Optics and Precision Engineering, Volume. 32, Issue 23, 3479(2024)
Real-time deblurring of wideband small target based on attention mechanism
Figures & Tables(14)
Fig. 4. Deblurring results of models with different numbers of modules
Fig. 6. Experimental comparison results of some infrared simulation datasets
Fig. 7. Experimental comparison results of some visible light datasets
Fig. 8. Experimental comparison results of some infrared simulation datasets
Fig. 9. Experimental comparison results of some visible light datasets
Table 1. Running time of models with different numbers of modules
View tableView in ArticleTable 1. Running time of models with different numbers of modules
Module Time/s M2_2 0.027 0 M2_4 0.027 8 M2_6 0.034 7 M4_2 0.028 4 M4_4 0.035 4 M4_6 0.042 2
Table 2. Details of image datasets
View tableView in ArticleTable 2. Details of image datasets
Dataset Blur type Image size Object Relative size Number GoPro Motion blur 1 280×720 Various - 3 214 Infrared ray Mixed blur 640×512 Plane 0.05-0.10 180 Visible light Mixed blur 960×540 Plane 0.03-0.07 1 130
Table 3. Comparison of evaluation indicators for deblurring infrared small targets using different algorithms
View tableView in ArticleTable 3. Comparison of evaluation indicators for deblurring infrared small targets using different algorithms
Method PSNR/dB SSIM Time/s MRDNet 31.544 0.875 0.056 Restormer 31.770 0.878 0.086 Fftformer 31.585 0.877 0.088 Swintormer 31.641 0.877 0.125 MSSANet 32.083 0.877 0.029
Table 4. Comparison of evaluation indicators for deblurring visible small targets using different algorithms
View tableView in ArticleTable 4. Comparison of evaluation indicators for deblurring visible small targets using different algorithms
Method PSNR/dB SSIM Time/s MRDNet 16.748 0.930 0.061 Restormer 16.851 0.929 0.084 Fftformer 16.970 0.931 0.087 Swintormer 16.983 0.931 0.104 MSSANet 17.118 0.930 0.028
Table 5. Ablation experiment results
View tableView in ArticleTable 5. Ablation experiment results
Algorithm Infrared simulation dataset Visible light dataset PSNR/dB SSIM PSNR/dB SSIM MSSN 31.630 0.877 17.086 0.932 MSSN+CBAM 31.661 0.877 17.103 0.931 MSSN+F-D 31.757 0.877 17.093 0.931 MSSANet 32.083 0.878 17.118 0.931
Tools
Get Citation
Copy Citation Text
Deyan ZHU, Yongqi AO, Jiayi XU, Chengcheng LI, Yufan ZHANG. Real-time deblurring of wideband small target based on attention mechanism[J]. Optics and Precision Engineering, 2024, 32(23): 3479
Paper Information
Category:
Received: Sep. 2, 2024
Accepted: --
Published Online: Mar. 10, 2025
The Author Email: Deyan ZHU (zdy_nuaa@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20