2:
“In order to avoid this, Hu et al. [13] initially proposed block-based method after learning human visual system (HVS) [18] and applying Xiang et al. idea.”
“When one current pixel is modified, it is sorted according to the mean square error (MSE) of the block, …”
Point 4: The caption of figure 8 is rather long. The comments to the figure should be places in the main text, reducing the length of the caption.
Response 4: Many thanks for your excellent suggestion for improving our manuscript. We have corrected them, as shown below:
In the caption of Figure 8:
“Figure 8. Robustness to cropping, crossed and rotation for Baboon and Barbara (M=2, T=5 and C=24). (a) and (b) the cropped watermarked images of Baboon and Barbara; (c) and (d) the crossed watermarked images of Baboon and Barbara; (e) and (f) the rotation watermarked images of Baboon and Barbara; (g) and (h) the AWGN watermarked images of Baboon and Barbara; (i) and (j) the JPEG watermarked images of Baboon and Barbara.”
Point 5: Table 3 could be placed in only one page.
Response 5: Thank you for providing the advice for us. We have modified the Table 3 in the Section 4.1.2 in the revised version, as shown below:
Table 3. Parameter settings for different algorithms.
Payload(bit) | 32 | 48 | 56 | 63 | |
Xiang et al. [13] | embedding range | [0.3A,1.7A] | [0.3A,1.7A] | - | - |
T | 2 | 6 | - | - | |
M | 2 | 2 | - | - | |
bins’ number | 64 | 96 | - | - | |
Zong et al. [17] | embedding range | [0,255] | [0,255] | - | - |
T | 2 | 4 | - | - | |
M | 3 | 2 | - | - | |
bins’ number | 64 | 96 | - | - | |
Scheme I [18] | embedding range | [15,244] | [15,244] | [15,244] | [15,244] |
T | 2 | 1.5 | 1.25 | 1.5 | |
M | 2 | 2 | 2 | 2 | |
bins’ number | 64 | 96 | 112 | 126 | |
Scheme II [18] | embedding range | [15,244] | [15,244] | [15,244] | [15,244] |
T | 3 | 2+sqrt(0.5) | 2.5 | 37/16+sqrt(0.5) | |
M | 2 | 4 | 8 | 4 | |
bins’ number | 64 | 64 | 64 | 84 | |
The proposed | embedding range | [0,255] | [0,255] | [0,255] | [0,255] |
T | 2 | 2 | 2 | 2 | |
M | 2 | 2 | 2 | 2 | |
bins’ number | 32 | 48 | 56 | 63 | |
Point 6: Conclusions are brief.
Can they be extended with more discussion about results, applications, and future work?
Response 6: Many thanks for your excellent suggestion for improving our manuscript. The conclusion has been rewritten as:
“In this paper, a large capacity histogram-based watermark algorithm is proposed. The watermark design of the proposed method has satisfactory functionalities in terms of robustness to geometric attacks and common image processing operations. In the embedding range selection, possible bad bins will suffice to be eliminated successfully with the assist of a predefined threshold. In the watermark embedding design, by taking the approximately-equal number of adjacent histogram bins into account, the proposed method makes every three consecutive bins carry 3-bits watermark; or rather, it realizes the maximum embedding rate (1bpb) at the first time. The experiment results demonstrate that the proposed watermark algorithm can provide a preferable tradeoff between embedding capacity and robustness, especially suitable for applications where robustness and embedding capacity are essential, such as copyright protection requirement for enough amount of watermark information and higher robustness. However, a possible limitation for the proposed histogram-based watermarking is the situation where certain operations may make the histogram shape distorted much. In this case, the watermark extraction will be affected directly. In our future research, one consideration is to improve the newly designed appropriately-equal condition so that when the histogram shape is distorted seriously, the accuracy of watermark extraction is still acceptable.”
Point 7: English needs many corrections. They are annotated in the pdf file in attachment.
