基于退化学习的图像超分辨综述

Overview of Image Super-Resolution Based on Degenerate Learning

图像超分辨通过将分辨率低的单个图像或者视频通过一些技术手段进行放大仍然保持清晰来获取信息,涉及的领域在医疗图像、视频监控等。对超分辨率单图像重建的研究进展进行了综述。讨论单图像超分辨率在退化阶段的方法,将其分成有监督退化和无监督退化学习的两种方式,监督退化主要介绍基于字典学习和基于深度学习的超分辨方法,至今无监督退化的方法还不是很多,所以之后的研究可以以此为重点。

Image super-resolution is to obtain information by amplifying a single image or video with low resolution through some technical means and still keeping it clear.The fields involved are medical images and video surveillance.This article reviews the research progress of super-resolution single image reconstruction.Mainly discuss the method of single-image super-resolution in the degradation stage,and divide it into two ways of supervised degradation and unsupervised degradation learning.Supervised degradation mainly introduces super-resolution methods based on dictionary learning and deep learning.There are not many methods of degradation,so subsequent research can focus on this.