Unsupervised Wavelet-Feature Correlation Ratio Markov Clustering Algorithm for Remotely Sensed Images
Round 1
Reviewer 1 Report (Previous Reviewer 2)
Comments and Suggestions for AuthorsThe paper, titled "Unsupervised Wavelet-Feature Correlation Ratio Markov Clustering Algorithm for Remotely Sensed Images," presents a novel clustering algorithm designed for remotely sensed data, emphasizing its application to multispectral and hyperspectral images. This algorithm, named Wavelet-feature Correlation Ratio Markov Clustering Algorithm (WFCRMCA), leverages the spectral characteristics of objects under different conditions to evaluate the differences among remotely sensed pixels. By applying wavelet transform to spectral data and using the correlation ratio as a statistical tool, the algorithm aims to capture and identify peak points in the data, allowing for efficient and accurate clustering. The paper evaluates the algorithm's performance using Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Thermal Mapping (TM) data, showcasing its high convergence velocity and acceptable clustering accuracy.
The WFCRMCA introduces a novel approach to clustering remotely sensed images, distinguishing itself from traditional methods like K-means or ISO-DATA through its use of wavelet transforms and correlation ratios. The algorithm's development is grounded in a thorough understanding of spectral data characteristics, and its design is supported by theoretical underpinnings and mathematical models. The empirical results demonstrate the algorithm's effectiveness, particularly in terms of clustering accuracy and speed, using real-world data sets. This work could significantly impact remote sensing image analysis, offering a more efficient and potentially more accurate clustering method.
Some other points should be considered as well:
The algorithm is complex, and the paper might be challenging for readers not well-versed in wavelet theory. The algorithm's performance appears to be highly dependent on the correct tuning of its parameters, which may limit its practicality in diverse scenarios.
While the algorithm is shown to be effective, the computational cost, especially for large datasets, is not thoroughly discussed. Finally the paper could benefit from a more extensive comparison with existing clustering algorithms to contextualize its performance better.
Major Revisions:
1. Include a more comprehensive comparison with established clustering algorithms, demonstrating the proposed method's advantages and limitations in various scenarios.
2. Provide a detailed discussion on the computational complexity of the algorithm, especially in handling large-scale datasets.
3. Conduct an in-depth analysis of the algorithm's sensitivity to its parameters and provide guidelines for parameter selection in different applications.
4. Discuss the practical implementation of the algorithm, including potential challenges and strategies for overcoming them in real-world applications.
Minor Revisions:
1. Simplify the technical language where possible to make the paper more accessible to a broader audience.
2. Include more graphical illustrations to help visualize the algorithm's process and its effects on clustering.
3. Clarify the abstract and conclusion to more succinctly summarize the main contributions and findings of the paper.
The English can be improved yet is reasonably well.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report (New Reviewer)
Comments and Suggestions for Authors1. The article introduces a novel approach, the Wavelet-Feature Correlation Ratio Markov Clustering Algorithm (WFCRMCA), to cluster remotely sensed images. This approach is based on wavelet transforms and correlation ratios, offering a fresh perspective on image analysis. It employs band-pass wavelet filters to detect critical points, including extreme points and zero crossings, allowing for the identification of abrupt spectral features. This focused approach to feature detection can aid in accurate clustering.
2. The method claims to provide efficient clustering accuracy by using correlation ratio-based criteria instead of traditional methods like K-means clustering, which can be limited in their accuracy or convergence speed. Clustering Speed. By avoiding the computation of Euclidean distances and utilizing simulated annealing Markov clustering, the WFCRMCA claims to accelerate the clustering speed, offering potential efficiency gains in the analysis process.
Though, there are major concerns,
3. Despite the methodology's potential, there are parameters that require manual adjustment (e.g., Tpeak and Scale2) for specific applications, which might impact its ease of use and applicability in different scenarios.
4. The Markov clustering method used in this approach is not easily parallelizable. While it can offer an optimal solution, it might be limiting in terms of scalability and processing speed for larger datasets or in parallel computing environments.
5. The approach focuses primarily on certain specific features derived from wavelet analysis. This might limit its ability to handle diverse or complex datasets where other features are equally important for accurate classification.
6. Although the method is intended for remotely sensed images, its application to other fields or types of data is not extensively discussed. The applicability beyond the intended scope might need further exploration.
7. The methodology's complexity might require a deeper understanding of wavelet analysis and specific feature extraction, potentially limiting its adoption without specialized knowledge.
8. The process of adjusting thresholds and scales to achieve optimal clustering results might be complex and time-consuming, impacting its practical implementation.
In essence, while the proposed WFCRMCA demonstrates promise in improving clustering accuracy and speed for remotely sensed images, its manual parameter adjustments and limited applicability might require further development and exploration for broader and more automated usage in the field of image analysis.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report (New Reviewer)
Comments and Suggestions for AuthorsThis paper presents a Wavelet-feature correlation ratio Markov clustering algorithm for remotely sensed images.