Response 7: We are sorry for these careless mistakes. We have corrected them, as shown below.
In the abstract:
“According to these cases, we embed information number 0, 1, 2, 3, 4, 5, 6 or 7, respectively.”
In the second paragraph of the section Introduction:
“A variety of watermarking related methods have been proposed so far, and can be classified into three categories: …”
“Feature area embedding based methods were to extract feature points [11] or salient points such as the eyes and the mouth [12] firstly, and then embed watermark into them.”
In the fourth paragraph of the section Introduction:
“With their method, some possible bad bins can also be eliminated to some degree.”
“Obviously, this restricts many applications, especially limited for those with a strict requirement for watermark capacity.”
In the first paragraph of the section 2.1:
“… 1) modifying a portion of pixels to embed watermark is only related to the modification count of each grayscale, not to the position of pixels; …”
In the first paragraph of the section 2.2:
“… then they divided these bins into a series of groups where each of them contains two neighboring bins,…”
“…as shown in Figure 1.”
In the second paragraph of the section 2.2:
“This will have a direct influence on the watermark embedding capacity, in turn, the watermark capacity, lowered greatly, restricts many application scenarios.”
In the first paragraph of the section 3.2:
“…Hu et al. [14] initially proposed block-based method after learning human visual system (HVS) [19] and applying Xiang et al. idea.”
In the second paragraph of the section 3.3.1:
“Note that having eight types of situations is mainly attributed into two factors: one is that the approximately-equal pixel number of adjacent bins is considered here.”
“Unlike other histogram-based watermark methods [14-16], these methods do not consider this case in their watermark designs, thus only 1-bit watermark data can be embedded into each three consecutive bins.”
In the third paragraph of the section 3.3.1:
“Suppose a, b and c denote the original three consecutive bins before watermark embedding,…”
“Others are all similar to the above, except for a special case,…”
“…the default value (the case which does not satisfy certain proportional relationships is set as an approximately-equal treatment) is used to weaken the fragile relationship for the adjacent bins which are approximately-equal.”
In the Case 1 of the section 3.3.2:
“…other operations are as follows:”
“…the order in the above steps cannot be changed because these pixels chosen from Bin2 should not contain those that have been adjusted in Bin1.”
In the Step4 of the section 3.3.2:
“According to the above histogram modification/adjustment procedure to modify the histogram shape of each group, when adjusting the pixel value,…”
In the Step5 of the section 3.3.2:
“These above steps are repeated until the whole PN sequence is embedded. The final modified image is the watermarked image.”
In the section 3.3.3:
“When the watermarked images are illegally manipulated during transmission,…”
“In a similar way, the proposed method can extract all of the other watermark information.”
“Similarly, we also give a partial version of pseudo-code of the watermark extraction process, as shown in Algorithm 2.”
In the section 4.1:
“In Xiang et al. method [13], two consecutive bins are combined into a group and the bin width is set as M=2, thus each four bins carry 1-bit watermark.”
In the section 4.1.1:
“…Structural Similarity Index Measure (SSIM) are used for judging distortion degree between the original and watermarked images.”
“This is because there are much more pixels needed to be modified to embed more payload bits.”
In the section 4.1.2:
“It can be observed that our method provides a preferable visual quality of the watermarked image, meanwhile, it ensures more information can be embedded into the pixels located at the same bins. Our algorithm, with respect to the existing ones, improves the embedding capacity by at least 100%.”
In the section 4.2:
“For the four test images, we set the experimental parameter as the following: 1) the bin’s width is set as 2; 2) the threshold is set as 5; 3) the capacity is set as 24 bits.”
In the last paragraph of the section 4.2:
“Because the histogram of modified image incurred during the cropping and crossed attacks keeps approximately the same as that of the original image without suffering from attack, the watermark algorithm has higher robustness to cropping and crossed attacks.”
