A Light Field Depth Estimation Algorithm Considering Blur Features and Prior Knowledge of Planar Geometric Structures

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes early and accurate depth detection in light field cameras by merging blur features and prior knowledge about planar geometric structures. The algorithm optimizes matching costs in low-texture regions, eliminates mismatched points using epipolar geometry, and uses an adaptive process to improve accuracy. The experiments demonstrate superior performance over existing methods regarding reconstruction, edge clarity, and scene coverage.

The study presented by the authors is interesting. However, several points deserve to be taken into account:

·        Authors should list their contributions as separate bullet points at the end of the introduction. This approach provides readers with an instant understanding of the study’s contributions. In addition, integrating a structured presentation of the document from the introduction would allow readers to easily navigate the different sections and understand the study's overall progress.

·        Authors should separate the literature review from the introduction to clarify the study's positioning. A concise introduction should focus on the general background and objectives of the research. At the same time, a dedicated state-of-the-art section would provide an in-depth overview of existing work, thereby better highlighting the proposed approach's originality and relevance.

·        The discussion of experimental results should be separated from the raw presentation of the data. Such a distinction would enable better identification of the difficulties encountered and proposed improvements. This separation would also make the analysis more accessible and enhance understanding of the scientific and practical implications of the results.

·        The current conclusion is dense and lacks structure, which makes it difficult to read. It could be organized into three distinct sections:

o   A summary of the main contributions to recall the significant contributions.

o   A summary of the results obtained to highlight the strong points of the study.

 

o   A clear presentation of prospects for improvement to guide future research.

Author Response

Comments 1：This paper proposes early and accurate depth detection in light field cameras by merging blur features and prior knowledge about planar geometric structures. The algorithm optimizes matching costs in low-texture regions, eliminates mismatched points using epipolar geometry, and uses an adaptive process to improve accuracy. The experiments demonstrate superior performance over existing methods regarding reconstruction, edge clarity, and scene coverage.

 

Response 1：We sincerely thank you for your positive acknowledgment of our work. Your recognition of our innovative approach in merging blur features and prior knowledge , is highly encouraging and validates our efforts. Your support is greatly appreciated.

 

Comments 2： Authors should list their contributions as separate bullet points at the end of the introduction. This approach provides readers with an instant understanding of the study’s contributions. In addition, integrating a structured presentation of the document from the introduction would allow readers to easily navigate the different sections and understand the study's overall progress.

 

Response 2：Thank you very much for your thorough review of our paper and for your valuable suggestions. We fully appreciate and agree with your recommendation to separately list the contributions at the end of the introduction and to integrate a structured overview of the document from the introduction. This will not only help readers immediately understand our research findings but also allow them to easily navigate through different sections and grasp the overall progress of the study.

In response to your suggestions, we make the following revisions to the paper:

Separately List Contributions at the End of the Introduction: We will clearly and concisely list the main contributions and innovations of our study in the final paragraph of the introduction. This will highlight our research outcomes and make it easier for readers to quickly capture the core content of the paper, Please see the details in Lines 40-52.

Integrate a Structured Overview of the Document: In the introduction, we will provide a more detailed overview of the paper's overall structure and the main content of each section. By doing so, readers will have a general understanding of the paper's framework when reading the introduction, making it easier for them to navigate to the sections they are interested in for in-depth reading , Please see the details in Lines 53-58.

Once again, we extend our heartfelt gratitude for your constructive feedback and insightful comments. They have greatly improved the quality and readability of our paper.

 

Comments 3： Authors should separate the literature review from the introduction to clarify the study's positioning. A concise introduction should focus on the general background and objectives of the research. At the same time, a dedicated state-of-the-art section would provide an in-depth overview of existing work, thereby better highlighting the proposed approach's originality and relevance.

 

Response 3：Thank you very much for your valuable comments, which have been instrumental in refining the structure of our paper. In accordance with your guidance, we have made significant adjustments by clearly separating the Introduction from the Related Work to better define the research orientation and context.

In the Introduction section, we have focused on the general background and purpose of the study, striving for brevity and clarity in presenting the rationale and significance of our work. We have ensured that this part captures the reader's interest and lays a solid theoretical foundation for the subsequent content.

Meanwhile, in the Related Work section, we have provided an in-depth overview of the latest developments in the relevant field, analyzing in detail the existing research findings, methodologies, and challenges. Through this section, we aim to highlight the uniqueness and relevance of our proposed approach in comparison to the existing work. We hope to demonstrate the novelty of our method and its advantages in addressing practical problems.