The studied topic is meaningful and the proposed method seems reasonable.
The authors may further improve the paper from the following aspects.
As discussed in some surveys and studies, e.g., Perceptual image quality assessment: a survey; Screen content quality assessment: overview, benchmark, and beyond; the quality of the input data like images and videos is an important aspect of various intelligent systems, including clustering and segmentation systems.
High-quality images are important for the successful usage of these intelligent systems, while low-quality media may degrade the performance of these systems.
The authors may give some discussions on this aspect as well as the above-mentioned works.
The authors only conduct experiment in high-quality images.
As discussed in Blind quality assessment based on pseudo-reference image, Blind image quality estimation via distortion aggravation, artifacts existing in images of various contents are more difficult to detect or evaluate. The authors are suggested to give some discussions on these aspects and the above works, and give some discussions on whether it is possible to conduct evaluation when the quality of the input image is not good.
The authors evaluate the proposed method on multi-spectral images which are typical multimodal data. As described in some studies, e.g., A multimodal saliency model for videos with high audio-visual correspondence; Study of subjective and objective quality assessment of audio-visual signals, integrating information of multiple modalities could further improve various intelligent systems. The authors are suggested to give some discussions on these aspects and the above works.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report (New Reviewer)
Comments and Suggestions for AuthorsPlease refer to the attached file.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report (New Reviewer)
Comments and Suggestions for Authors- The paper demonstrates a strong level of scientific soundness and originality in using WFCRMCA for clustering analysis in remote sensing data and showcases promising results. The presentation of the methodology and results is comprehensive, providing a detailed understanding of the WFCRMCA algorithm's application and performance. Providing more contextual explanations, especially regarding the wavelet transformation, would help readers in understanding the methodology better. Also, perhaps some of the older references could be replaced with new ones.
Some minor edits.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report (New Reviewer)
Comments and Suggestions for AuthorsI am satisfied with the revised manuscript.
Reviewer 3 Report (New Reviewer)
Comments and Suggestions for AuthorsNone
Reviewer 4 Report (New Reviewer)
Comments and Suggestions for AuthorsNA
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn this manuscript, the author proposed a new algorithm, i.e., wavelet-feature correlation ratio Markov clustering algorithm. The results in the manuscript show some advantages of this algorithm. The method is novel and the manuscript is well organized. I think this work can be accepted after some improvements.
My comments are listed below:
1. The appearance of Fig2 can be improved.
2. In the Conclusions, the author claimed that "WFCRMCA accelerates clustering speed...", can the author provide some detatiled data or materials about how or how much the WFCRMCA accelerates clustering speed?
3. I noticed that this manuscript just has only one author, so I have a little puzzle that is why the author used "we" not "I" in the manuscript.
Reviewer 2 Report
Comments and Suggestions for Authors
The work is interesting and the authors worked properly on the topic.
An algorithm called Wavelet-feature correlation ratio Markov clustering algorithm (WFCRMCA) is presented in this article. Regardless of the conditions under which they observe it, the spectrums of a particular object exhibit consistent features, including upward, downward, protruding, and concave patterns.
This method optimizes the Markov clustering process in the wavelet feature space to evaluate the differences among remotely sensed pixels. Wavelet transforms are used to capture and identify peak points in the spectral data. An adjustable spectrum domain or wavelet scale is used to calculate the correlation ratio between two samples.
At each step of the Markov clustering process, the authors create class centers based on the correlation ratio threshold. Unlike K-means and ISO-DATA, this approach does not compute Euclidean distance, which slows down clustering. The authors use strategies like simulated annealing and gradually shrinking cluster sizes to control clustering convergence. As a result, they can quickly determine the best clustering. By using data from the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and Thermal Mapping (TM), the authors validate the WFCRMCA method.
Results demonstrate high convergence velocity and acceptable clustering accuracy for remote sensing data analysis.
Anyway there is necessity to operate in this sense to improve the general work:
1. The authors need to consider explicitly all the computational phases of the algorithms
they propose, create some specialized draws to show them in practices and more importantly
aiming replicability should disclose the computational procedure they follows, the language
used, the libraries and so on.
2. The discussion is too short. The authors should discuss extensively their results
considering also possibilities to improvement in the use of this approach at a more
general level
3. possible real applications of this approach are not presented. In the discussion
the authors should discuss also possible applications
4. in the conclusions authors should express clearly what strong and weak points of the
approach proposed. At the same time there is the need to discuss possible extensions of the
approach and future directions.
5. A comparison with existing methods and the improvements should be emphasized.
6. No discussion on the data-preprocessing: the authors should discuss about possible
data preprocessing approaches.
7. References should improved throughout the text.
Comments on the Quality of English LanguageThe English level should be improved.
Reviewer 3 Report
Comments and Suggestions for AuthorsPlease find my comments in the uploaded pdf.
Comments for author File: Comments.pdf
Please find my comments in the uploaded pdf.