“Finally, the embedding of watermark information mainly depends on the statistical property of the histogram and is independent of the size of an image, thus this proposed watermark method can be applied to images with any size.”
In the caption of Figure 8:
“Figure 8. Robustness to cropping, crossed and rotation for Baboon and Barbara (M=2, T=5 and C=24). (a) and (b) the cropped watermarked images of Baboon and Barbara; (c) and (d) the crossed watermarked images of Baboon and Barbara; (e) and (f) the rotation watermarked images of Baboon and Barbara; (g) and (h) the AWGN watermarked images of Baboon and Barbara; (i) and (j) the JPEG watermarked images of Baboon and Barbara.”
In addition, after checking the total manuscript, we have corrected all possible error grammar expression thoroughly in the revised version. Please refer to the revised version for detailed contents.
Author Response File: Author Response.pdf
Reviewer 2 Report
The paper presents an interesting approach for image processing-based method. The following are some suggestion to improve the quality of the paper: 1. Line 16: Please describe why the histogram watermarking is a hot research topic. What is the general application of the proposed method? 2. Please increase the thickness of lines in Figure 8. Most of the lines are not clear enough. Authors may consider to remove the grid on the Figures and change the green line color with another color. 3. The caption of Figure 8 is too long. If possible please remove some description and make it more concise. Authors may put these content in the body of the paper. 4. Please add more description in the Conclusion section. Authors may add some more contents about the potential application of the proposed method and the limitation of the method (if any).Author Response
Response to Reviewer 2 Comments
We would like to express our gratitude to you for your time and effort in reviewing and processing our manuscript and especially providing constructive comments and valuable suggestions for significantly improving the manuscript. Following these comments and suggestions, we have made changes in the revised manuscript. A point-by-point reply to the reviewer’s comments is given below, where in each case we quote the referee’s comments and then explain how we have revised the paper to accommodate the revisions requested.
Point 1: Line 16: Please describe why the histogram watermarking is a hot research topic. What is the general application of the proposed method?
Response 1: Many thanks for your excellent suggestion for improving our manuscript. We have added some explanations to the revised version, as shown below:
In the abstract:
“Histogram watermark, which performs watermark embedding by slightly modifying the histogram of original image, has been a hot research topic in information hiding technology due to the superiority that its pixel modification during the watermark embedding process is independent of the pixel position. This property makes histogram-based watermark strong resistance to geometric attacks, such as cropping attack, crossed attack, rotation attack, etc.”
Point 2: Please increase the thickness of lines in Figure 8. Most of the lines are not clear enough. Authors may consider to remove the grid on the Figures and change the green line color with another color.
Response 2: Thanks for your helpful and valuable comment. We have modified the Figure 8:
| |
(a) | (b) |
(c) | (d) |
(e) | (f) |
(g) | (h) |
(i) | (j) |
Figure 8. Robustness to cropping, crossed and rotation for Baboon and Barbara (M=2, T=5 and C=24). (a) and (b) the cropped watermarked images of Baboon and Barbara; (c) and (d) the crossed watermarked images of Baboon and Barbara; (e) and (f) the rotation watermarked images of Baboon and Barbara; (g) and (h) the AWGN watermarked images of Baboon and Barbara; (i) and (j) the JPEG watermarked images of Baboon and Barbara.
Point 3: The caption of Figure 8 is too long. If possible please remove some description and make it more concise. Authors may put these content in the body of the paper.
Response 3: Thank you for providing the advice for us. We have corrected the caption of Figure 8:
“Figure 8. Robustness to cropping, crossed and rotation for Baboon and Barbara (M=2, T=5 and C=24). (a) and (b) the cropped watermarked images of Baboon and Barbara; (c) and (d) the crossed watermarked images of Baboon and Barbara; (e) and (f) the rotation watermarked images of Baboon and Barbara; (g) and (h) the AWGN watermarked images of Baboon and Barbara; (i) and (j) the JPEG watermarked images of Baboon and Barbara.”