Thank you again.

 

Comments 4：The discussion of experimental results should be separated from the raw presentation of the data. Such a distinction would enable better identification of the difficulties encountered and proposed improvements. This separation would also make the analysis more accessible and enhance understanding of the scientific and practical implications of the results.

 

Response 4：Thank you very much for your valuable feedback. You have pointed out the importance of separating the discussion of experimental results from the raw presentation of data to better identify encountered difficulties and propose improvements. This distinction is indeed crucial for enhancing the readability and understanding of the scientific and practical implications of the results.

While preparing the manuscript, I have taken care to include a detailed discussion of the experimental results in Section 5, titled "Discussion." , Please see the details in Lines 587-596.In this section, I have summarized the key findings from the experimental data. My aim was to provide readers with a clear and coherent analysis framework that facilitates their understanding of the scientific significance and practical applications of the results.

Once again, thank you for your insightful guidance. I look forward to your further feedback and will make necessary revisions based on your suggestions.

Comments 5：The current conclusion is dense and lacks structure, which makes it difficult to read. It could be organized into three distinct sections:

• A summary of the main contributions to recall the significant contributions.
• A summary of the results obtained to highlight the strong points of the study.
• A clear presentation of prospects for improvement to guide future research.

 

Response 5：Thank you very much for your thorough review and valuable suggestions on our paper. We greatly appreciate your input and have carefully considered your recommendations.

Based on your guidance, we have revised the conclusion section to make it more readable and structured, Please see the details in Lines 589-606.. Specifically, we have:

• Summarized the main contributions of our research, highlighting the key findings and their significance.
• Provided a detailed summary of the experimental results, emphasizing the strengths and practical implications of our study.
• Outlined future research directions and potential improvements, based on the limitations of our current work.

We believe that these changes will enhance the clarity and coherence of the conclusion section, making it easier for readers to understand and evaluate our research.

Thank you again.

Reviewer 2 Report

Comments and Suggestions for Authors

The article introduces a novel approach to light field depth estimation, combining blur features and planar geometric structure priors. This integration addresses critical limitations of existing methods, such as neighborhood selection and mismatches in weak texture regions. The algorithm's emphasis on iterative propagation, dynamic pixel matching, and incorporation of epipolar geometry relationships represents a significant advancement over prior methodologies. The experimental results further demonstrate the effectiveness of the method compared to state-of-the-art algorithms, which strengthens its claim of originality.

However, the novelty score is slightly reduced because while the combination of techniques is innovative, the components themselves (e.g., blur features and planar priors) are built on well-established concepts in the field. A deeper exploration of how this work diverges fundamentally from similar existing methods would further enhance its contribution.

Comments on the Quality of English Language

The language is clear and technical, making it suitable for a scientific audience. Key concepts are explained systematically, and the article provides sufficient details for replication and understanding. However, there are areas for improvement:

• Grammar and Syntax: There are frequent minor grammatical errors, such as "we proposes" instead of "we propose." These errors can distract readers and impact the professionalism of the article.
• Redundancy: Some sections, particularly in the introduction and methodology, are repetitive, which dilutes the focus of the narrative.
• Readability: The text is dense, making it challenging for readers who are not deeply familiar with light field technology to follow. Simplifying some explanations without losing technical depth would improve accessibility.

Author Response

Comments 1：The article introduces a novel approach to light field depth estimation, combining blur features and planar geometric structure priors. This integration addresses critical limitations of existing methods, such as neighborhood selection and mismatches in weak texture regions. The algorithm's emphasis on iterative propagation, dynamic pixel matching, and incorporation of epipolar geometry relationships represents a significant advancement over prior methodologies. The experimental results further demonstrate the effectiveness of the method compared to state-of-the-art algorithms, which strengthens its claim of originality.

However, the novelty score is slightly reduced because while the combination of techniques is innovative, the components themselves (e.g., blur features and planar priors) are built on well-established concepts in the field. A deeper exploration of how this work diverges fundamentally from similar existing methods would further enhance its contribution.

The language is clear and technical, making it suitable for a scientific audience. Key concepts are explained systematically, and the article provides sufficient details for replication and understanding.

