A Method Used to Improve the Dynamic Range of Shack–Hartmann Wavefront Sensor in Presence of Large Aberration
Round 1
Reviewer 1 Report
Wen Yang, Jianli Wang and Bin Wang: A New Method to Improve the Dynamic Range of Shack–Hartmann Wavefront Sensor in Large Aberration Situation
The authors present a neural-network solution to the problem of the limited dynamic range of Hartmanngrams. The idea is nice, and should be advanced.
A previous relevant method is Fourier analysis of Hartmanngrams, where the lenslet range is not an issue, since spot movement in the Fourier domain is simply a phase change, even by a large motion. This is quite similar to the proposed method here, which uses the cross-correlation between the spots and a template, which is the positions of the spots in the lack of aberration (Eq 9). Very early references are given in Canovas and Ribak (Applied Optics 46, 1830, 2007), references which the authors might have missed. These studies also deal with the cases of noisy systems, not mentioned in this work.
I have some practical questions, such as what is the extracted global feature (page 8)? Is it the approximate location of the spots? Some of the terms are not familiar to optics readers (one-hot after Fig 3, mlp and max pool in Fig 4, k classes in page 8, g and gamma in Fig 6, softmax loss). A very short explanation should be enough.
The calculation of the dynamic range (Eq 9) is interesting, but simply finding the difference between the given wavefronts or their slopes and the calculated ones, as a function of the RMS of the input slopes, could be more informative.
It is not clear what is the wavefront color scale in Figs 7, 8. Is it possible also to invert the spot array (negative image) to better see the spots and grid?
Author Response
Thank you very much for your valuable comments. We have revised the manuscript and uploaded the response letter. Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The study proposes a novel method to increase the dynamic range of Shack–Hartmann Wavefront Sensors (SHWFS) aberration detection. The proposed method can allow measuring of the wavefront aberrations with values that would be impossible to measure for conventional SHWFS without the huge errors due to the displacement of the focusing spot. In general such large aberrations can be present in ophthalmology, where it would be hard to study the patients with severe sight problems or patients who undergone laser treatment. In this case, usually other devices are applied, such as interferometers or holographic wavefront sensors. However, SHWFS being much more compact would be ideal choice if their dynamic range would allow such measurements. The dynamic range of any traditional SHWFS is limited by their constructive parameters such as lens size and focus length. Thus, digital methods of increasing the dynamic range would allow applied to increase their dynamic range and expand the application of the already produced devices.
In this work, the authors devised the method to increase the dynamic range of SHWFS based on the application of the neural networks to determine the displaced focusing spot even if it enters the other adjacent sub-aperture range. The authors performed numerical simulations using a set of randomly generated wavefronts, to test the performance of their method. The manuscript's structure is clear. However, there are certain parts authors need to clarify before I can recommend it for publishing.
1. The manuscript’s title should be revised. It is assumed that the method proposed in the manuscript is novel, thus, there is no need to emphasize the word “new”. In addition, “large aberration situation” is a very strange construct. I suggest going for “in presence of large aberrations”.
2. The Introduction should be expanded in part of the applications for the large aberrations detection. Currently it’s only mentioned in one sentence. For now it seems that other fields doesn’t require such solutions, but what about for example measuring high order parabolic mirrors during manufacturing.
3. Part 2.1 of the manuscript describes the working principle of the conventional SHWFS, but authors doesn’t cite any relevant articles or books, hinting that the algorithm is original that is clearly false. Please cite relevant articles to the topic.
4. In part 2.2 the following sentences closely repeat:
“Although the light spot detected by the CCD is affected by the aberrations of the human eye, the aberrations of the human eye mainly show defocus and astigmatism; the higher order aberration component is relatively small, and the defocus aberration and astigmatism are both central symmetric. Therefore, it is reasonable to approximate the actual spot distribution to the normal distribution.”
And next paragraph
“Due to the dynamic characteristics of human eye aberrations, although it is impossible to estimate the influence of human eye aberrations on the spot when selecting the autocorrelator template, human eye aberrations are mainly defocused and astigmatism, which are centrosymmetric, and the spot under its action still conforms to the normal distribution.”
The second part should be revised also because it’s confusing. Мне кажется тут я уже придираюсь…
5. In part 3.1. the authors mentioned that the incident wavefront was derived from the first 27 Zernike polynomials, but once again didn’t cite any relevant literature to prove it. From my knowledge for the most wavefront sensing problems, the first 15 Zernike polynomials are sufficient (it contain approximately 90% of all energy [Rukosuev, A.; Nikitin, A.; Belousov, V.; Sheldakova, J.; Toporovsky, V.; Kudryashov, A. Expansion of the Laser Beam Wavefront in Terms of Zernike Polynomials in the Problem of Turbulence Testing. Appl. Sci. 2021, 11, 12112. https://doi.org/10.3390/app112412112]). In addition, the authors derive from their dataset from the atmospheric turbulence model, where higher modes can be present, but what about the human eye aberrations that were mentioned previously? As authors said, the main components there are the defocus and astigmatism.
6. While the authors mentioned the setup used in the experiment, they have neglected to show the performance of their method. How much time does it take to compute the solution of the centroid search? Can it still be used in real-time measurements?
7. The authors mentioned two extreme cases, where the proposed method doesn’t work. Can you speculate on the possible solution?
Author Response
Thank you very much for your valuable comments. We have revised the manuscript and uploaded the response letter. Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors took into account all the comments, so the article can be recommended for publication. I would like to wish the authors to bring their result to a wide practical implementation.