Point 4: Please add more description in the Conclusion section. Authors may add some more contents about the potential application of the proposed method and the limitation of the method (if any).
Response 4: Many thanks for your excellent suggestion for improving our manuscript. The conclusion has been rewritten as:
“In this paper, a large capacity histogram-based watermark algorithm is proposed. The watermark design of the proposed method has satisfactory functionalities in terms of robustness to geometric attacks and common image processing operations. In the embedding range selection, possible bad bins will suffice to be eliminated successfully with the assist of a predefined threshold. In the watermark embedding design, by taking the approximately-equal number of adjacent histogram bins into account, the proposed method makes every three consecutive bins carry 3-bits watermark; or rather, it realizes the maximum embedding rate (1bpb) at the first time. The experiment results demonstrate that the proposed watermark algorithm can provide a preferable tradeoff between embedding capacity and robustness, especially suitable for applications where robustness and embedding capacity are essential, such as copyright protection requirement for enough amount of watermark information and higher robustness. However, a possible limitation for the proposed histogram-based watermarking is the situation where certain operations may make the histogram shape distorted much. In this case, the watermark extraction will be affected directly. In our future research, one consideration is to improve the newly designed appropriately-equal condition so that when the histogram shape is distorted seriously, the accuracy of watermark extraction is still acceptable.”
Author Response File: Author Response.pdf
Reviewer 3 Report
Overall the paper is sound, since it addresses an interesting and deeply felt problem, by employing well-known techniques. However, such a paper presents some serious shortcomings that need to be necessarily addressed.
First of all, I think the writing style/clarity needs more effort before this work will be ready to publish. Indeed, I found the paper extremely difficult to read, due not only to the poor grammar used throughout, but also the unclear structure of the argument being put across. In particular, the quality of the presentation significantly weakens this paper. Therefore, the language used in this paper needs to be run through a professional editor, in order make it easier to follow and comprehend. This paper would be substantially improved by thoroughly rewriting the prose with the help of a good English-language writer. In general, this paper needs such a treatment before being considered any further.
Furthermore, presentation aside, by reading the paper, it still was not entirely clear what to expect with the direction of the article. Indeed, the contribution proposed in this paper has been only marginally compared and contextualised with respect to the state of the art. As a result, it is extremely difficult to understand the real novelty introduced by the paper. The aforementioned aspects should be carefully addressed before the paper can be considered any further.
The algorithms and figures within the paper have been very poorly described. Therefore, more effort should be devoted to this aspect.
The mathematical notation used is extremely poor. As a result, the paper is extremely difficult to follow and understand. To address this problem, I would suggest adopting a LaTeX-Like style.
The set of references should be expanded by taking into account the following related works:
[a] https://doi.org/10.1145/2893474
[b] https://doi.org/10.1007/s11042-015-3065-8
As minor remarks, a thorough proofreading is strongly recommended, since the paper is affected by several typos and grammatical errors.
Author Response
Response to Reviewer 3 Comments
We would like to express our gratitude to you for your time and effort in reviewing and processing our manuscript and especially providing constructive comments and valuable suggestions for significantly improving the manuscript. Following these comments and suggestions, we have made changes in the revised manuscript. A point-by-point reply to the reviewer’s comments is given below, where in each case we quote the referee’s comments and then explain how we have revised the paper to accommodate the revisions requested.
Point 1: First of all, I think the writing style/clarity needs more effort before this work will be ready to publish. Indeed, I found the paper extremely difficult to read, due not only to the poor grammar used throughout, but also the unclear structure of the argument being put across. In particular, the quality of the presentation significantly weakens this paper. Therefore, the language used in this paper needs to be run through a professional editor, in order make it easier to follow and comprehend. This paper would be substantially improved by thoroughly rewriting the prose with the help of a good English-language writer. In general, this paper needs such a treatment before being considered any further.
Response 1: Thanks for your helpful and valuable comment. We have adjusted the presentation style in the revised version. In addition, we have rewritten the paper with the help of a good English-language writer. At last, the manuscript has been edited with LaTeX.
Point 2: Furthermore, presentation aside, by reading the paper, it still was not entirely clear what to expect with the direction of the article. Indeed, the contribution proposed in this paper has been only marginally compared and contextualised with respect to the state of the art. As a result, it is extremely difficult to understand the real novelty introduced by the paper. The aforementioned aspects should be carefully addressed before the paper can be considered any further.
Response 2: Many thanks for your excellent suggestion for improving our manuscript. We have added some explanations to the revised version, as shown below:
In the abstract:
“Histogram watermark, which performs watermark embedding by slightly modifying the histogram of original image, has been a hot research topic in information hiding technology due to the superiority that its pixel modification during the watermark embedding process is independent of the pixel position. This property makes histogram-based watermark strong resistance to geometric attacks, such as cropping attack, crossed attack, rotation attack, etc. In this paper, we propose a large capacity histogram-based robust watermarking algorithm based on three consecutive bins at the first time. In our scheme, we divide the shape of three consecutive bins into eight cases. According to these cases, we embed information number 0, 1, 2, 3, 4, 5, 6 or 7, respectively. The embedded information capacity reaches one bit per bin (bpb), and the amount of embedded information equals to 200% of the previous existing algorithms. Experimental results show that the new algorithm not only has large capacity of embedding information, but also has strong robustness to geometric attacks as well as common image processing operations.”
In the section Conclusions:
“In this paper, a large capacity histogram-based watermark algorithm is proposed. The watermark design of the proposed method has satisfactory functionalities in terms of robustness to geometric attacks and common image processing operations. In the embedding range selection, possible bad bins will suffice to be eliminated successfully with the assist of a predefined threshold. In the watermark embedding design, by taking the approximately-equal number of adjacent histogram bins into account, the proposed method makes every three consecutive bins carry 3-bits watermark; or rather, it realizes the maximum embedding rate (1bpb) at the first time. The experiment results demonstrate that the proposed watermark algorithm can provide a preferable tradeoff between embedding capacity and robustness, especially suitable for applications where robustness and embedding capacity are essential, such as copyright protection requirement for enough amount of watermark information and higher robustness. However, a possible limitation for the proposed histogram-based watermarking is the situation where certain operations may make the histogram shape distorted much. In this case, the watermark extraction will be affected directly. In our future research, one consideration is to improve the newly designed appropriately-equal condition so that when the histogram shape is distorted seriously, the accuracy of watermark extraction is still acceptable.”
Point 3: The algorithms and figures within the paper have been very poorly described. Therefore, more effort should be devoted to this aspect.
Response 3: Thank you for providing the advice for us. We have added some explanations to the revised version, as shown below:
In the section 3.3.1:
First, we transform the original watermark into its binary type. For various watermark data, pixel modification rules are slightly different. Note that having eight types of situations is mainly attributed into two factors: one is that the approximately-equal pixel number of adjacent bins is considered here. The other is that every three consecutive bins are partitioned into a group to perform watermark embedding. Unlike other histogram-based watermark methods [14-16], these methods do not consider this case in their watermark designs, thus only 1-bit watermark data can be embedded into each three consecutive bins. In the proposed methods, every 3-bits watermark information can be simultaneously embedded into each three consecutive bins. This is also why pixel modification rules for various watermark data are required.
