A Texture-Considerate Convolutional Neural Network Approach for Color Consistency in Remote Sensing Imagery

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study proposes a convolutional neural network-based color consistency method for remote sensing cartography that considers both global and local color mapping and texture mapping constrained by the source domain.

This is a good work which should consider the following comments:

1.       For remote sensing applications, the authors should consider more examples, such as detection, classification, and segmentation, high-speed ship detection in sar images based on a grid convolutional neural network, depthwise separable convolution neural network for high-speed sar ship detection, a mask attention interaction and scale enhancement network for sar ship instance segmentation, cfar-dp-fw: a cfar-guided dual-polarization fusion framework for large scene sar ship detection.

2.       This a texture CNN, so why texture, not others? What about LBP?

3.       For Figure 2, please give the detailed implementation of layers.

4.       The authors should compare other previous work.

5.       Pay attention to a polarization fusion network with geometric feature embedding for sar ship classification, injection of traditional hand-crafted features into modern cnn-based models for sar ship classification: what, why, where, and how, hyperli-net: a hyper-light deep learning network for high-accurate and high-speed ship detection from synthetic aperture radar imagery.

6.       Some language can be improved.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents an innovative and effective approach to addressing color consistency in remote sensing imagery, supported by comprehensive evaluation and practical relevance. However, addressing the following errors is essential before the manuscript can be published.
•
Line 152: The descriptions of downsample and upsample are reversed. Correcting this will improve the clarity and accuracy of the methodological description.
•
It is unclear what metrics were used to calculate the figures in Table 1. Are these figures derived from the metric defined in Eq.7? Clarifying this point will help in understanding the evaluation criteria used.
•
L2 loss is prone to outliers and emphasizes high frequencies. Did the authors consider using Huber loss or perceptual loss (e.g., SSIM)? Including the results of such alternative loss functions would provide a more comprehensive understanding of the method's performance.
•
The details of the four experimental settings are insufficient. For example, what constitutes small and large temporal differences? Providing a numerical definition or specific time ranges would add clarity to the experimental design.
•
While the proposed method performs well on the datasets used, its generalization to different geographical areas or other landscape types might not be fully addressed. Further discussion or additional experiments on diverse datasets would strengthen the manuscript.
•
The performance of the proposed method may be highly dependent on the specific sensors used in the experiments. The manuscript might benefit from discussing the method's adaptability to other sensor types not included in the study.
•
There is a lack of discussion on the computational complexity or inference speed of the proposed method, especially given its two-stage structure. Addressing how the method scales with very large datasets or high-resolution imagery would be valuable.

Comments on the Quality of English Language

none

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript tries to address the issue of color discrepancies in remote sensing images caused by various factors such as radiation, atmospheric conditions, sensor perspectives, and land cover, which hinder effective image mosaicking and applications. Existing color consistency adjustment methods often result in texture distortions due to their inability to handle complex nonlinear color-mapping relationships. To overcome these challenges, the authors propose a convolutional neural network-based method that incorporates both global and local color mapping and texture mapping constrained by the source domain. This approach effectively manages complex color-mapping relationships while minimizing texture distortions. Comparative experiments with remote-sensing images from different times, sensors, and resolutions showed that the proposed method achieved superior color consistency, preserved fine texture details, and produced visually appealing results, aiding in the creation of large-area data products.

This manuscript is well written, just a few suggestions to improve this paper:

1)    Figure 1: The Equation’s variables in the Fig. 1 needs to be explained to help the reader to better understand.

2)    Eqs. (2, 4, 14, 15): The norm operators in the these Equations need to be explained.

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

It seems that all my comments are not solved. I have no more comments.

Let AE decide it.