 

Response 1：Thank you very much for your meticulous review and valuable feedback on our work.  We greatly appreciate your recognition of the novelty in combining blur features and planar geometric structure priors for light field depth estimation, as well as your concern regarding the reliance on well-established concepts within the field for the method's components.  Below is our detailed response to your comments:

Deepening the Innovation: You pointed out that while the combination of techniques is innovative, the individual components (such as blur features and planar priors) are built upon established concepts in the field. We fully concur and wish to clarify that our innovation lies in the novel integration of these classic concepts to overcome limitations of existing methods in neighborhood selection and mismatch handling in weak texture regions.  Specifically, we introduced iterative propagation mechanisms, dynamic pixel matching strategies, and incorporated epipolar geometry relationships, which represent substantial extensions and integrations of classic concepts aimed at achieving more accurate and robust depth estimation.

In summary, we highly value your feedback and plan to fully incorporate your suggestions in the revised manuscript to further enhance the innovation and contribution of our work. We believe that with these improvements, our work will better showcase its unique value and practical significance in the field of light field depth estimation.

Thank you once again for your review and valuable comments. We look forward to your further guidance.

 

Comments 2： There are frequent minor grammatical errors, such as "we proposes" instead of "we propose." These errors can distract readers and impact the professionalism of the article.

 

Response 2：Thank you for pointing out the grammatical and syntactic errors. We acknowledge that these errors can distract readers and undermine the professionalism of our papers. We carefully reviewed the manuscript and corrected all similar errors in the corresponding sections of the article separately to ensure the accuracy and consistency of the language, Please see the details in Lines 10,41,47,291,357,382,599.

Thank you again.

 

Comments 3： Some sections, particularly in the introduction and methodology, are repetitive, which dilutes the focus of the narrative.

 

Response 3：

We appreciate your interest in the introduction and methods sections. In response to your concern about repetitiveness, I have revised the introduction by streamlining the content, providing a concise overview, and enhancing the section with a clear statement of our contributions, Please see the details in Lines 35~58. This adjustment aims to sharpen the focus of the narrative. 

Once again, I greatly appreciate your insightful comments.

 

Comments 4： The text is dense, making it challenging for readers who are not deeply familiar with light field technology to follow. Simplifying some explanations without losing technical depth would improve accessibility.

 

Response 4： We understand that dense text can pose challenges for readers unfamiliar with light field technology. To enhance accessibility without compromising technical depth, we have provided simplified descriptions and explanations of light field-related technologies in sections 3.1.1, 3.1.3, 3.2.1, and 3.2.2, Please see the details in lines 157~160, lines 227~241, lines 284~292, lines 316~335. We hope these changes will make the newspaper more approachable to a wider audience.

Thank you once again for your review and valuable comments. We look forward to your further guidance.

 

Special thanks to you for your good comments.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper presents a novel light field depth estimation algorithm that combines blur features with prior knowledge of planar geometric structures. The proposed approach is innovative and well-structured, offering a clear contribution to the field.

- The paper relies on a specific dataset, which may not adequately capture the diversity of real-world scenes. Expanding the dataset would strengthen the findings.

- The algorithm's computational demands are not explicitly analyzed. A discussion on time complexity or efficiency would be valuable.

- While the results are promising, the theoretical basis for combining blur features and planar geometric structures could be elaborated further.

- he choice of evaluation metrics is not adequately justified. Additional metrics, such as computational speed or robustness to noise, could provide a more comprehensive evaluation.

- Add more visual examples to demonstrate the algorithm’s effectiveness in diverse scenarios, especially challenging ones like low-light or high-reflectance environments.

- Conduct ablation studies to quantify the contribution of blur features and planar geometric priors individually.

Author Response

Comments 1： The paper presents a novel light field depth estimation algorithm that combines blur features with prior knowledge of planar geometric structures. The proposed approach is innovative and well-structured, offering a clear contribution to the field.

 

Response 1：Thank you very much for your valuable recognition of our light field depth estimation algorithm. We appreciate your comments on its innovation and well-structured presentation. We will carefully consider your feedback and make corresponding improvements to further enhance our work.

 

Comments 2： The paper relies on a specific dataset, which may not adequately capture the diversity of real-world scenes. Expanding the dataset would strengthen the findings.

 

Response 2：We are very grateful for your valuable suggestions and the opportunity to revise our paper, which has allowed us to further improve its quality. We acknowledge that expanding the dataset to include more diverse scenarios could further strengthen our research results.

However,regarding the dataset used in our paper, we believe that the dataset is representative of the scenarios commonly encountered within the range of the focused light field cameras currently available on the market, which typically have an operating distance of 0-20 meters.