For clarity, we take a simple example to explain the relationship between binary watermark information and its corresponding formula. Suppose , and denote the original three consecutive bins before watermark embedding, and , and denote the corresponding modified bins after watermark embedding, respectively. Let us define the proportional relationship of binary watermark information "0" and "1" as 1:2 to distinguish various types of watermark. For instance, given the watermark "000", one should modify the original bins , and using the formula, in the second line of Table 1, so that the pixel number of the modified bins , and closes to 1:1:1. In a similar modification/adjustment manner, the proportional relationship of the modified bins , and after watermark embedding will close to 1:1:2 if the given watermark is "001". Others are all similar to the above, except for a special case, i.e., the watermark "111". It can be distinguished by predefining the proportional relationship of the modified bins , and as 1:2:3 roughly. With the assist of this predefinition, one can accomplish the corresponding modification/adjustment before and after watermark embedding by making use of the formula, as shown in the last line of Table 1. Based on the above analysis, the extraction accuracy of embedded watermark can be ensured.
After that, we can embed 3-bits watermark into each group containing three consecutive bins. To explicitly analyze the embedding rules, an illustration of embedding 3-bits watermark in eight cases is also shown in Figure 3. Assume that there remains an original group containing three consecutive bins, as shown in Figure 3(a), let us analyze the case . Obviously, the original group in Figure 3 (a) does not satisfy the condition in Case 3, shown in the fourth line in Table 1, a part of pixels in both bins and , shown in Figure 3(a), will be modified to so that the modified result satisfies , and , here, represents a down integer function, and , as shown in Figure 3(d). For watermark embedding process of other cases, the modification process is similar to the above.
Table 1. Pixel modification rules for various watermark information (where, ).
Case | Original watermark | Its binary type | Pixel modification rules |
1 | 0 | 000 | |
2 | 1 | 001 | |
3 | 2 | 010 | |
4 | 3 | 011 | |
5 | 4 | 100 | |
6 | 5 | 101 | |
7 | 6 | 110 | |
8 | 7 | 111 |
Figure 3. Illustration of embedding 3-bits watermark in eight cases.
In Figure 8:
(a) | (b) |
(c) | (d) |
(e) | (f) |
(g) | (h) |
(i) | (j) |
Figure 8. Robustness to cropping, crossed and rotation for Baboon and Barbara (M=2, T=5 and C=24). (a) and (b) the cropped watermarked images of Baboon and Barbara; (c) and (d) the crossed watermarked images of Baboon and Barbara; (e) and (f) the rotation watermarked images of Baboon and Barbara; (g) and (h) the AWGN watermarked images of Baboon and Barbara; (i) and (j) the JPEG watermarked images of Baboon and Barbara.
Point 4: The mathematical notation used is extremely poor. As a result, the paper is extremely difficult to follow and understand. To address this problem, I would suggest adopting a LaTeX-Like style.
Response 4: Thank you very much for your valuable comment. The paper has been rewritten using the LaTeX editor.
Point 5: The set of references should be expanded by taking into account the following related works:
[a] https://doi.org/10.1145/2893474
[b] https://doi.org/10.1007/s11042-015-3065-8
Response 5: Many thanks for your excellent suggestion and helpful website for improving our manuscript. These papers are mentioned and discussed in the Introduction. Please refer to [1] and [3]:
“With the rapid growth of popular and low-cost access to image editing applications, some intentional or unintentional manipulations on digital media are becoming available during transmission [1,2].”
“Robust watermarking [3] is the technique that embeds information in the image to provide authentication, copyright protection, copy control, etc.”
Point 6: As minor remarks, a thorough proofreading is strongly recommended, since the paper is affected by several typos and grammatical errors.
Response 6: We are sorry for these careless mistakes. We have corrected them, as shown below.
In the abstract:
“According to these cases, we embed information number 0, 1, 2, 3, 4, 5, 6 or 7, respectively.”
In the second paragraph of the section Introduction:
“A variety of watermarking related methods have been proposed so far, and can be classified into three categories: …”
“Feature area embedding based methods were to extract feature points [11] or salient points such as the eyes and the mouth [12] firstly, and then embed watermark into them.”
In the fourth paragraph of the section Introduction:
“With their method, some possible bad bins can also be eliminated to some degree.”
