Dynamic Light Field Reconstruction Based on Gradient Descent Deep Equilibrium Model (Invited)

Ruixue Wang; Xue Wang; Guoqing Zhou; Zhaolin Xiao; Qing Wang

doi:10.3788/LOP241400

Laser & Optoelectronics Progress, Volume. 61, Issue 16, 1611006(2024)

Dynamic Light Field Reconstruction Based on Gradient Descent Deep Equilibrium Model (Invited)

Ruixue Wang¹, Xue Wang¹, Guoqing Zhou¹, Zhaolin Xiao², and Qing Wang^1、*

Author Affiliations

¹School of Computer Science, Northwestern Polytechnical University, Xi'an 710072, Shaanxi, China

²School of Computer Science and Engineering, Xi'an University of Technology, Xi'an 710048, Shaanxi, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Flowchart of dynamic light field temporal domain SCI method

Download full size

Fig. 2. Flowchart of DEQ-GD model

Download full size

Fig. 3. Construction of convolutional neural network for regularized gradient term

Download full size

Fig. 4. Comparison of reconstructed sub-aperture images on five test scenes by different methods. (a) Gound truth images; (b) proposed method; (c) method in Ref. [27]

Download full size

Table 1. PSNR and SSIM of reconstructed sub-aperture images on five test scenes by different methods

View table

Table 1. PSNR and SSIM of reconstructed sub-aperture images on five test scenes by different methods

Test scene	Method in Ref.［27］		Proposed method
Test scene	PSNR /dB	SSIM	PSNR /dB	SSIM
Average	23.83	0.656	27.36	0.848
ambushfight_5	24.03	0.719	25.38	0.825
chickenrun_1	20.30	0.565	23.33	0.776
foggyrocks_1	24.33	0.731	25.10	0.836
questbegins_1	27.70	0.671	35.71	0.961
thebigfight_2	22.80	0.594	27.29	0.843

Table 2. PSNR and SSIM of reconstructed monocular video on five test scenes by different methods

View table

Table 2. PSNR and SSIM of reconstructed monocular video on five test scenes by different methods

Test scene	Method in Ref.［40］		Proposed method
Test scene	PSNR /dB	SSIM	PSNR /dB	SSIM
Average	29.54	0.953	29.62	0.954
ambushfight_5	29.86	0.978	29.69	0.978
chickenrun_1	25.88	0.932	25.99	0.933
foggyrocks_1	26.16	0.949	26.15	0.949
questbegins_1	37.71	0.979	38.05	0.981
thebigfight_2	28.09	0.929	28.21	0.931

Table 3. Average PSNR and SSIM of five test scenes under different number of feature interaction layers
View table
Table 3. Average PSNR and SSIM of five test scenes under different number of feature interaction layers
Number of feature interaction layer PSNR /dB SSIM
1 28.67 0.883
3 28.95 0.889
5 18.97 0.463

Table 4. Average PSNR and SSIM of five test scenes under different resolutions
View table
Table 4. Average PSNR and SSIM of five test scenes under different resolutions
Resolution PSNR /dB SSIM
3×3×128 pixel×128 pixel 28.95 0.889
5×5×128 pixel×128 pixel 27.36 0.848

Tools

Get Citation

Copy Citation Text

Ruixue Wang, Xue Wang, Guoqing Zhou, Zhaolin Xiao, Qing Wang. Dynamic Light Field Reconstruction Based on Gradient Descent Deep Equilibrium Model (Invited)[J]. Laser & Optoelectronics Progress, 2024, 61(16): 1611006

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Imaging Systems

Received: May. 31, 2024

Accepted: Jul. 15, 2024

Published Online: Aug. 12, 2024

The Author Email: Qing Wang (qwang@nwpu.edu.cn)

DOI:10.3788/LOP241400

CSTR:32186.14.LOP241400

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. PSNR and SSIM of reconstructed sub-aperture images on five test scenes by different methods

Table 1. PSNR and SSIM of reconstructed sub-aperture images on five test scenes by different methods

Table 2. PSNR and SSIM of reconstructed monocular video on five test scenes by different methods

Table 2. PSNR and SSIM of reconstructed monocular video on five test scenes by different methods

Table 3. Average PSNR and SSIM of five test scenes under different number of feature interaction layers

Table 3. Average PSNR and SSIM of five test scenes under different number of feature interaction layers

Table 4. Average PSNR and SSIM of five test scenes under different resolutions

Table 4. Average PSNR and SSIM of five test scenes under different resolutions