An Improved Multi-Exposure Image Fusion Method for Intelligent Transportation System

Round 1

Reviewer 1 Report

The authors propose a solution for intelligent transportation systems by an image fusion method. In addition, a new multi-exposure image dataset for traffic signs, TrafficSign, is introduced to evaluate the method. The proposed solution takes into account the weather factor and environmental noise by employing a bilateral filtering and dehazing algorithm. The experimental results on the proposed dataset demonstrate the effectiveness and of the proposed solution.

Minor comments:
(1) The labels and captions of figures 5 and 6 should be increased since it is difficult to read them.

(2) On page 8, 2nd paragraph, the sentence fragment "frequencyP_r(R_k)of" should be spaced: "frequency P_r(R_k) of".

Major comments:
(1) Equations (1) and (2) the exponents are very small and should be enlarged or replaced by a function so the reader can see it easily.

(2) In Equation (3), the main operation between the exponents is not readable. I infer it should be an addition, but it is not clear.

Overall, I believe this work is relevant and presents interesting ideas. It tackles a real problem and provides a feasible solution.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, an image fusion method for Intelligent Transportation Systems is proposed and tested on the TrafficSign dataset.

 

Major Comments

1. It would be interesting to analyze the benefits of the proposed method in a real-world application, such as traffic sign classification, compared to having a single source of the image, and report the results in terms of classification accuracy improvement.
2. In Section 3.1, the preprocessing procedure is described. However, from the scientific perspective, it would be interesting to understand the reasons why such kind of equations are applied, and what are the effects. The authors are encouraged to comment more on this.

 

Minor Comments

1. Algorithm 1 should appear after its reference in the text, i.e., page 4.
2. In Equation 5, JC is indicated with the superscript ‘C’, while in the supporting text it is indicated with the subscript ‘C’. The notation should be uniform throughout the manuscript.
3. Figures 5 and 6 have a bad resolution. It is recommended to export it in vectorial form.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript is devoted to the development of an improved multi-exposure image fusion method for intelligent 12 transportation systems.

 

The manuscript is well structured, it contains all sections for this type of publication. The abstract briefly reflects the content of the manuscript. However, to my mind, the content of this paper should be revised since there are many mistakes and incorrectness.  Below, I present some of them.

1. English grammar is unsatisfactory. The article contains a lot of language mistakes and incorrectness.
2. The Related Works section is very short and it does not contain the current work in this subject area (for example, the use of wavelet analysis, etc.). Moreover, the paper will look better, if after the Introduction and Related Works sections will be allocated the unsolved part of the general problem and the following formulation of the paper objectivity.
3. Tables 1 and 2 both contain the appropriate algorithms. However, there is a shape for algorithm representation. This presentation is not correct.
4. Figure 1. What is the term Optimized processing? How this procedure was implemented within the framework of this research? What type of criteria were used?
5. The sections Experimental Result and Analysis and Conclusions are not complete too.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The effort made by the authors in revising the manuscript is appreciated.

However, the two major comments of the first revision have not been properly addressed yet. They are reformulated as follows:

1. By performing experimental results on traffic sign classification (e.g., on the German Traffic Sign Recognition Benchmark, GTSRB), it is possible to concretely evaluate the proposed method, and compare it against the state-of-the-art. The comparison metric, in this case, would be the classification accuracy on a traffic sign dataset. Due to the quality improvement of the image, the expectation of higher accuracy of a classifier having as inputs the images processed with the proposed Multi-exposure Image Fusion Method should be supported by the results. Hence, the authors are encouraged to extend the results section with these experiments.
2. The explanations of the formulas, in terms of what the algorithms do, are appreciated. However, it would be useful for the readers to understand why such formulas and algorithms are used, instead of other formulas and algorithms. The motivation behind the choice of the equations used to build the proposed method is key for a comprehensive presentation of the idea.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Thanks for the manuscript correction. Of course, it looks better. However, to my best mind, it should be revised again.
1. English grammar. The mistakes are a lot of. I do not think that I should point to each of them.
2. Table and algorithm have differed from each other. I think that Table 1 and Table 2 should be renamed: Algorithm 1 and Algorithm 2. There are not tables.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

The efforts made by the authors in immediately addressing the reviewers’ comments in a comprehensive way is appreciated.

In this third version, the quality of the manuscript is significantly improved.

There is only a minor comment:

• The authors should report the values of classification accuracy as the ratio between the images correctly classified and the total images, considering the sum for all the classes. More specifically, Tables III and IV only report the accuracies relative to a single class. The accuracy values over the whole dataset should be reported.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf