基于稀疏先验的空间目标图像盲反演方法
Sparse Prior-based Space Objects Image Blind Inversion Algorithm
1 重庆大学 微电子与通信工程学院, 重庆 400044
2 重庆大学 信息物理社会可信服务计算教育部重点实验室, 重庆 400044
3 中国科学院光电技术研究所, 成都 610209
4 中国科学院光束控制重点实验室, 成都 610209
图 & 表
图 1. Intensity histogram and gradient histogram images of space object in deep space background
Fig. 1. Intensity histogram and gradient histogram images of space object in deep space background
下载图片 查看原文
图 2. Intensity histogram and gradient histogram images of space object at different exposure levels
Fig. 2. Intensity histogram and gradient histogram images of space object at different exposure levels
下载图片 查看原文
图 3. Intensity histogram and gradient histogram images of space object in ground background
Fig. 3. Intensity histogram and gradient histogram images of space object in ground background
下载图片 查看原文
图 4. Fitting the gradient distribution of space object image with each prior
Fig. 4. Fitting the gradient distribution of space object image with each prior
下载图片 查看原文
图 5. Sparse representation of space object image
Fig. 5. Sparse representation of space object image
下载图片 查看原文
图 6. 海事卫星图像高斯退化反演结果比较Compare of inversion results of Gaussian degraded maritime satellite image
Fig. 6. 海事卫星图像高斯退化反演结果比较Compare of inversion results of Gaussian degraded maritime satellite image
下载图片 查看原文
图 7. 空间站图像高斯退化反演结果比较Compare of inversion results of Gaussian degraded space station image
Fig. 7. 空间站图像高斯退化反演结果比较Compare of inversion results of Gaussian degraded space station image
下载图片 查看原文
图 8. 海事卫星图像运动退化反演结果比较Compare of inversion results of motion degraded maritime satellite image
Fig. 8. 海事卫星图像运动退化反演结果比较Compare of inversion results of motion degraded maritime satellite image
下载图片 查看原文
图 9. 空间站图像运动退化反演结果比较Compare of inversion results of motion degraded space station image
Fig. 9. 空间站图像运动退化反演结果比较Compare of inversion results of motion degraded space station image
下载图片 查看原文
图 10. 月球观测图像反演结果比较Compare of inversion results of lunar observation image
Fig. 10. 月球观测图像反演结果比较Compare of inversion results of lunar observation image
下载图片 查看原文
图 11. 土星退化图像反演结果比较Compare of inversion results of real Saturn degraded image
Fig. 11. 土星退化图像反演结果比较Compare of inversion results of real Saturn degraded image
下载图片 查看原文
表 1高斯退化海事卫星图像反演结果的SSIM和GMG
Table1. SSIM and GMG of the inversion results of Gaussian degraded maritime satellite image
Standard deviation of Gaussian blur kernel | SSIM/GMG | Krishnan’s | Zhang’s | Perrone’s | Pan’s | Lin’s | Our method | σ=1 | 0.868/2.369 | 0.862/3.445 | 0.857/3.580 | 0.931/4.001 | 0.890/4.133 | 0.938/3.542 | σ=2 | 0.868/1.780 | 0.873/2.391 | 0.864/2.314 | 0.923/2.362 | 0.917/2.612 | 0.928/2.632 | σ=3 | 0.884/1.505 | 0.866/2.015 | 0.865/1.639 | 0.906/2.078 | 0.901/2.053 | 0.918/2.264 | σ=4 | 0.878/1.423 | 0.869/1.908 | 0.864/1.670 | 0.879/1.877 | 0.890/1.997 | 0.914/2.116 | σ=5 | 0.874/1.393 | 0.845/1.886 | 0.863/1.787 | 0.857/1.854 | 0.904/2.036 | 0.902/1.894 |
