Single Image Super-Resolution: Depthwise Separable Convolution Super-Resolution Generative Adversarial Network

Round 1

Reviewer 1 Report

The authors discuss on the advanced problem in image processing of image super- resolution. The proposed strategy is coherent with the topic. The introduced method does appear efficient and correct. The numerical results are quite impressive. I suggest the authors to view at the Cellular Nonlinear Networs- Cellular Neural Network in order to consider for future studies the use of another approach in image super reolution. I suggest the following paper that invite the authors to include in the references. International Journal of Bifurcation and Chaos in Applied Sciences and EngineeringVolume 15, Issue 7, July 2005, Pages 2063-2090  

The CNN paradigm: Shapes and complexity(Article)

Arena, P.,   Bucolo, M.,   Fazzino, S.,   Fortuna, L.Email Author,   Frasca, M.

The intrinsic parallelism of CNN, in my opinon, is quite interesting in order to perform image super resolution.

Some words of comments about my remark is welcome.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper proposes a new method (DSCSRGAN) to further enhance the visual quality of the SuperResolution Generative Adversarial Network (SRGAN) by using a deep learning approach. The paper is interesting and generally well written. However some issues need to be solved:

Line 40: instead of “Maximum A Posteriori (MAP)” it is written “Maximum a Posteriori (MAP)”; Line 80: instead of “Goodfellow et al.” it is written “Goodfellow”. Moreover, there is a need to include a reference there; There are many places in the manuscript where initial capital letters are not needed. Some examples: line 82 (“Using”) line 127 and 128 (“Residual unit”), line 138 (“Equation”), lines 193 (“Two”), etc. Line 84 and also line 87: The sequence “As follows” is not properly used. I suggest to link it with the previous sentence (“, as follows”); Line 105: I suggest to replace “ideas” by “goals”; Line 112: instead of “representation” it is written “represent”; Title of figure 2 is somehow confusing (mainly “our mode”) and needs to be reshaped; Line 129:  a full stop is missing before “The” (“. The”); Lines 141-143: the sentence is badly-shaped/unclear (mainly the part with epsilon) and needs to be reshaped; Not all variables used in equations are explained. For example, see eq. (10); Lines 168-169: Some details are needed to better understand this sentence; Line 175: A reference is needed for the theorem; Line 212: it is written that the new method increases the number of parameters with 2% compared with SRGAN. Can you give an estimation of how this value (number of parameters) affects the algorithm speed or the memory used?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors have successfully solved all my previous comments/concerns.