Dense-HR-GAN: A High-Resolution GAN Model with Dense Connection for Image Dehazing in Icing Wind Tunnel Environment

Round 1

Reviewer 1 Report

The authors suggested a scheme for dehazing the blurred image caused by the ice wind tunnel. A dense GAN model is proposed. The proposed method is compared with several dehazing techniques. Though the authors did the hard work still have some queries and suggestions to improve the manuscript:

1. Many statements are needed to be revised, like the abstract statement, “To address the issue of blurred images generated during ice wind tunnel tests due to attenuation caused by scattering and absorption when light passes through cloud and fog droplets, we propose a high-resolution dense connection GAN model called as Dense-HR-GAN for image dehazing, which is designed for this environment.” It is a very long statement and difficult to read.
2. In Fig. 3, what is the significance of fake and real?
3. How is it ensured that in the process of dehazing, the critical information of the image is not missed?
4. The metric defined in equation 9 is not precisely defined.
5. Why authors took the only one type of image for evaluation?
6. Is the proposed scheme suitable for other types of blurred images?
7. The contribution of the authors is not clear.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper proposes a high-resolution dense connection GAN model called Dense-HR-GAN for image dehazing in icing wind tunnel environment. Experiments show the competitive performance of the proposed model compared to some state-of-the-arts. Detailed comments are as follows:

1. One main concern of this paper is that the specific design of icing wind tunnel environment is not clear, since it can also be applied to traditional image dehazing. Please clarify this point.

2. Only one metric is used for evaluation. Other quality metric for natural images can also be exploited for comparisons, such as NIQE.

3. The compared methods are all before 2020. Can more recent ones be compared in the experiments?

4. Some quality metrics for image dehazing/enhancement/super-resolution are suggested to be pointed out in the paper, including Dehazed image quality evaluation: from partial discrepancy to blind perception, Perceptual quality assessment of low-light image enhancement, Blind quality assessment for image superresolution using deep two-stream convolutional networks, etc.

5. Please further proofread the paper, for example, “can achieves” -> “can achieve”.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have incorporated most of the suggestions. I don't have any further significant comments.

Reviewer 2 Report

The authors have addressed my comments.