Elastic Downsampling: An Adaptive Downsampling Technique to Preserve Image Quality

Round 1

Reviewer 1 Report

The paper describes an adaptive downsampling technique based on a novel Perceptual Relevance metric and a new Elastic Downsampling method. The paper is interesting but has drawbacks:

1. The motivation in part 1 is not clearly presented.
2. Part 2 should be expanded with more references.
3. It'll better to present the processing pipeline at the beginning of part 3 graphically.
4. The authors should present the limitations of the proposed approach.
5. It'll be better to present a comparison in terms of more metrics. 
6. The paper needs part Discussion. Authors should discuss the results and how they can be interpreted from the perspective of previous studies and of the working hypotheses. Future research directions may also be highlighted.
7. The implementation of the proposed technique using a low-cost device (Raspberry Pi3) should be clearly presented.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is, overall, of good quality. Unfortunately, in this form, there are some aspects that raise questions and also some aspect that would improve the readability and experiment reproduction by other researchers.

The main aspect that draws attention is the limited number of images that were used in the testing stage, only 24. The dataset should consist of at least hundreds of images, to be considered relevant for the general case. Therefore, the authors must increase the number of images used in the testing phase to demonstrate the performance of the proposed algorithm. 

The second aspect that should be improved is the comparison with other methods. In the introduction, the authors describe multiple methods for the same operation, but in the comparison, only one reference scenario is used. It is difficult to assess the performance of the algorithm from only one comparison.

I present the elements that would increase the readability and understanding of the paper onwards.

An aspect that would greatly increase the readability of the paper is providing the whole proposed algorithm as pseudocode, clearly and in detail explaining all the variables that are used. In this way, the core of the contribution will be concentrated and would be of much value for experiment reproduction by other researchers. In the current state of the paper, the algorithm is mainly described in words, which is good only if it is complemented by rigorous mathematical formulas, to remove any possible misunderstanding. 

All the variables in a formula should be clearly defined, with as many details as possible, right after the formula. This is also true for figures. All figures should be described in detail right before or after them, making sure all the elements are covered. Also, the axis choice should be clearly explained. 

The abstract, in this form, it contains elements that should be part of the introduction. The abstract should be rewritten to state the current state of the art, the knowledge gap that is targeted by this paper, briefly explain how the knowledge gap is covered by the paper and highlight the main performances, in this order, with very concise phrases.

The references should also be improved. There are only 17 papers presented, and the newest are from 2016. Arguments should be presented about how this collection can represent the state of the art.

In conclusion, the paper is promising, but it requires some improvements, both in experiment phase and results presentation, but also in the general presentation of the method. The overall paper must be improved in form also.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors addressed all my concerns. The paper can be accepted in the present form.

Reviewer 2 Report

The authors properly answered to all reviewer's suggestions and now the paper is more clear and valuable.