|
查看原文
表 2高斯退化空间站图像反演结果的SSIM和GMG
Table2. SSIM and GMG of the inversion results of Gaussian degraded space station image
Standard deviation of Gaussian blur kernel | SSIM/GMG | Krishnan’s | Zhang’s | Perrone’s | Pan’s | Lin’s | Our method | σ=1 | 0.873/8.617 | 0.776/8.656 | 0.515/9.442 | 0.903/9.348 | 0.926/9.784 | 0.947/9.981 | σ=2 | 0.795/4.074 | 0.808/5.399 | 0.565/4.940 | 0.833/4.835 | 0.849/5.558 | 0.857/5.620 | σ=3 | 0.734/3.640 | 0.710/4.734 | 0.558/4.200 | 0.772/4.654 | 0.801/4.811 | 0.824/4.665 | σ=4 | 0.695/5.195 | 0.757/4.397 | 0.549/4.766 | 0.652/5.829 | 0.829/5.121 | 0.806/4.746 | σ=5 | 0.674/5.013 | 0.685/4.483 | 0.548/5.069 | 0.636/6.383 | 0.824/5.005 | 0.834/4.955 |
|
查看原文
表 3运动退化海事卫星图像反演结果的SSIM和GMG
Table3. SSIM and GMG of the inversion results of motion degraded maritime satellite image
Blur scale of motion blur | SSIM/GMG | Krishnan’s | Zhang’s | Perrone’s | Pan’s | Lin’s | Our method | 9 | 0.859/3.934 | 0.885/3.532 | 0.841/3.706 | 0.891/4.126 | 0.936/4.063 | 0.951/3.983 | 13 | 0.875/3.369 | 0.863/3.187 | 0.856/3.163 | 0.850/3.711 | 0.928/3.627 | 0.935/3.814 | 17 | 0.876/3.617 | 0.859/3.059 | 0.853/2.998 | 0.822/3.434 | 0.916/3.561 | 0.921/3.729 | 21 | 0.820/3.143 | 0.852/2.916 | 0.854/2.953 | 0.799/3.455 | 0.913/3.485 | 0.915/3.492 | 25 | 0.815/3.248 | 0.846/2.713 | 0.853/2.652 | 0.814/3.179 | 0.904/3.259 | 0.898/3.310 |
|
查看原文
表 4运动退化空间站图像反演结果的SSIM和GMG
Table4. SSIM and GMG of the inversion results of motion degraded space station image
Standard deviation of Gaussian blur kernel | SSIM/GMG | Krishnan’s | Zhang’s | Perrone’s | Pan’s | Lin’s | Our method | 9 | 0.719/5.480 | 0.799/7.227 | 0.515/8.104 | 0.652/8.131 | 0.842/8.761 | 0.869/8.794 | 13 | 0.649/4.819 | 0.763/6.490 | 0.520/8.190 | 0.627/7.942 | 0.807/8.175 | 0.844/8.203 | 17 | 0.615/6.608 | 0.752/5.568 | 0.531/6.862 | 0.550/7.770 | 0.836/6.842 | 0.850/6.711 | 21 | 0.567/5.967 | 0.741/5.402 | 0.546/6.821 | 0.495/7.108 | 0.847/6.859 | 0.851/6.617 | 25 | 0.543/5.975 | 0.726/4.889 | 0.547/6.000 | 0.493/5.859 | 0.813/5.758 | 0.836/6.294 |
|
查看原文
表 5真实空间目标反演图像客观评价结果
Table5. Objective evaluation results of the estimated real space object image
Degraded image | Evaluation indexes | Krishnan’s | Zhang’s | Perrone’s | Pan’s | Lin’s | Ours | lunar image | GMG | 2.8182.845 | 2.8202.685 | 2.8643.264 | 2.8092.580 | 2.8493.110 | 2.8973.623 | Saturn image | GMG | 3.719 22.924 9 | 3.648 52.698 5 | 3.742 93.012 4 | 3.530 32.407 90 | 3.644 92.700 8 | 3.729 84.328 6 |
|
查看原文
李正周, 卿琳, 李博, 陈成, 亓波. 基于稀疏先验的空间目标图像盲反演方法[J]. 光子学报, 2020, 49(2): 0210001. Zheng-zhou LI, Lin QING, Bo LI, Cheng CHEN, Bo QI. Sparse Prior-based Space Objects Image Blind Inversion Algorithm[J]. ACTA PHOTONICA SINICA, 2020, 49(2): 0210001.