New residual neural network for rapid imaging of scattering

Fang Liu; Bochao Zhang; Baoxing Xiong; Fan Gao; Xiang Zhang; Xiao Yuan

doi:10.1117/12.2587308

2 December 2020 New residual neural network for rapid imaging of scattering

Fang Liu, Bochao Zhang, Baoxing Xiong, Fan Gao, Xiang Zhang, Xiao Yuan

Proceedings Volume 11717, 24th National Laser Conference & Fifteenth National Conference on Laser Technology and Optoelectronics; 117172G (2020) https://doi.org/10.1117/12.2587308
Event: 24th National Laser Conference & Fifteenth National Conference on Laser Technology and Optoelectronics, 2020, Shanghai, China

Abstract

There are various scattering media in nature. How to achieve rapid imaging through scattering media is an important issue in the field of optical imaging. In order to improve the speed of scattering imaging, this paper adopts a new type of residual neural network called XRNet, which mainly uses the advantages of easy optimization of residual neural network(ResNet) and combines U-Net structure and depthwise separable convolution instead of standard convolution to reduce the complexity of training model, thereby improving recovery speed of imaging. The image restoration speed and imaging quality under this model are simulated and calculated, and the convolutional neural network (CNN), ResNet and XRNet are analyzed by Pearson Correlation Coefficient (PCC), Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). The results show that XRNet has the fastest imaging speed when the imaging quality is basically unchanged. Finally, the imaging quality and recovery speed of XRNet at different depths are evaluated, and the best values are obtained at a depth of 31 layers.

Citation Download Citation

Fang Liu, Bochao Zhang, Baoxing Xiong, Fan Gao, Xiang Zhang, and Xiao Yuan "New residual neural network for rapid imaging of scattering", Proc. SPIE 11717, 24th National Laser Conference & Fifteenth National Conference on Laser Technology and Optoelectronics, 117172G (2 December 2020); https://doi.org/10.1117/12.2587308

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available