Algorithm for Cloud Removal from Optical Remote Sensing Images Based on the Mechanism of Fusion and Refinement

Fig. 6. Results of the RICE1 dataset. (a) Cloudy images; (b) cloud-free images, (c) CAP; (d) BCCR; (e) DCP; (f) DehazeNet; (g) MSCNN; (h) proposed algorithm

Download full size

Fig. 7. Results of the RICE2 dataset. (a) Cloudy images; (b) cloud-free images, (c) CAP; (d) BCCR; (e) DCP; (f) DehazeNet; (g) MSCNN; (h) proposed algorithm

Download full size

Fig. 8. Results of real images. (a) Cloudy images; (b) CAP; (c) BCCR; (d) DCP; (e) DehazeNet; (f) MSCNN; (g) proposed algorithm

Download full size

Table 1. Quantitative analysis results of the RICE1 test set
View table
Table 1. Quantitative analysis results of the RICE1 test set
Algorithm Evaluation index
SSIM PSNR /dB
CAP 0.7950 20.5820
BCCR 0.6525 17.1802
DCP 0.6814 16.7349
DehazeNet 0.8019 20.8613
MSCNN 0.7549 20.1289
Proposed algorithm 0.8964 23.9876

Table 2. Quantitative analysis results of the RICE2 test set
View table
Table 2. Quantitative analysis results of the RICE2 test set
Algorithm Evaluation index
SSIM PSNR /dB
CAP 0.7683 20.7902
BCCR 0.5904 16.6502
DCP 0.5538 15.8937
DehazeNet 0.7351 20.8093
MSCNN 0.6499 17.4240
Proposed algorithm 0.8397 22.8645

Table 3. Quantitative analysis results of the ablation experiment of loss functions

View table

Table 3. Quantitative analysis results of the ablation experiment of loss functions

Dataset	Removed loss function	Evaluation index
Dataset	Removed loss function	SSIM	PSNR /dB
RICE1	None（baseline）	0.8964	23.9876
	Perceptual loss	0.8737	22.9699
	Gradient loss	0.8697	22.2485
RICE2	None（baseline）	0.8397	22.8645
	Perceptual loss	0.8298	21.1345
	Gradient loss	0.8176	20.6549

Table 4. Quantitative analysis results of the ablation experiment of discriminator network

View table

Table 4. Quantitative analysis results of the ablation experiment of discriminator network

Dataset	Discriminator	Evaluation index
Dataset	Discriminator	SSIM	PSNR /dB
RICE1	None	0.8494	20.0365
	Discriminator^［21］	0.8789	21.0264
	Proposed	0.8964	23.9876
RICE2	None	0.7947	19.1643
	Discriminator^［21］	0.8096	21.0645
	Proposed	0.8397	22.8645

Table 5. Model parameters, FLOPs, and average running time
View table
Table 5. Model parameters, FLOPs, and average running time
Algorithm Parameters /10³ FLOPs /10⁶ Running time /s
CAP 1.19
BCCR 2.92
DCP 1.35
DehazeNet 8.1 5.1 1.29
MSCNN 8.0 4.8 2.14
Proposed algorithm 964.1 26.9 1.21

Tools

Get Citation

Copy Citation Text

Xiaoyu Wang, Yuhang Liu, Yan Zhang. Algorithm for Cloud Removal from Optical Remote Sensing Images Based on the Mechanism of Fusion and Refinement[J]. Laser & Optoelectronics Progress, 2024, 61(4): 0428012

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Remote Sensing and Sensors

Received: Nov. 14, 2022

Accepted: Mar. 6, 2023

Published Online: Feb. 20, 2024

The Author Email: Wang Xiaoyu (wxyhit197745@sina.com), Liu Yuhang (lyhang95@163.com)

DOI:10.3788/LOP223038

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Quantitative analysis results of the RICE1 test set

Table 1. Quantitative analysis results of the RICE1 test set

Table 2. Quantitative analysis results of the RICE2 test set

Table 2. Quantitative analysis results of the RICE2 test set

Table 3. Quantitative analysis results of the ablation experiment of loss functions

Table 3. Quantitative analysis results of the ablation experiment of loss functions

Table 4. Quantitative analysis results of the ablation experiment of discriminator network

Table 4. Quantitative analysis results of the ablation experiment of discriminator network

Table 5. Model parameters, FLOPs, and average running time

Table 5. Model parameters, FLOPs, and average running time