Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (22): 28-41.DOI: 10.3778/j.issn.1002-8331.2106-0055

• Research Hotspots and Reviews • Previous Articles     Next Articles

Review of Single Image Super-Resolution Based on Deep Learning

ZHANG De, LIN Qingyu, GUO Maozu   

  1. School of Electrical and Information Engineering & Beijing Key Laboratory of Intelligent Processing for Building Big Data, Beijing University of Civil Engineering and Architecture, Beijing 100044, China
  • Online:2021-11-15 Published:2021-11-16



  1. 北京建筑大学 电气与信息工程学院&建筑大数据智能处理方法研究北京市重点实验室,北京 100044


Image Super-Resolution(SR) is a technique that uses signal processing and machine learning to reconstruct High Resolution(HR) images from single or multiple Low Resolution(LR) images. Due to the unpredictability of sub-pixel displacement between multiple LR images, Single Image Super-Resolution(SISR) has gradually become the main direction of super-resolution research. In recent years, deep learning method has been developed rapidly and widely used in the field of image processing. Therefore, the deep learning algorithms and network models used in SISR are summarized systematically. Firstly, the setting and evaluation index of SR are introduced. Then, the deep learning algorithms of SISR reconstruction are discussed and compared, mainly from three aspects of network structure design, loss function and upsampling method. Next, the commonly used standard data sets are introduced, and several typical algorithms based on different network models are selected for experimental comparative analysis. Finally, the future research trend and development direction of SR technology are prospected.

Key words: image super-resolution, deep learning, Convolutional Neural Network(CNN), image processing


图像超分辨重建(Super-Resolution,SR)是指利用信号处理和机器学习等方法,从单幅或者多幅低分辨率图像(Low Resolution,LR)中重建对应的高分辨率图像(High Resolution,HR)的技术。由于多幅LR图像之间亚像素位移的不可预知性,单幅图像超分辨重建(Single Image Super-Resolution,SISR)逐渐成为超分辨研究的主要方向。近年来,深度学习方法得到迅速发展,并广泛应用到图像处理领域。因此,针对单幅图像超分辨重建所使用的深度学习相关算法和网络模型进行系统的总结。介绍图像超分辨问题的设置和评价指标;讨论和比较单幅图像超分辨重建的深度学习算法,主要从网络结构设计、损失函数和上采样方式三方面进行论述;介绍常用的标准数据集,并选用基于不同网络模型的几种典型算法进行实验对比分析;展望图像超分辨技术未来的研究趋势和发展方向。

关键词: 图像超分辨, 深度学习, 卷积神经网络(CNN), 图像处理