Speckle Reduction in Digital Holography by Fast Logistic Adaptive Non-Local Means Filtering

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1 The innovation of the paper is to integrate the Logistic function into the weight calculation of NLM algorithm, but the description is not very clear, especially in the derivation of mathematical expressions. For example, how is Eq. (2) derived to Eq. (14)?

2 In Figure 7, ds=3 pixels and Ds=17 are set to verify the optimal values of ds and Ds, respectively. Has the effect of ds and other combinations of Ds on image denoising performance been verified?

3 In the experiments of section 4, only the parameters of SI are shown. Please add the values of PSNR and SSIM.

4 Figure 12 shows the holographic reconstruction results for NLM, improved NLM and LA-NLM. Are the other methods mentioned in Table 2 validated?

5 There is no vertical coordinate heading in Figure 4. Please indicate the physical quantity and unit.

6 The physical quantities and units are not indicated in Figure 5. Please add the related contents.

Comments on the Quality of English Language

No comment.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Speckle noise is a multiplicative noise that is present in coherent digital holography. The authors write digital holography generally but the problem exists only in coherent holography. There is another branch of digital holography with spatially incoherent light without speckle noises. Non local means algorithm is one of the widely used methods for speckle reduction. The authors have cited most of the important articles. However, given the long list of published articles, I suggest adding more relevant references. 

The optical experiment has been carried out using a commonly used Mach-Zehnder interferometer for both transmission as well as reflective objects. The reconstruction results are satisfactory as the reflection results are noisier than the transmission ones.  

The proposed method LA-NLM uses a modified weight function called the Logistic function in the shape of the letter S which has an improved smoothening effect compared to the previous NLM method. This is the gist of the paper.

The new method has been compared with many existing methods such as NLM, improved NLM, BM3D etc and the performance of the new method is proved to be better than the existing methods in terms of speckle noise. 

The idea is interesting with promising simulation and experimental results. I only have few two concerns. Comparing the performances of LA-NLM with other methods in simulation and experiments, in experiments the difference is not high which must be addressed. 

While the new method reduces speckle better than other methods, it looses sharpness. Comparing the experimental results of improved NLM and LA-NLM, the RHS features appear blurred in the case of LA-NLM. The authors must address this aspect also.

Another suggestion is to include the photograph of the set up in addition to the experimental set up. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The paper proposed a new imaging processing algorithm to reduce the speckle noise in digital holography. The algorithm is called as Logistic adaptive non-local means (LA-NLM), which is a variant to the NLM algorithm by introducing the logistic function to the weight function. The author performed a detailed study on optimizing the parameters of this LA-NLM. The simulation and experiment also showed that this LA-NLM can effectively reduce the speckle noise. Overall, I think this is a good study which is valuable to the field of image processing. Below are some further comments and questions.

1) In your results, you compared the images processed by different methods incl. yours and showed that the LA-NLM performs the best. Could you compare your method with other non-imaging processing methods, like using the uncorrelated pattern. 

2) recommend you also add the comparison of time cost for different image processing method. I think one of benefit of using the integral is the improved time cost. This may be another good show case for your method. 

3) do you think this method trade off the resolution of imaging given that there is some kind of averaging effect. 

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf