End-to-End Phase Reconstruction of Digital Holography Based on Improved Residual Unet

Kunge Li; Huaying Wang; Xu Liu; Jieyu Wang; Wenjian Wang; Liu Yang

doi:10.3788/LOP220881

Laser & Optoelectronics Progress, Volume. 60, Issue 6, 0610016(2023)

End-to-End Phase Reconstruction of Digital Holography Based on Improved Residual Unet

Kunge Li¹, Huaying Wang^1,2,3, Xu Liu^1,2,3、*, Jieyu Wang¹, Wenjian Wang¹, and Liu Yang¹

¹School of Mathematics and Physics Science and Engineering, Hebei University of Engineering, Handan 056038, Hebei, China

²Hebei Computational Optical Imaging and Photoelectric Detection Technology Innovation Center, Handan 056038, Hebei, China

³Hebei International Joint Research Center for Computational Optical Imaging and Intelligent Sensing, Handan 056038, Hebei, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Experimental setup for digital holographic recording optical path. (a) Optical path diagram for loading blood cell data; (b) optical path diagram for loading handwrite number data

Download full size

Fig. 2. Dataset examples. (a) Sample hologram; (b) sample phase

Download full size

Fig. 3. Network structure of Udanet

Download full size

Fig. 4. Schematic of different attention mechanisms^[31]. (a) Channel attention; (b) frequency attention

Download full size

Fig. 5. Residual block comparison. (a) DDA residual block; (b) ordinary residual block; (c) attention residual block

Download full size

Fig. 6. Loss function curve

Download full size

Fig. 7. Phase reconstruction of the blood cells. (a) Ground truth; (b) reconstruction result of Udanet; (c) three-dimensional image of error map between ground truth and Udanet reconstruction result

Download full size

Fig. 8. Phase reconstruction of the handwrite number dataset. (a) Ground truth; (b) reconstruction result of Udanet;(c) three-dimensional image of error map between ground truth and Udanet reconstruction result

Download full size

Table 1. Network parameter setting
View table
Table 1. Network parameter setting
Parameter Udanet Unet Uanet
Batchsize 32 32 32
Learning-rate 1×10^-3 1×10^-3 1×10^-3
Epoch 200 200 200
Number of images on training set 26000，33000 26000 26000
Block parameter Lowk，Topk

Table 2. SSIM value of Udanet with different frequency components
View table
Table 2. SSIM value of Udanet with different frequency components
Parameter Low8 Top1 Top2 Top4 Top8
SSIM average value 0.861 0.855 0.853 0.868 0.848
SSIM peak value 0.914 0.915 0.910 0.922 0.907

Table 3. SSIM value of different networks
View table
Table 3. SSIM value of different networks
Parameter Unet Uanet Udanet（Top4）
SSIM average value 0.844 0.851 0.868
SSIM peak value 0.901 0.913 0.922

Tools

Get Citation

Copy Citation Text

Kunge Li, Huaying Wang, Xu Liu, Jieyu Wang, Wenjian Wang, Liu Yang. End-to-End Phase Reconstruction of Digital Holography Based on Improved Residual Unet[J]. Laser & Optoelectronics Progress, 2023, 60(6): 0610016

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites