基于时间维超分辨率的视频快照压缩成像重构

陈赞,李冉,冯远静,李永强

Video snapshot compressive imaging reconstruction based on temporal super-resolution

Zan CHEN,Ran LI,Yuanjing FENG,Yongqiang LI

表 1 模拟数据重构结果和运行时间对比，单元格中左侧数据为PSNR，右侧数据为SSIM

Tab.1 Comparison results and running time of simulated data reconstruction, the left data of the cell is PSNR and the right is SSIM

对比方法 PSNR/dB, SSIM tr/s Kobe Traffic Runner Drop Aerial Crash 平均值 Tensor-FISTA[19] 25.02, 0.804 22.71, 0.822 30.32, 0.942 34.36, 0.971 25.95, 0.876 25.50, 0.891 27.31, 0.884 0.0166 E2E-CNN[17] 26.24, 0.820 24.53, 0.888 33.80, 0.974 36.66, 0.989 27.29, 0.915 26.30, 0.912 29.14, 0.917 0.0098 RevSCI[18] 27.51, 0.884 24.87, 0.898 34.05, 0.976 37.70, 0.990 26.97, 0.912 26.31, 0.915 29.57, 0.929 0.1412 ISTA-Rev-AE[29] 25.91, 0.811 24.09, 0.870 33.07, 0.977 38.05, 0.970 26.73, 0.903 26.18, 0.908 28.96, 0.909 0.0481 GAP-Unet-S12[20] 27.48, 0.856 25.55, 0.907 35.29, 0.980 37.18, 0.992 27.90, 0.924 26.83, 0.925 30.04, 0.931 0.0327 HQS-RevSCI[30] 27.59, 0.875 25.14, 0.902 34.30, 0.977 38.15, 0.990 27.27, 0.917 26.32, 0.914 29.79, 0.929 0.4136 FISTA-Rev-AE-3D[31] 27.58, 0.865 25.59, 0.909 35.03, 0.979 38.59, 0.991 27.58, 0.921 26.57, 0.915 30.16, 0.930 0.1281 SCI-OF[32] 29.03, 0.916 26.83, 0.933 35.32, 0.980 39.68, 0.992 28.07, 0.932 27.34, 0.939 31.04, 0.949 0.2476 EfficientSCI[33] 25.25, 0.826 22.65, 0.828 31.34, 0.962 35.51, 0.984 26.02, 0.890 25.52, 0.898 27.71, 0.898 0.0206 Res2former[34] 26.54, 0.858 24.32, 0.868 33.42, 0.973 37.90, 0.989 27.30, 0.916 26.58, 0.924 29.34, 0.921 0.0216 本文方法 28.87, 0.912 27.05, 0.935 36.29, 0.982 39.82, 0.993 28.29, 0.933 27.35, 0.941 31.27, 0.949 0.2509