“Obviously, this restricts many applications, especially limited for those with a strict requirement for watermark capacity.”
In the first paragraph of the section 2.1:
“… 1) modifying a portion of pixels to embed watermark is only related to the modification count of each grayscale, not to the position of pixels; …”
In the first paragraph of the section 2.2:
“… then they divided these bins into a series of groups where each of them contains two neighboring bins,…”
“…as shown in Figure 1.”
In the second paragraph of the section 2.2:
“This will have a direct influence on the watermark embedding capacity, in turn, the watermark capacity, lowered greatly, restricts many application scenarios.”
In the first paragraph of the section 3.2:
“…Hu et al. [14] initially proposed block-based method after learning human visual system (HVS) [19] and applying Xiang et al. idea.”
In the second paragraph of the section 3.3.1:
“Note that having eight types of situations is mainly attributed into two factors: one is that the approximately-equal pixel number of adjacent bins is considered here.”
“Unlike other histogram-based watermark methods [14-16], these methods do not consider this case in their watermark designs, thus only 1-bit watermark data can be embedded into each three consecutive bins.”
In the third paragraph of the section 3.3.1:
“Suppose a, b and c denote the original three consecutive bins before watermark embedding,…”
“Others are all similar to the above, except for a special case,…”
“…the default value (the case which does not satisfy certain proportional relationships is set as an approximately-equal treatment) is used to weaken the fragile relationship for the adjacent bins which are approximately-equal.”
In the Case 1 of the section 3.3.2:
“…other operations are as follows:”
“…the order in the above steps cannot be changed because these pixels chosen from Bin2 should not contain those that have been adjusted in Bin1.”
In the Step4 of the section 3.3.2:
“According to the above histogram modification/adjustment procedure to modify the histogram shape of each group, when adjusting the pixel value,…”
In the Step5 of the section 3.3.2:
“These above steps are repeated until the whole PN sequence is embedded. The final modified image is the watermarked image.”
In the section 3.3.3:
“When the watermarked images are illegally manipulated during transmission,…”
“In a similar way, the proposed method can extract all of the other watermark information.”
“Similarly, we also give a partial version of pseudo-code of the watermark extraction process, as shown in Algorithm 2.”
In the section 4.1:
“In Xiang et al. method [13], two consecutive bins are combined into a group and the bin width is set as M=2, thus each four bins carry 1-bit watermark.”
In the section 4.1.1:
“…Structural Similarity Index Measure (SSIM) are used for judging distortion degree between the original and watermarked images.”
“This is because there are much more pixels needed to be modified to embed more payload bits.”
In the section 4.1.2:
“It can be observed that our method provides a preferable visual quality of the watermarked image, meanwhile, it ensures more information can be embedded into the pixels located at the same bins. Our algorithm, with respect to the existing ones, improves the embedding capacity by at least 100%.”
In the section 4.2:
“For the four test images, we set the experimental parameter as the following: 1) the bin’s width is set as 2; 2) the threshold is set as 5; 3) the capacity is set as 24 bits.”
In the last paragraph of the section 4.2:
“Because the histogram of modified image incurred during the cropping and crossed attacks keeps approximately the same as that of the original image without suffering from attack, the watermark algorithm has higher robustness to cropping and crossed attacks.”
“Finally, the embedding of watermark information mainly depends on the statistical property of the histogram and is independent of the size of an image, thus this proposed watermark method can be applied to images with any size.”
To sum up, after checking the total manuscript, we have added and explained all of the possible poor expressions/presentations thoroughly in the revised version. Please refer to the revised version for detailed contents.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Dear Authors,
Thank you for providing the revised paper.
Kind regards,
- Reviewer -
Reviewer 3 Report
p.p1 {margin: 0.0px 0.0px 0.0px 0.0px; font: 15.0px Arial}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. In my opinion, the paper is now ready to be accepted for publication.
