Deep Learning-Based Interference-Free Hologram Generation

Fig. 2. Hologram spectrum and reconstructed images with different sizes of filter windows^[11]. (a) Hologram spectrum and filter window; (b) reconstructed image with large size filter window; (c) reconstructed image with small size filter window

Download full size

Fig. 3. Deep learning-based interference-free hologram generation network

Download full size

Fig. 4. Simulated data sets. (a1)(a2) Original maps; (b1)(b2) holograms; (c1)(c2) reconstructed images; (d1)(d2) spectrograms

Download full size

Fig. 5. Simulation experiment data. (a) Original images; (b) holograms; (c) reconstructed images; (d) spectrograms; (e) interference-free holograms; (f) reconstructed images of interference-free holograms

Download full size

Fig. 6. Actual shooting light path diagram

Download full size

Fig. 7. Actual shooting of Fresnel hologram for verification. (a) Original object; (b) holograms taken actually; (c) spectrogram; (d) interference-free hologram; (e) reconstructed image

Download full size

Fig. 8. Loss function error curves. (a) RPN classification loss error curve; (b) RPN regression loss error curve;(c) Fast R-CNN classification loss error curve; (d) Fast R-CNN regression error curve; (e) total training error curve; (f) classification accuracy

Download full size

Fig. 9. Experimental procedure for artificial filtering hologram with interference

Download full size

Fig. 10. Images obtained by artificial filtering and HoloZL network. (a) Original images; (b) artificially filtered interference-free holograms; (c) reconstructed images of artificially filtered interference-free holograms; (d) interference-free holograms generated by HoloZL network; (e) reconstructed images of interference-free holograms generated by HoloZL network

Download full size

Table 1. Quality assessment metrics of interference-free hologram

View table

Table 1. Quality assessment metrics of interference-free hologram

Angle α /rad	$C_{1}$ （↑）	$C_{2}$ （↑）	MAE（↑）	PSNR（↑）	SSIM（↑）	Generated time $t_{G} (↓) / s$	Reconstruction time $t_{R} (↓) / s$
π/3.00	1.34	1.50	1.03	48.11	0.76	2.30	0.60
3π/4.00	1.29	1.85	0.57	48.11	0.76	2.20	0.74
π/4.00	1.30	1.89	0.77	48.10	0.76	2.44	0.69
π/2.02	1.31	1.49	1.44	48.10	0.75	1.95	0.56
π/2.00	1.29	1.59	0.59	48.11	0.76	2.00	0.61

Table 2. Quantitative assessment indexes of comparative experiments

View table

Table 2. Quantitative assessment indexes of comparative experiments

Data set		PSNR（↑）	SSIM（↑）	C（↑）	MAE（↑）	Generated time $t_{G} (↓) / s$	Reconstruction time $t_{R} (↓) / s$
Mnist	Artificial filtering	48.00	0.76	1.36	0.83	2.01	0.41
Mnist	HoloZL	49.70	0.77	1.39	0.77	1.93	0.33
Guang	Artificial filtering	27.82	0.91	2.98	1.32	1.93	0.54
Guang	HoloZL	28.35	0.95	2.98	1.24	1.71	0.66
Wukong	Artificial filtering	17.60	0.82	2.72	6.12	1.89	0.41
Wukong	HoloZL	18.42	0.82	2.73	5.73	1.33	0.34
Lianpu	Artificial filtering	18.82	0.80	2.99	7.76	1.61	0.75
Lianpu	HoloZL	19.31	0.81	2.99	7.76	1.60	0.53

Tools

Get Citation

Copy Citation Text

Jiaxue Wu, Jinbin Gui, Junchang Li, Tai Fu, Wei Cheng. Deep Learning-Based Interference-Free Hologram Generation[J]. Acta Optica Sinica, 2022, 42(14): 1409001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Holography

Received: Dec. 14, 2021

Accepted: Feb. 21, 2022

Published Online: Jul. 15, 2022

The Author Email: Gui Jinbin (jinbingui@163.com)

DOI:10.3788/AOS202242.1409001

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology