Journal of Applied Optics, Volume. 46, Issue 3, 632(2025)
Colorization of grayscale images combined with semantic segmentation
Figures & Tables(14)
Fig. 4. Comparison of colorization effects of different network structures
Fig. 5. Comparison of colorization effects of segmented networks with and without attention mechanisms
Fig. 6. Comparison of colorization effects with different loss functions
Fig. 7. Physical drawing of multi-pixel photon counter experimental platform
Fig. 8. Comparison of LLL image effects of different colorization methods
Table 1. Comparison of image quality metrics (PSNR and SSIM) for different colorization methods
View tableView in ArticleTable 1. Comparison of image quality metrics (PSNR and SSIM) for different colorization methods
图像 LIZUKA S等人 LARSSON G等人 ANTIC J等人 KANG X等人 VITORIA P等人 本文方法 PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM PSNR SSIM orange 17.868 0.778 20.027 0.751 20.803 0.767 20.335 0.608 22.482 0.781 23.231↑ 0.829 ↑grassland 21.166 0.899 21.789 0.947 27.990 ↑0.959 26.95 0.960 26.822 0.895 27.145 0.963 ↑desktop 16.555 0.808 17.589 0.811 17.772 0.767 17.634 0.787 19.734 0.816 20.818 ↑0.891 ↑sheep 27.118 0.918 25.506 0.932 27.001 0.948 26.044 0.901 29.985 0.952 30.343 ↑0.960 ↑pavement 26.429 0.908 23.613 0.940 27.681 0.953 27.504 0.895 30.135 0.965 32.386 ↑0.973 ↑zebra 19.586 0.894 21.227 0.951 23.729 0.895 27.595 0.967 ↑28.583 ↑0.959 27.372 0.964
Table 2. Comparison of image quality metrics (LPIPS and IE) for different colorization methods
View tableView in ArticleTable 2. Comparison of image quality metrics (LPIPS and IE) for different colorization methods
图像 LIZUKA S等人 LARSSON G等人 ANTIC J等人 KANG X等人 VITORIA P等人 本文方法 LPIPS IE LPIPS IE LPIPS IE LPIPS IE LPIPS IE LPIPS IE orange 0.152 7.134 0.188 6.979 0.169 7.248 0.190 6.873 0.126 7.286 ↑0.106 ↓6.922 grassland 0.074 7.218 0.057 7.269 0.065 7.285 0.069 7.144 0.052 7.237 0.051 ↓7.305 ↑desktop 0.139 7.428 0.115 7.372 0.101 7.191 0.106 7.459 0.129 7.502 0.079 ↓7.515 ↑sheep 0.063 7.635 0.096 7.625 0.059 ↓7.667 0.095 7.619 0.062 7.671 ↑0.060 7.641 pavement 0.156 7.682 0.162 7.676 0.073 7.722 0.123 7.671 0.075 7.639 0.051 ↓7.694 ↑zebra 0.093 7.215 0.067 7.189 0.066 7.173 0.079 7.222 0.061 7.261 0.059 ↓7.287 ↑
Table 3. Image quality metrics for experiments with different network structures
View tableView in ArticleTable 3. Image quality metrics for experiments with different network structures
方法 PSNR SSIM LPIPS IE Baseline 22.731 0.894 0.138 7.422 +分类子网络 25.042 0.922 0.133 7.415 +分割子网络 25.949 0.931 0.128 7.494 本文方法 26.793 ↑0.938 ↑0.073 ↓7.579 ↑
Table 4. Image quality metrics for experiments with and without attention mechanisms
View tableView in ArticleTable 4. Image quality metrics for experiments with and without attention mechanisms
方法 PSNR SSIM LPIPS IE 无注意力机制 25.652 0.915 0.132 7.458 本文方法 26.793 ↑0.938 ↑0.125 ↓7.579 ↑
Table 5. Image quality metrics for experiments with different loss functions
View tableView in ArticleTable 5. Image quality metrics for experiments with different loss functions
方法 PSNR SSIM LPIPS IE Baseline 25.853 0.898 0.088 7.136 + Lcls 26.343 0.905 0.084 7.242 + Lseg 26.291 0.913 0.079 7.307 本文方法 26.793 ↑0.938 ↑0.073 ↓7.579 ↑
Table 6. Quality metrics of LLL images by different colorization methods
View tableView in ArticleTable 6. Quality metrics of LLL images by different colorization methods
方法 PSNR SSIM LPIPS IE LARSSON G等人 10.847 0.207 0.747 7.332 ANTIC J等人 10.550 0.206 0.739 7.332 VITORIA P等人 10.762 0.212 0.737 7.337 本文方法 11.194 ↑0.236 ↑0.709 ↓7.408 ↑
Tools
Get Citation
Copy Citation Text
Xiaoning GAO, Liju YIN, Feng WANG, Yulin DENG, Yiming QIN. Colorization of grayscale images combined with semantic segmentation[J]. Journal of Applied Optics, 2025, 46(3): 632
Paper Information
Category:
Received: Apr. 22, 2024
Accepted: --
Published Online: May. 28, 2025
The Author Email: Liju YIN (尹丽菊)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20