To further clarify, the datasets mentioned in our paper are carefully selected to represent a wide range of real-world scenarios. Specifically:

• The HS260 dataset covers a scene range of 10-23 meters and includes a variety of textures, such as calibration boards, tripods, electrical distribution boxes, road signs, car license plates, and car logos. It also includes weak texture scenarios such as playground surfaces, walls, leaves, railings, and car windows.
• The R12 dataset covers a scene range of 0.4-1.5 meters and includes a rich variety of textures, such as calibration boards, shoe boxes, textured cardboard boxes, packaging boxes, umbrellas, and basketballs. It also includes weak texture scenarios such as desktops, reflective drink bottles, and fire extinguishers.
• The R32 dataset covers a scene range of 2-4 meters and includes scenarios with a variety of textures, such as multiple types of dolls and calibration boards. It also includes a range of other scenarios, such as radios, postcards, and perforated boards.

We are once again grateful for the valuable suggestions and comments put forward by the reviewer. Your insights have not only helped us identify potential areas for improvement in our research, but have also inspired us to further contemplate ways to enhance the quality and depth of our paper. We firmly believe that by continually embracing and incorporating feedback from professional reviewers like you, our research work can become more rigorous and comprehensive.

 

Comments 3：The algorithm's computational demands are not explicitly analyzed. A discussion on time complexity or efficiency would be valuable.

 

Response 3：Thank you for your valuable feedback on our paper. We appreciate your suggestion to include a discussion on the algorithm's time complexity or efficiency.

In response to your concern, we have included a quantitative experiment on efficiency in Section 4.6 of our paper. This section presents the results of our experiments, which were designed to evaluate the computational demands of our algorithm Please see the details in lines 569~586.

In the experiment, we measured the time required for our algorithm to process various input sizes and compared it to other relevant algorithms in the field. The results demonstrate that our algorithm is efficient and performs well .Furthermore, we have included a detailed analysis of the results in Section 4.6, discussing the factors that contribute to the algorithm's efficiency .

We believe that the inclusion of this quantitative experiment and corresponding result analysis addresses your concern and enhances the paper's overall quality.The efficiency experiment for depth estimation using light field cameras aims to evaluate the performance and efficiency of light field cameras in performing depth estimation for three-dimensional scenes.

Thank you once again for your review and valuable comments. We look forward to your further guidance.

 

Comments 4：While the results are promising, the theoretical basis for combining blur features and planar geometric structures could be elaborated further.

 

Response 4：Thank you very much for your thorough review of our paper and for your valuable feedback. We fully concur with your suggestion that the theoretical foundation of Blur Features Transfer and Plane Prior Optimization could benefit from further elaboration to make it accessible to readers unfamiliar with light field technology.

In response to your comment, we have made significant improvements in Sections 3.2.1 and 3.2.2 of our paper, Please see the details in lines 284~292,lines316~335. Specifically, in Section 3.2.1 on "Blur Features Transfer," we have delved deeper into the role and significance of blur features in light field image processing . We have added supplementary explanations to enhance the exposition of the theoretical foundation.

Similarly, in Section 3.2.2 on "Plane Prior Optimization," we have further clarified the principles and steps of the plane prior optimization. We have explained in detail how planar geometric structure information is utilized to optimize the image reconstruction process and how this method improves image quality and processing efficiency.

Through these improvements, we believe that Sections 3.2.1 and 3.2.2 of the paper now convey the theoretical foundation of Blur Features Transfer and Plane Prior Optimization more clearly and make it easier for readers unfamiliar with light field technology to comprehend the content.

Once again, I greatly appreciate your insightful comments.

 

 

Comments 5：he choice of evaluation metrics is not adequately justified. Additional metrics, such as computational speed or robustness to noise, could provide a more comprehensive evaluation.

 

Response 5：Thank you for your constructive feedback on our paper, particularly your suggestion regarding the evaluation metrics used. We appreciate your highlighting the importance of a comprehensive evaluation, which includes not only accuracy but also computational speed .

As mentioned in our response to Question 2, we have indeed incorporated computation time metrics into our efficiency experiments, Please see the details in lines 569~587. Specifically, we have measured and reported the computation time for various algorithms across different datasets to provide a quantitative assessment of their performance in terms of speed. This addition aims to offer a more holistic view of the algorithms' efficiency.

Thank you again.

