Researching | Passive compressive ghost imaging with low rank clustering

Optics and Precision Engineering, Volume. 32, Issue 1, 12(2024)

Passive compressive ghost imaging with low rank clustering

Teng LEI^1,2,3, Yiming ZHANG^1,2,3, Yizhe MA^1,2,3, Xuezhuan DING^1,2,3, Yingyue WU^1,2,3, and Shiyong WANG^1,2,3、*

Author Affiliations

¹Shanghai Institute of Technical Physics， Chinese Academy of Sciences， Shanghai200083， China

²University of Chinese Academy of Sciences， Beijing100049， China

³Laboratory of Infrared Detection and Imaging Technology， Chinese Academy of Sciences， Shanghai20008， China

show less

Abstract Get PDF(in Chinese)

Figures&Tables (10)

References (33)

Paper Information

Figures & Tables(10)

Fig. 1. Principle for ghost imaging camera

Download full size

View in Article

Fig. 2. Schematic diagram of light field modulation

Download full size

View in Article

Fig. 3. Flow chart of reconstruction algorithm with joint iteration

Download full size

View in Article

Fig. 4. Simulation results of gray images

Download full size

View in Article

Fig. 5. MSE variation curve of PCGI-LRC under different iterations at sampling rate of 0.25

Download full size

View in Article

Fig. 6. Experimental optical path of passive compressed ghost imaging

Download full size

View in Article

Fig. 7. Target image and its scattering field

Download full size

View in Article

Fig. 8. Comparison of imaging effects of different algorithms

Download full size

View in Article

Table 1. [in Chinese]

View in Article

Table 1. [in Chinese]

算法1：低秩聚类被动压缩鬼成像算法

输入：测量值 $Y$ ，测量矩阵 $Φ$ ，最大迭代次数 $K$ ，迭代终止阈值 $ε$

1：初始化： $X_{0} = 0$

2：令 $i = 1 : K$

3：采用共轭梯度法求解最小二乘问题，获得预重

构图像 ${\hat{X}}^{(k + 1)}$ ：

${\hat{X}}^{(k + 1)} = a r g \underset{{\hat{X}}^{(k)}}{m i n} 1 / 2 {\hat{X}}^{(k)}^{T} Φ^{T} Φ {\hat{X}}^{(k)} - {\hat{X}}^{(k)}^{T} Φ^{T} y$

4：对图像 ${\hat{X}}^{(k + 1)}$ 进行分块得到局部块 ${\hat{x}}_{l}^{(k + 1)}$

5：寻找局部块 ${\hat{x}}_{l}^{(k + 1)}$ 的相似块矩阵组 ${\hat{X}}_{l}^{(k + 1)}$

6：对每个有噪声的数据矩阵 ${\hat{X}}_{l}^{(k + 1)}$ 进行奇异值分解： $(U_{l}^{(k + 1)}, Σ_{l}^{(k + 1)}, V_{l}^{(k + 1)}) = S V D ({\hat{X}}_{l}^{(k + 1)})$

7：利用广义软阈值算法进行低秩矩阵近似

8：图像更新，对所有块进行加权平均，得到噪声

抑制后的图像 $q^{(k + 1)}$

9：若 $(q^{(k + 1)} - {\hat{X}}^{(k)}) / {\hat{X}}^{(k)} < ε$ ，则停止循环，设置 $X = q^{(k + 1)}$ ；否则设置 ${\hat{X}}^{(k)} = q^{(k + 1)}$ ，联合迭代继续

10：若迭代次数 $i > K$ ，则停止循环

输出：重构图像 $X$

Table 1. Comparison of reconstruction effects of different ghost imaging algorithms

View in Article

Table 1. Comparison of reconstruction effects of different ghost imaging algorithms

Algorithm	Rate=0.062 5			Rate=0.125			Rate=0.25			Rate=0.5
Algorithm	SSIM	PSNR/dB	MSE	SSIM	PSNR/dB	MSE	SSIM	PSNR/dB	MSE	SSIM	PSNR/dB	MSE
GI-SBC（OMP）	0.118 4	16.396 0	0.149 1	0.138 9	16.892 5	0.133 0	0.141 3	17.012 1	0.128 7	0.147 2	17.863 2	0.106 3
GI-SBC（FISTA）	0.079 0	14.572 2	0.226 9	0.092 6	15.535 3	0.181 8	0.105 6	16.448 2	0.147 3	0.123 3	17.752 4	0.109 1
GI-TV（TVAL3）	0.091 7	17.247 6	0.122 6	0.092 4	17.388 6	0.118 6	0.104 6	17.673 9	0.110 9	0.106 0	17.284 9	0.121 5
GI-TV（GISCNL）	0.104 1	17.662 1	0.111 4	0.113 4	17.731 0	0.109 6	0.115 0	17.950 1	0.104 3	0.115 8	18.026 8	0.102 4
PCGI-LRC	0.147 2	18.751 5	0.086 8	0.152 6	18.784 9	0.086 0	0.156 3	19.178 4	0.078 6	0.153 5	19.164 6	0.078 8

Tools

Get Citation

Copy Citation Text

Teng LEI, Yiming ZHANG, Yizhe MA, Xuezhuan DING, Yingyue WU, Shiyong WANG. Passive compressive ghost imaging with low rank clustering[J]. Optics and Precision Engineering, 2024, 32(1): 12

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Jun. 26, 2023

Accepted: --

Published Online: Jan. 23, 2024

The Author Email: Shiyong WANG (s_y_w@sina.com)

DOI:10.37188/OPE.20243201.0012

Topics

laser devices and laser physics

Lasers and Laser Optics

laser manufacturing

Instrumentation, Measurement and Metrology