基于参考图像的超分辨率重建算法综述  

A Review of Reference-Based Image Super-Resolution Reconstruction Algorithms

作　　者：张东晓[1] 唐妮 ZHANG Dongxiao;TANG Ni(School of Science,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学理学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2023年第5期442-453,共12页Journal of Jimei University：Natural Science

基　　金：福建省自然科学基金项目(2020J01710);集美大学国家基金培育计划项目(ZP2020063)。

摘　　要：单帧图像超分辨率重建是指从低分辨率图像重建出高分辨率图像的过程,是一类典型的不适定性问题。近期出现一种新的重建思路:将已有的高分辨率图像作为参考,利用参考图中丰富的纹理来补偿低分辨率图像缺失的细节信息,从而缓解单帧图像超分辨重建的不适定性。这种基于参考图的超分辨率重建,与普通的单帧重建不同,除输入低分辨率图像外,还需额外输入一张或多张与低分辨率图像内容或纹理相似的高分辨率参考图像。尽管该重建思路的相关研究开展时间较短,但也涌现出了很多先进的重建算法,因此急需对这些方法做一个全面的梳理。在分析和借鉴相关研究成果的基础上,从基于图像对齐的方法和基于图像块匹配的方法入手,对近年来出现的新方法进行综述性介绍,并探讨其未来的发展方向,以期抛砖引玉,为该领域的后续研究提供有益参考。Single image super-resolution reconstruction refers to the process of reconstructing high-resolution images from low-resolution images,which is a typical ill-posed problem.Recently,a new reconstruction idea has emerged:Taking the existing high-resolution image as a reference,the rich texture in the reference image is used to compensate for the missing detail information of the low-resolution image,so as to alleviate the ill-posedness of single image super-resolution reconstruction.This new method is called reference-based image super-resolution reconstruction.Different from ordinary single image super-resolution reconstruction,the input of this method also needs one or more high-resolution reference images with similar content or texture to the low-resolution image.Although the related research on this reconstruction method has been carried out for a short period of time,many advanced reconstruction ideas and algorithms have emerged,and a comprehensive review is urgently needed.Therefore,on the basis of analyzing and drawing on relevant research results about alignment-based and patch-matching-based methods,this paper gives a general introduction to the new methods emerging in recent years,and discusses the future development direction.It is hoped that this will be a useful reference for subsequent research in this field.

关 键 词：图像超分辨率重建 参考图 图像对齐 图像块匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]