A Deep Learning Framework for Multi-Plane Computer-Generated Holography

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This work proposed a physics-informed deep learning framework that directly generates holograms for 3D multi-plane displays. The proposed method obtained phase hologram successfully reconstructs sparse multi-plane structured patterns with low visible crosstalk. The author has done a lot of work and experiments to verify the feasibility and performance of the method. However, there are still some problems that are not explained clearly, so clear explanations and careful revisions are required before publishing this work. The specific questions are as follows:

1. The manuscript demonstrates impressive reconstruction results for specific sparse patterns. However, deep learning-based methods frequently encounter challenges regarding generalization. It is recommended that the authors provide additional test results using more complex patterns that extend beyond the current dataset, such as natural images with continuous grayscale or significantly denser structures. If the proposed model is inherently optimized for sparse patterns, the limitations regarding its practical scope of application should be explicitly discussed in the text.
2. Several deep learning-based CGH frameworks, such as HoloNet or architectures based on algorithm unrolling, have been proposed in recent years. To better underscore the unique advantages of the proposed framework in multi-plane tasks, the authors should provide a comparative analysis with these mainstream deep learning models. Specifically, comparisons regarding computational efficiency (e.g., inference time) and reconstruction accuracy (e.g., PSNR or SSIM) would be highly valuable in validating the superiority of this approach.
3. The manuscript introduces a "trainable Fourier transform layer." The authors should clarify the accuracy and validity of this layer when handling scenarios that do not satisfy long-distance approximations, such as near-field diffraction. Furthermore, real-world optical systems are inevitably affected by coherent noise and laser speckle. I suggest the authors discuss whether these physical factors are incorporated into the physics-informed training process or if the network possesses a inherent degree of robustness against such optical noise.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Please see the attached file.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

no

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Having carefully reviewed the response letter, I find that the authors have addressed my concerns thoroughly and comprehensively. All my concerns have been addressed, and I recommend the paper for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript can be accepted.