Abstract

We propose an effective and robust method for terahertz (THz) image super-resolution based on a deep convolutional neural network (CNN). A deep CNN model is designed. It learns an end-to-end mapping between the low- and high-resolution images. Blur kernels with multiple width and noise with multiple levels are taken into the training set so that the network can handle THz images very well. Quantitative comparison of the proposed method and other super-resolution methods on the synthetic THz images indicates that the proposed method performs better than other methods in accuracy and visual improvements. Experimental results on real THz images show that the proposed method significantly improves the quality of THz images with increased resolution and decreased noise, which proves the practicability and exactitude of the proposed method.

© 2019 Optical Society of America

Super-resolution reconstruction of terahertz images based on a deep-learning network with a residual channel attention mechanism

Xiuwei Yang, Dehai Zhang, Zhongmin Wang, Yanbo Zhang, Jun Wu, Biyuan Wu, and Xiaohu Wu
Appl. Opt. 61(12) 3363-3370 (2022)

Adaptive terahertz image super-resolution with adjustable convolutional neural network

Yade Li, Weidong Hu, Xin Zhang, Zhihao Xu, Jiaqi Ni, and Leo P. Ligthart
Opt. Express 28(15) 22200-22217 (2020)

Polarized image super-resolution via a deep convolutional neural network

Haofeng Hu, Shiyao Yang, Xiaobo Li, Zhenzhou Cheng, Tiegen Liu, and Jingsheng Zhai
Opt. Express 31(5) 8535-8547 (2023)

Deblurring adaptive optics retinal images using deep convolutional neural networks

Xiao Fei, Junlei Zhao, Haoxin Zhao, Dai Yun, and Yudong Zhang
Biomed. Opt. Express 8(12) 5675-5687 (2017)

More realistic low-resolution OCT image generation approach for training deep neural networks

Hongming Pan, Di Yang, Zhuoqun Yuan, and Yanmei Liang
OSA Continuum 3(11) 3197-3205 (2020)