A Remote Sensing Image Quality Interpretation Scale Characterization Method Based on the TTP Criterion
Round 1
Reviewer 1 Report
The manuscript demonstrates a commendable level of creativity which addresses the issue of existing image quality evaluation models can not represent the perceptual characteristics of human visual image information. By introducing the TTP criterion as a key parameter in the remote sensing image quality equation and utilizing more appropriate function to represent key components, it finally constructs a novel scale characterization method for remote sensing image quality interpretation to effectively quantify image confidence. This method significantly improves the reliability and accuracy of the model. Throughout the whole argumentation process, the author exhibits clear thinking, logical structure, and accurate expression of viewpoints, effectively combines their expertise with the research topic and meets the publication requirements in terms of both the topic selection and viewpoint justification.
Here are some comments:
1. When objectively presenting research results published by others, the general present tense is usually used. Therefore, please change the tense of the section describing the current status of the research in paragraph 4 of the introduction to the general present tense.
2. The language of communication manuscripts should be more concise. However, the second sentence of chapter 2, paragraph 3 in the manuscript "However, the evaluation of image interpretability is a specific human behavior driven by subjective consciousness, and its results will be influenced by the intuitive perception of the observer, i.e., the perceptual response characteristics of the human eye to the valid information contained in the image will have a direct impact on the results of the human prediction interpretation task." is wordy and suggest re-expression.
3. The format of the manuscript subsection headings must be uniform. And the title of the first section in Chapter 4 should be modified to "4.1 Construction of the New Model" (capitalized);
4. The conclusion part is not concise enough and suggested to be rephrased.
The manuscript contains some grammar and expression errors. It is recommended to thoroughly read the article and make necessary revisions.
Please pay attention to the use of articles in the writing process. The definite article "the" should be added when GIQE4 is mentioned in the manuscript.
Author Response
请参阅附件。
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1. A brief summary
The article is devoted to the study of the issues of determining the quality of remote sensing images using model representations that take into account both objective technical characteristics of raster data and subjective criteria of human visual perception. An improved NORSIQE-model representation of image quality is proposed, which is generally of some interest. However, there are principal questions and comments about the formulation and methodology of the study.
2. General concept comments.
1. The presented study postulates that the low reliability/accuracy of GIQE are due to the lack of consideration of the features of human visual perception. Based on this, accounting for the transfer characteristics of the human eye is introduced. However, this formulation of the study seems to be unfounded. From the conceptual purpose of GIQE, as well as from the diagram in Fig. 1, it follows that GIQE is used primarily for satellite and aerial data from remote sensing of the Earth domain. At the same time, the actual relevant formulation of the problem of thematic processing and classification of such data is associated either with machine methods, or with the visualization of images on the display and its interactive visual study using scaling functions in both spatial and brightness dimensions. Note that in raster remote sensing products, the dynamic range of bands usually exceeds 8 bits, and the raster is not intended to be analyzed instantly over the entire range of available pixel values. Arbitrarily high coefficients of raster zooming are used, as well as methods for correcting brightness histograms with contrasting objects of interest within the selected arbitrarily narrow dynamic range. In such setup, it is not become important the visual functions of the operator, but his ability to manipulate the spatial scaling and dynamic range of the image is critical. At the same time, the image quality is determined by its technical characteristics and can be measured using traditional objective criteria. With this in mind, it seems that the formulation of the proposed study is poorly connected with actuality, or is not clearly presented. It is proposed to major revise the article in order to solve this problem.
2. An experiment is being carried out, showing the effectiveness of the proposed model and an 87% improvement in the quality of modeling. However, this result looks unconvincing, since the method of obtaining and the source of test data are obscurely described. The following ambiguous formulations about test data are given, generally confusing the reader:
- expert artificial NIIRS value (line 173)
- screened by experts (line 238)
- based on the professionally trained manual interpretation of NIIRS values (line 255).
In addition, it is recommended to explain why the test data were examined only in the quality range from 4 to 8.
3. Specific comments.
1. Figure 1. It is desirable to clearly explain to readers with the help of this figure the meaning of the parameters included in expression 1. The figure shows graphs of some parameters, but they are given in the form of arbitrary pictograms, the details of which are difficult to distinguish. Nevertheless, such graphs are important in order for the reader to assess the composition of the parameters for calculating NIIS. In addition, the title of the Fig.1 seems to be incorrect, since the picture includes not only GIQE concept.
2. Figure 3. Please, shorten the title without repeating the same phrases.
3. Figure 4. Check the correctness of the formula.
4. «Contrast Threshold Function of the human eye». Where were the specific values of this function used in the calculation taken from?
5. «∆NIIRS». It is suggested to give a formula for calculating this parameter, since further ∆NIIRS is constantly used, and there are questions about it. For example, in Figure 5a, the graphs grow monotonically (as if ∆NIIRS is a cumulative value), and in Figure 6 there is no growth.
6. Line 293. «a large number». Debatable. It is advisable to give a specific number.
7. Line 298. «is of great significance». Debatable. It is necessary to prove this statement.
8. I found earlier references to the description of TTP than those given in the article, it is advisable to refer specifically to the primary sources.
9. Some formulations are difficult to understand, for example:
«interpret scale prediction»
«remote sensing image quality interpretation model»
«image quality interpretation scale representation»
«images.model»
and others.
Thank you very much for your work!
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors attempt to establish an effective mapping relationship between the objective image quality of satellite remote sensing images and human subjective perception features using the TTP criterion, forming a more accurate remote sensing image quality equation, which is interesting. However, there are several critical points that need to be addressed before the manuscript can be considered for publication.
1. The experimental results are not convincing. GIQE5 was released around 2019, so why is this paper not compared with GIQE5, or GIQE4C which is used for continuity data. The argument would be more convincing after such a comparison.
2. From section 4, the authors are using TTP instead of the two parameters RER as well as H in the GIQE4 formula, and the innovation stems from the fact that using the TTP criterion to cover the MTF parameter. This method could be able to measure the diffusion phenomenon and the image clarity in the remote sensing optical imaging system. In section 4.2 of the paper, the authors show that the prediction accuracy of this paper is 14% higher when comparing the MTF-Nyquist model, the MTF-50 model and the MTF-Area model. Is the accuracy comparison based on the same validation dataset? This is not explained in the paper. The data in rows 273-275 of the paper could be considered for inclusion in a figure or table description.
3. The number of references cited throughout the piece is too sparse (20 whole references). A large number of arguments need to be supported by relevant literature, for example:
(a) The argument in lines 87 to 90 should be supported by some references. Here the author only gives the conclusion that "image interpretability evaluation is a specific human behaviors driven by subjective awareness".
(b) In lines 132 to 134, the introduction of TTP to build a bridge between subjective and objective evaluation for methods to improve subjective image evaluation quality. Authors should consider whether there are similar (but not limited to remotely sensed imagery) methods and literature. If there is, then it should be cited here to support it.
(c) In lines 149 to 150, the authors argue that the TTP criterion is directly used as an independent variable in the remote sensing image quality equation. There is no valid theoretical as well as methodological support here. At least some explanation of the argument is needed in conjunction with the above equation.
4. Typically, the evaluation results of subjective evaluations tend to vary from one evaluator to another, with poor stability, as well as inefficiency when faced with large data volumes of imagery. In this paper, whether regression analyses conducted by the authors using a sample of 280 samples are generalizable to the analysis of other remotely sensed images?
Overall, the manuscript is well written and clearly organized.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
1. A brief summary.
The second version of the article "Remote sensing image quality interpretation scale characterization method based on TTP criterion" is submitted for review. The reviewer thanks the authors for the revision of the article and answers to a significant part of the comments. It is proposed to finalize the article in a minor way and respond to unresolved and to additional specific comments.
2. Specific comments (from review 1, unresolved).
2.1 The diagrams in Figure 1 are difficult to see. A flowchart with obviously indistinguishable details does not adorn an article dedicated to assessing the quality of imagery. Please elaborate the Figure 1.
2.2. Figure 4. Please check the expression (RER?).
2.3. In Figure 5a, the graphs are monotonously growing, and in Figure 6, there is no monotonous growth on similar-meaning graphs. This fact is somewhat confusing. It turns out that only the samples for Figure 5a were ranked by delta-NIIRS?
3. Specific comments (additional).
3.1 Line 120 («of the system»). It is proposed to specify exactly which components are included in the system in the presented formulation of the study. Is the display the part of the system?
3.2. It is proposed to put directly into the manuscript a specific method and formulas for calculating the MTF of the system and CTF, then the effort will meet the important criterion of "reproducibility", which is crucial for recognize the research by community.
Thank you very much for your work!
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript has been revised carefully, the quality of this paper has been improved significantly. It is well-written and well-organized. I recommend accepting the manuscript.
Author Response
We thank you very much for accepting this manuscript and for your valuable suggestions.
