Journal of Infrared and Millimeter Waves, Volume. 43, Issue 4, 541(2024)
Terahertz imaging super-resolution algorithm based on Hilbert spatial curve filling
Figures & Tables(15)
Fig. 1. Network structure diagram of terahertz image super-resolution based on curve filling in Hilbert space
Fig. 5. Schematic diagram of the experimental environment，（a） experimental equipment； （b） image acquisition
Fig. 6. Hilbert space curve layer training and test loss comparison diagram
Fig. 8. Comparison of peak signal-to-noise ratio of different models
Fig. 9. Structure similarity comparison diagram of different models
Fig. 10. Superresolution reconstruction comparison of terahertz images of different models，（a） original low-resolution image； （b） ViT； （c） ViT without Hilbert kernel size=3； （d） ViT without Hilbert kernel size=5； （e） ViT without Hilbert kernel size=7； （f） ViT with Hilbert kernel size=3； （g） ViT with Hilbert kernel size=5； （h） ViT with Hilbert kernel size=7； （i） real high-resolution image
Fig. 11. Comparison diagram of Hilbert transform effect on different frequency imaging methods
Table 1. Comparison of neighborhood retention rates of different curves
View tableView in ArticleTable 1. Comparison of neighborhood retention rates of different curves
8-Nbr 24-Nbr 48-Nbr Z curve 2.132 3 3.193 3 4.409 0 Peano curve 2.176 4 3.462 9 4.274 3 Hilbert curve 1.815 9 2.836 5 3.979 0
Table 2. Super-resolution image quality evaluation index
View tableView in ArticleTable 2. Super-resolution image quality evaluation index
Different algorithm MSE PSNR/dB SSIM ViT 1.251 47.108 7 0.946 3 ViT without Hilbert kernel size=3 1.331 4 46.892 9 0.952 1 ViT without Hilbert kernel size=5 1.213 47.102 6 0.947 5 ViT without Hilbert kernel size=7 1.254 5 47.149 8 0.949 6 ViT with Hilbert kernel size=3 1.113 5 47.564 2 0.952 4 ViT with Hilbert kernel size=5 1.123 5 47.566 2 0.953 7 ViT with Hilbert kernel size=7 0.998 2 47.922 6 0.947 8
Table 3. Objective evaluation index of super resolution image quality based on different algorithms
View tableView in ArticleTable 3. Objective evaluation index of super resolution image quality based on different algorithms
Different algorithm MSE PSNR/dB SSIM Nearest 25.793 6 34.015 7 0.868 3 Bicubic 27.339 4 33.762 9 0.898 SRCNN
SRResNet
14.322 4
1.489 2
36.570 6
46.401 3
0.901 7
0.924 1
SRGAN 1.592 3 46.110 6 0.956 2 ESPCNN 1.412 84 46.629 9 0.935 1 ViT 1.251 0 47.108 7 0.946 3 MDM 1.148 2 47.530 6 0.939 9 ViT with Hilbert kernel size=7 0.998 2 47.922 6 0.947 8 HR 0 ∞ 1
Tools
Get Citation
Copy Citation Text
Mo-Xuan YANG, Yuan-Meng ZHAO, Hao-Xin LIU, Yi LIU, You WU, Cun-Lin ZHANG. Terahertz imaging super-resolution algorithm based on Hilbert spatial curve filling[J]. Journal of Infrared and Millimeter Waves, 2024, 43(4): 541
Paper Information
Category: Research Articles
Received: Oct. 14, 2023
Accepted: --
Published Online: Aug. 27, 2024
The Author Email: Yuan-Meng ZHAO (zhao.yuanmeng@cnu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20