Undistorted and Consistent Enhancement of Automotive SAR Image via Multi-Segment-Reweighted Regularization

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors are describing a novel method for enhancement of automotive synthetic aperture radar images through multi-segment-reweighted regularization (MSR). The paper is well written and organized, it contains both thorough theoretical considerations and experimental validation, and finally the presented approach is novel and original. Thus, as such it could be considered for publication in MDPI Remote Sensing journal if some flaws are remedied.

My major concern is the lack of the computational complexity analysis (in terms of time and space requirements) of the proposed MSR method since it introduces multiple iterations and complex weighting schemes. Please state what is the computational cost and how it compares to more standard methods? How does this impact practical implementation feasibility on resource-constrained automotive systems?

Can the authors also include the execution times of the proposed algorithm and compare it to other algorithms on their workstation computer?

Also, since the performance of the proposed MSR algorithm depends on hyperparameters such as the number of segments (P) and the super-resolution factor (fSR), the authors should consider including the sensitivity analysis on these parameters.

Please comment is there a sense to use ℓ₂ regularization method besides the ℓ₁ regularization?

Please compare and benchmark the proposed method not only with conventional ℓ₁ regularization methods, but also with some recent deep-learning-based SAR enhancement techniques.

A question to the authors would be: is there a possibility to theoretically and rigorously prove the convergence of their iterative algorithm?

Annotate figures 2, 3, 4, 5 and 6 giving axis labels and indicating where, e.g., -λ₁ and λ₁ ticks are on the x-axis.

Please give more information on the used 79GHz SAR system design by Beijing Autoroad in reference [36]. At least some elementary data such as which chip does it use, how many transmitters and receivers does it have, etc.

For figure 7. please indicate distances in meters.

Figure 10 caption the last one should be: "f) RL1 with our weighting scheme in Equation (41)" not b).

Finally, consider open-sourcing MATLAB code implementation of the proposed method on some publicly available repository on GitHub, GitLab or Bitbucket as it will greatly ease the reproducibility of your results.

Comments on the Quality of English Language

The paper has several typos:

 - in the Abstract, line 4 it writes: "Sparse regularization methods ave the potential to enhance image quality" whereas it should be "Sparse regularization methods have the potential to enhance image quality"

 - line 197: there is a double a in "denotes a a small positive constant."

 - Capitalize title of the subsection 3.2: Weighting Scheme instead of weighting scheme

 - line 344: "can not" -> "cannot"

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This research addresses issues in automotive SAR imaging such as poor resolution, noise, and sidelobe artifacts. Conventional ℓ1 regularization methods struggle with radar cross section (RCS) distributions across wide dynamic ranges, resulting in insufficient sidelobe suppression and amplitude distortion. The authors propose Multi-Segment-Reweighted Regularization (MSR), a novel weighted regularization approach that localizes global scattering point enhancement to the mainlobe scale, effectively mitigating sidelobe interference regardless of RCS variations. MSR employs multi-segment regularization to constrain amplitude within mainlobes, preserving original response characteristics, and introduces a new thresholding function called Thinner Response Undistorted THresholding (TRUTH). Real-data experiments validate the feasibility and effectiveness of the proposed method for automotive SAR image restoration.
For real application, I think the author may further investigate methods to reduce computational complexity for real-time processing in automotive applications, potentially through algorithm simplification or hardware acceleration. In addition, the performance evaluation for moving target will be increase the practical impact of the proposed technology.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper proposed a novel reweighted regularization method, termed Multi-Segment-Reweighted Regularization (MSR), for automotive SAR image restoration to effectively mitigate sidelobe interference. It leverages the advantages of the existing two frameworks, and a new thresholding function is presented. In addition, an iterative algorithm is presented, and a lot of experimental results are given. Overall organization is good. However, some minor comments should be solved before final publication:

1. It would be better if the insufficient capacity of the existing penalty terms and the reason to propose a new penalty term could be described more clear in Introduction.
2. It would be better if the definition of symbols such as supp, ||_k and E() could be provided.
3. Please check if there are spelling mistakes in Eq. (11) and the sentence below Eq. (11), and the symbol of the log-sum penalty function in Eq.(11) should be consistent with it in Fig. 1.
4. It would be better if the legend for the red and blue lines in Fig.2 and Fig.3 or their definition could be provided.
5. It would be better if the particularity of automotive SAR image reconstruction could be analyzed, and the relationship between automotive SAR image reconstruction mechanism and the design idea of the proposed method should be elaborated more clearer.
6. It would be better if the computation complexity could be analyzed.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed almost all the issues I have raised to the best of their abilities, therefore  I recommend the paper for publication by MDPI Remote Sensing.