Diffusion Model with Detail Complement for Super-Resolution of Remote Sensing

Round 1

Reviewer 1 Report

This article proposes an adaptation of a diffusion model, called Diffusion Model with Detail Complement (DMDC), applied to high resolution image reconstruction. Their main contribution with respect to the baseline model is the implementation of a "detail complement" component, which allows a better reconstruction of small details and objects in complex scenes, as compared to other state-of-the-art reconstruction methods. 

 

Based on the illustrations and on the evaluation scores presented in the article, I think that the proposed method is a relevant contribution to the field of remote sensing super resolution. The introduction and related work sections provide a fair amount of literature review, however the remaining sections should be improved.

 

In the methodology section, there are several notations/variables that are not explained and not even mentioned in the text, as, for instance : $\omega$ in Equation 2, $q$ in line 176 and $\overline{\alpha}$ in Equation 4. 

 

In the “Comparison with state-of-the-art” section, interesting results are given in Table 1, but a more detailed discussion is missing.

 

Finally, we have two very short conclusion and discussion sections.

 

Globally, I believe that the paper provides enough contributions, but some parts are hard to read since they lack explanation, including the “detail complement” component.

 

A few minor suggestions/questions :

- Lines 118-121 : what is the difference between "optimized generative models" and "generation models" ? Doesn't the latter need optimization as well?

 

- Lines 151-153 : "We introduce DDPM into..." -> should DDPM be replaced by DMDC as defined  previously in the paper?

 

Some acronyms are used without being previously defined :

- Line 26 : SR

- Line 35 : LR...

 

- Line 81

"... we review and conclude..." -> "...we review and discuss.." ?

 

- Line 88

"The work adopted deep learning..." -> "Deep learning based methods..."

 

If a sentence starts with a citation, add the author’s name first as, for example :

- Line 140: ".[50] proposed..." -> "Nair et al. [50] proposed..."

- Page 6, bellow Equation (11) : "... [24] found..." -> "... Ho et al [24] found ..."

 

Figure 1 should appear later in the article since it is only cited in Page 5.

 

- Line 175 : "...which is only affected by $X_{t-1}$." -> Should $X_{t-1} be replaced by $X_t$ in this sentence ?

 

Below Equation (9) : "" ...distributions p and q:" -> put both p and q in math mode

 

Equation (10) : KL has not been mentioned nor described anywhere in the text. 

 

- Line 185 : "...almost difficult..." -> "...difficult..."

 

Algorithms 1 and 2 are not detailed enough. What are the input and output of both algorithms ? They should be given at the header, just below the caption.

 

- Line 210 : "...ruins..." -> "...clutter..."

Author Response

Thanks to the reviewers' comments and the editor's efforts, we have completed revisions to the previous manuscript. The revised manuscript and point-by-point response to comments have been uploaded, hoping to meet the requirements of the Remote Sensing journal. We express our sincerest gratitude to you and the reviewers for your efforts.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments for author File: Comments.pdf

Author Response

Thanks to the reviewers' comments and the editor's efforts, we have completed revisions to the previous manuscript. The revised manuscript and point-by-point response to comments have been uploaded, hoping to meet the requirements of the Remote Sensing journal. We express our sincerest gratitude to you.

Author Response File: Author Response.pdf