 

Comments 6：Add more visual examples to demonstrate the algorithm’s effectiveness in diverse scenarios, especially challenging ones like low-light or high-reflectance environments.

 

Response 6：Thank you very much for your valuable suggestions and insights. We fully agree with your perspective that visual examples are crucial for demonstrating the effectiveness of algorithms in light field camera depth estimation, as they provide intuitive representations of the algorithm's performance.

We appreciate your emphasis on the importance of including more visual examples, especially in challenging scenarios such as low-light or high-reflectance environments.     These examples would indeed serve to highlight the precision of our algorithm in corresponding regions.

Due to current time constraints, we are unable to incorporate additional visual examples into the manuscript immediately. However, please be assured that we are committed to addressing this point in future revisions. We plan to add a comprehensive set of visual examples that encompass a wide range of challenging environments, including low-light and high-reflectance scenarios, to further demonstrate the accuracy and robustness of our algorithm.

Thank you again.

 

Comments 7：Conduct ablation studies to quantify the contribution of blur features and planar geometric priors individually.

 

Response 7：Thank you for your detailed feedback and constructive suggestions, which have been invaluable in guiding the refinement of our work.

In response to your request for ablation studies to quantify the individual contributions of blur features and planar geometric priors, we are pleased to inform you that we have conducted such experiments and incorporated the results into our manuscript.

Specifically, in Section 4.3, we have added an ablation study focused on the introduction of blur features, Please see the details in lines 518~530. This study compares the performance of our method with and without the inclusion of blur features, using three quantitative metrics to evaluate changes in accuracy. The results, presented in a table, demonstrate a significant improvement in performance when blur features are incorporated, underscoring their critical role in enhancing the algorithm's ability to handle complex scenes.

Similarly, in Section 4.4, we have conducted an ablation study on the impact of planar geometric priors, Please see the details in lines 531~549. This study evaluates the performance of our method in scenarios where planar geometric priors are either included or excluded. Again, using the same three quantitative metrics, we have documented the changes in performance. The results clearly show that the inclusion of planar geometric priors leads to improved accuracy and stability.

In both sections, we have provided a thorough analysis of the experimental results, discussing the observed trends and possible explanations for the improvements observed when each feature is included.These analyses further support the effectiveness of our proposed method and highlight the significant contributions of both blur features and planar geometric priors.

We are grateful for your guidance in this area and hope that you find the revised manuscript to be more comprehensive and persuasive.

 

Special thanks to you for your good comments.

Reviewer 4 Report

Comments and Suggestions for Authors

Dear Authors,

Interesting article. I have a few questions of a problematic nature, to which I did not find answers in the text, and it seems that it would be good to mention them:

- how was the calibration process carried out?

- was there a problem with geometric distortion of the lenses - if so? how was it solved?

- at what distance range were the measurements taken?

- how will the authors comment on the obtained RMSE - does the method work well or badly and for what applications can it be used

Best regards.

Author Response

Comments 1：Interesting article. I have a few questions of a problematic nature, to which I did not find answers in the text.

 

Response 1：Thank you very much for acknowledging the interest in our article and for raising a few challenging questions that we did not address in the text.  Your questions reflect your thoughtful consideration of our research and provide us with an opportunity to further refine and enhance the quality of our paper. We highly value your feedback and look forward to addressing these questions to make our research more rigorous and comprehensive. 

Once again, thank you for your valuable suggestions and comments!

 

Comments 2：how was the calibration process carried out?

 

Response 2：Thank you so much for your careful review and comments.  Please allow me to make the following detailed explanation and description of the issue.

The calibration method employed in this paper is proposed in the paper titled "Leveraging Blur Information for Plenoptic Camera Calibration" (2022, CVPR), which can be accessed at the URL: https://link.springer.com/article/10.1007/s11263-022-01582-z，and the open source code for the calibration method and related data sets are available at https://github.com/comsee-research/compote.

