Effect of Bit Depth on Cloud Segmentation of Remote-Sensing Images

Round 1

Reviewer 1 Report

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

The paper titled "Effect of Bit Depth on Cloud Segmentation of Remote Sensing Images" uses the deep semantic segmentation algorithm UNet and a set of widely used cloud labeling data set "L8 Biome" to find the relationship between bit depth and segmentation accuracy on different surface landscapes. The results show that 16-bit images have slightly better results when normalization is used. So the 8-bit and 16 don't have many differences when considering the balance of efficiency and accuracy.

According to cloud segmentation results. There is less than a 1% difference between normalized results and actual results. It is also seen that when remote sensing images contain white backgrounds like ice and snow the accuracy is significantly reduced(around 20%) due to the method used. Kappa values also show that efficiency also decreased tremendously.  

Major Problem

Even though the paper is well written it has very little contribution to the literature since it only shows how the existent method (Unet) performs on 8-bit and 16-bit remote sensing images.  That would be great if you had any improvement in Unet. Or had additional feature extraction techniques which will help Unet to improve its detection accuracy. Or May be focused on increasing the accuracy of snow and ice background remote sensing images.

Minor Changes.

Line 257-265 line spacing is different than other lines. 

Minor English grammar check, please.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The author uses a widely used cloud marker dataset, "L8 Biome," as validation data to explore the relationship between bit depth and segmentation accuracy in different surface landscapes when the algorithm is used for cloud detection. My detailed comments are listed below, which hopefully can help the authors improve the quality of their work.

(1)The author only proves the accuracy of the proposed method in his manuscript. Is it possible that testing on other network structures will yield different results? Can a single network structure explain the universality of the impact of bit depth on cloud segmentation in remote sensing images. If possible, the author should conduct more rigorous experiments.

(2)In section 2, the author divides the selected data into training sets and test sets based on 4:1. Is the proportion of training sets too high. In reality, obtaining remote sensing image data is expensive and laborious. Can less training sets be used to verify the superiority of the proposed algorithm?

(3) The references are relatively old. Only about five papers have been published within three years. Some related works should be further systematically surveyed.

(4)The manuscript have some mistakes in grammar and syntax, typos and unprofessional English expressions. It is recommended that the author conduct a comprehensive review of the manuscript.

(5)In Table 1, the results of 8BN and 16B seem to differ little. Does it indicate that the normalization operation achieves the same effect as improving bit depth? It is recommended that the author explain this.

(6)The author mentioned in the abstract that this means that data selection and classification do not always need to follow the highest possible bit depth when doing cloud detection, but should consider the balance of efficiency and accuracy. However, in Table 2, there is no significant difference in training time between 8B and 16B, but their accuracy has significantly improved. Does the normalization operation have a more significant impact on efficiency?

(7)The novelty of this paper is not well stated. It is suggested to improve the novel parts of the design and give more motivations and analysis of those parts.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Thank you for addressing my concerns.