Adaptive Construction Method for Binary Measurement Matrix Based on Deep Learning

Fig. 4. Simulation results of different measurement matrices. (a)--(f) Results of random Gaussian matrix; (g)--(l) results of Toeplitz matrix; (m)--(r) results of Hadamard matrix; (s)--(x) results of proposed matrix

Download full size

Fig. 5. Ghost imaging system structure based on compressed sensing. (a) Principle diagram; (b) picture

Download full size

Fig. 6. Target images of laser imaging experiment. (a) Target1; (b) Target2

Download full size

Fig. 7. Results of laser imaging under different sampling rates. (a)--(f) Results of random Gaussian matrix; (g)--(l) results of Toeplitz matrix; (m)--(r) results of Hadamard matrix; (s)--(x) results of proposed matrix

Download full size

Table 1. PSNR results of different measurement matrices under different sampling rates
View table
Table 1. PSNR results of different measurement matrices under different sampling rates
Matrix SR=0.1 SR=0.2 SR=0.5
Ball Face Ball Face Ball Face
GS 14.04 16.08 17.45 19.98 31.71 28.95
TP 18.14 15.75 22.95 18.85 32.82 24.45
HM 29.25 23.25 31.68 27.48 37.82 33.28
Proposed 31.11 27.16 34.20 29.61 39.50 33.93

Table 2. SSIM results of different measurement matrices under different sampling rates
View table
Table 2. SSIM results of different measurement matrices under different sampling rates
Matrix SR=0.1 SR=0.2 SR=0.5
Ball Face Ball Face Ball Face
GS 0.205 0.311 0.254 0.450 0.766 0.796
TP 0.352 0.315 0.478 0.446 0.867 0.717
HM 0.878 0.646 0.910 0.795 0.977 0.941
Proposed 0.917 0.696 0.953 0.867 0.982 0.950

Table 3. PSNR of laser imaging results under different measurement matrices
View table
Table 3. PSNR of laser imaging results under different measurement matrices
Matrix SR=0.1 SR=0.2 SR=0.5
Target1 Target2 Target1 Target2 Target1 Target2
GS 9.28 8.53 11.84 10.92 10.36 11.98
TP 8.91 8.46 10.32 9.81 10.06 11.18
HM 13.33 12.97 13.49 12.88 13.18 12.34
Proposed 14.97 12.89 14.26 15.14 15.03 14.88

Table 4. SSIM of laser imaging results under different measurement matrices
View table
Table 4. SSIM of laser imaging results under different measurement matrices
Matrix SR=0.1 SR=0.2 SR=0.5
Target1 Target2 Target1 Target2 Target1 Target2
GS 0.089 0.080 0.138 0.113 0.114 0.128
TP 0.083 0.065 0.103 0.087 0.116 0.126
HM 0.181 0.115 0.181 0.136 0.165 0.114
Proposed 0.316 0.242 0.295 0.284 0.322 0.237

Tools

Get Citation

Copy Citation Text

Jiefei Han, Bobo Lian, Liying Sun. Adaptive Construction Method for Binary Measurement Matrix Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2220001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Optics in Computing

Received: May. 6, 2021

Accepted: May. 18, 2021

Published Online: Nov. 10, 2021

The Author Email: Liying Sun (kebersun@163.com)

DOI:10.3788/LOP202158.2220001

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. PSNR results of different measurement matrices under different sampling rates

Table 1. PSNR results of different measurement matrices under different sampling rates

Table 2. SSIM results of different measurement matrices under different sampling rates

Table 2. SSIM results of different measurement matrices under different sampling rates

Table 3. PSNR of laser imaging results under different measurement matrices

Table 3. PSNR of laser imaging results under different measurement matrices

Table 4. SSIM of laser imaging results under different measurement matrices

Table 4. SSIM of laser imaging results under different measurement matrices