基于残差编解码网络的单光子压缩成像 
下载： 1699次
Single-Photon Compressive Imaging Based on Residual Codec Network
图 & 表
图 1. 单光子压缩成像实验光路图
Fig. 1. Experimental optical path of single photon compressive imaging
下载图片 查看原文
图 2. 用SRIED-Net进行训练和测试的过程
Fig. 2. Process of training and testing with SRIED-Net
下载图片 查看原文
图 3. 残差映射原理图
Fig. 3. Residual mapping schematic
下载图片 查看原文
图 4. 不同算法在不同测量率下的平均峰值信噪比比较
Fig. 4. Comparison of average SNRs of different algorithms at different MR
下载图片 查看原文
图 5. 单光子压缩成像重建结果
Fig. 5. Reconstruction results of single photon compressive imaging
下载图片 查看原文
图 6. 单光子压缩成像重建结果
Fig. 6. Reconstruction results of single photon compressive imaging
下载图片 查看原文
图 7. 使用TVAL3算法重建的单光子压缩成像。(a)使用随机高斯矩阵进行采样;(b)使用SRIED-Net的第一层全连接层作为测量矩阵进行采样
Fig. 7. Single photon compressive imaging using TVAL3 to reconstruct. (a) Sampling by random Gaussian matrix; (b) sampling by measurement matrix from the first FC layer in SRIED-Net
下载图片 查看原文
图 8. 使用不同测量矩阵采样后重建结果的峰值信噪比
Fig. 8. SNRs of reconstruction results after sampling with different measurement matrices
下载图片 查看原文
图 9. 使用不同测量矩阵采样后重建结果的结构相似性
Fig. 9. SIMs of reconstruction results after sampling with different measurement matrices
下载图片 查看原文
表 1不同算法在不同测量率下重建结果的峰值信噪比
Table1. SNR of reconstruction results of different algorithms at different MRdB
| Image name | Method | MR is 0.25 | MR is 0.10 | MR is 0.04 | MR is 0.01 |
|---|
| TVAL3 | 24.12 | 22.00 | 19.94 | 15.85 | | Barbara | AD_RE-Net | 22.51 | 21.31 | 19.67 | 18.94 | | SRIED-Net | 24.30 | 23.28 | 22.37 | 20.91 | | TVAL3 | 27.10 | 23.37 | 20.73 | 15.72 | | Parrot | AD_RE-Net | 22.99 | 21.72 | 19.68 | 18.79 | | SRIED-Net | 27.38 | 25.51 | 22.87 | 21.24 | | TVAL3 | 31.43 | 26.45 | 22.22 | 16.90 | | House | AD_RE-Net | 25.48 | 23.87 | 20.57 | 19.87 | | SRIED-Net | 29.97 | 27.93 | 24.93 | 22.17 | | TVAL3 | 28.56 | 23.84 | 20.40 | 15.32 | | Boats | AD_RE-Net | 23.70 | 22.11 | 19.19 | 18.58 | | SRIED-Net | 27.83 | 25.93 | 23.48 | 20.84 | | TVAL3 | 22.62 | 18.55 | 16.60 | 13.64 | | Fingerprint | AD_RE-Net | 18.87 | 17.74 | 15.73 | 15.46 | | SRIED-Net | 25.00 | 22.66 | 19.48 | 16.02 | | TVAL3 | 23.74 | 18.72 | 15.89 | 12.42 | | Flinstones | AD_RE-Net | 17.95 | 16.98 | 14.72 | 14.37 | | SRIED-Net | 24.03 | 21.69 | 18.75 | 16.01 | | TVAL3 | 25.73 | 21.77 | 18.63 | 16.06 | | Cameraman | AD_RE-Net | 20.69 | 19.64 | 18.08 | 17.48 | | SRIED-Net | 24.83 | 23.57 | 21.52 | 19.28 | | TVAL3 | 26.19 | 22.10 | 19.10 | 15.13 | | Mean PSNR | AD_RE-Net | 21.74 | 20.48 | 18.23 | 17.64 | | SRIED-Net | 25.92 | 24.37 | 21.91 | 19.50 |
|
查看原文
管焰秋, 鄢秋荣, 杨晟韬, 李冰, 曹芊芊, 方哲宇. 基于残差编解码网络的单光子压缩成像[J]. 光学学报, 2020, 40(1): 0111022. Yanqiu Guan, Qiurong Yan, Shengtao Yang, Bing Li, Qianqian Cao, Zheyu Fang. Single-Photon Compressive Imaging Based on Residual Codec Network[J]. Acta Optica Sinica, 2020, 40(1): 0111022.
PDF全文
