基于深度学习的超分辨率图像重建研究综述  

A Review of Research on Super Resolution Image Reconstruction Based on Deep Learning

作　　者：郑璐 王保云[1,2] 杨昆[1] 孔艳[1,2] 张祝鸿 张玲莉 ZHENG Lu;WANG Baoyun;YANG Kun;KONG Yan;ZHANG Zhuhong;ZHANG Lingli(School of Information,Yunnan Normal University,Kunming 650500;Laboratory of Data Science,School of Information,Yunnan Normal University,Kunming 650500)

机构地区：[1]云南师范大学信息学院,昆明650500 [2]云南师范大学信息学院数据科学实验室,昆明650500

出　　处：《计算机与数字工程》2021年第4期804-808,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:41461038)资助。

摘　　要：目前超分辨率图像重建技术是计算机视觉领域的研究热点,随着深度学习的发展,基于深度学习的超分辨率图像重建技术已经取得了一定的研究成果。论文回顾了典型的超分辨率图像重建的深度网络模型,对超分辨率图像重建的深度学习算法和网络结构进行介绍,比较分析了不同模型的优缺点,从本质上发现并提出了超分辨率图像重建的一些问题。在此基础上,提出了基于深度学习的超分辨率图像重建方法未来的研究趋势。At present,super-resolution image reconstruction technology is a research hotspot in the field of computer vision.With the development of deep learning,super-resolution image reconstruction technology based on deep learning has achieved cer⁃tain research results.This paper reviews the deep network model of typical super-resolution image reconstruction,introduces the deep learning algorithm and network structure of super-resolution image reconstruction,compares the advantages and disadvantages of different models,essentially discovers and proposes some problems in super-resolution image reconstruction.On this basis,the future research trend of super-resolution image reconstruction based on deep learning is proposed.

关 键 词：超分辨率重建 深度学习 卷积神经网络 生成对抗网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]