This method takes into comprehensive consideration various types of micro-images.     Initially, defocus blur is incorporated into the camera model, introducing the Blur Aware Plenoptic feature (BAP). A novel pre-calibration step is introduced, leveraging BAP features derived from white images to provide a robust initial estimate of internal parameters. This is achieved by varying the main lens aperture and utilizing different micro-lens focal lengths, combined with parameters from the image space, to estimate the blur radius using raw white images captured with a light diffuser mounted on the main objective and taken at different apertures. The blur radii are then correlated with camera parameters, enabling their initialization.Subsequently, a set of chessboard calibration board images, captured at various distances and orientations using a light field camera, are employed. The points on the calibration board possess known physical dimensions and relative positions.Image matching techniques are utilized to extract corner features from the calibration images.It propose a novel reprojection error function that exploits BAP features to refine a more comprehensive model, particularly including the multiple micro-lens focal lengths. This checkerboard-based calibration is conducted in a unified optimization process. By using the extracted corner coordinates and the camera model, internal camera parameters such as focal length, principal point position (i.e., image center), micro-lens type, pixel size, and distortion parameters (e.g., radial and tangential distortions) are computed. The external parameters of the camera include its position and orientation. Finally, by minimizing the reprojection error—the discrepancy between the actual points in the image and the theoretical points computed through the calibration parameters—the external parameters of the camera are optimized, resulting in more accurate transformations between pixel coordinates in the image and three-dimensional coordinates in the real world.

Thank you again.

 

Comments 3：was there a problem with geometric distortion of the lenses - if so? how was it solved?

 

Response 3：Thank you for your inquiry regarding geometric distortions in the lens and how they were handled in our study. We appreciate your attention to this important aspect of our work.

Camera lens distortions can introduce straight-line deformation and shape distortion in images. Therefore, we consider the distortion of the main lens in the camera model. To correct these distortions, we utilized the Brown-Conrady model, which accounts forradial-tangential distortion. This model enabled us to accurately estimate the distortion parameters, which were then used to perform geometric transformations on the images to eliminate the distortions introduced by the lens. After distortion correction, straight lines in the images remained straight, and shapes were preserved as intended.

To validate the calibration results, we projected known three-dimensional points into the images and compared them to the calibration outcomes. By evaluating metrics such as reprojection error and distortion correction effectiveness, we were able to assess the accuracy and reliability of our calibration.

Prior to depth estimation, we performed this calibration process and applied the obtained distortion parameters to remove distortions from the light field raw images captured under the same conditions (focus distance, aperture). This step ensured that the images used for depth estimation were free from lens distortions, allowing for more accurate and reliable depth maps to be generated.

In conclusion, we have addressed the geometric distortions in the lens by performing a  calibration process using the Brown-Conrady model and applying the obtained distortion parameters to correct the images.

Thank you again.

 

Comments 4：at what distance range were the measurements taken?

 

Response 4：Thanks for your attention to the range of measured distances in the experimental section, the three light field image datasets included in the experimental section of this article are measured in a specific range of distances. For HS260 data set, the shooting distance is 10 ~ 23 meters; For the R32 dataset, it is between 2 and 4 meters; For the R12 dataset, the range is 0.4 to 1.5 meters. Other parameters of the experimental configuration are shown in Section 4.1, Please see the details in lines 412~439.We believe these range of distances represent typical use cases for our approach, and we specified them in the experimental section for completeness.

Thank you again.

 

Comments 5：how will the authors comment on the obtained RMSE - does the method work well or badly and for what applications can it be used

解释rmse

Response 5：Thank you for your attention to the RMSE metric in our work. Below is a detailed response to your questions regarding the effectiveness of the RMSE metric and its applications.

Regarding the Effectiveness of the RMSE Metric:

In our study, the RMSE metric has been used as a key indicator to measure the performance of light field image depth estimation algorithm. A lower RMSE value indicates a smaller discrepancy between the predicted depth or disparity and the actual values, thereby demonstrating the high accuracy and robustness of the algorithm. According to our experimental results, our algorithm has achieved lower RMSE values on multiple datasets, which suggests that our method performs well in depth estimation.

Regarding the Applications of the RMSE Metric:

The RMSE metric has a wide range of applications in multiple fields. In light field image depth estimation, a lower RMSE value means that the algorithm can more accurately estimate scene depth, which is crucial for 3D reconstruction and scene understanding in fields such as augmented reality, virtual reality, and autonomous driving. Furthermore, in traditional multi-view stereo matching, RMSE can be used to evaluate the accuracy of the disparity map obtained by the matching algorithm, which is important for applications such as 3D reconstruction, object recognition, and tracking. Additionally, RMSE can be applied to other computer vision tasks, such as image segmentation and object detection, to measure the difference between the algorithm's prediction and the actual results.

Thank you again.

 

Special thanks to you for your good comments.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed most of the concerns raised, and I recommend this article for publication.

Reviewer 3 Report

Comments and Suggestions for Authors

Following the second revision of this paper, I recommend that it be published.