摘要
图像修复技术是目前计算机视觉领域的研究热点之一,该技术主要利用缺失区域周边或者外部辅助数据来对图片受损区域进行信息推理和修复。随着大数据时代的到来,基于深度学习的图像修复技术以其出色的性能成为了图像处理领域内的关注点。对现有的图像修复算法进行总结,对各个算法的模型结构、性能表现及在常用数据集上的指标进行讨论说明,并对该领域目前所存在的问题和难点进行分析和展望。
Image inpainting is a hot topic in the field of computer vision.It is a process that enables filling in damaged regions with alternative contents by estimating the relevant information either from surrounding areas or external data.With the advent of big data,image inpainting methods based on deep learning have attracted significant attention in image processing because of their excellent performance.This paper presents a brief review of existing image inpainting approaches and discusses the network structure and performance of each algorithm,along with a comparison of widely used datasets.In view of the existing challenges in this field,this paper proposes potential research directions and developmental trends in image inpainting.
作者
李雪涛
王耀雄
高放
Li Xuetao;Wang Yaoxiong;Gao Fang(School of Electrical Engineering,Guangxi University,Nanning 530004,Guangxi,China;Institute of Intelligent Machines,Chinese Academy of Sciences,Hefei 230031,Anhui,China)
出处
《激光与光电子学进展》
CSCD
北大核心
2023年第2期19-34,共16页
Laser & Optoelectronics Progress
基金
广西科技基地和人才专项(2020AC19253)
安徽省重点研究与开发计划(202104a05020041)。
关键词
图像修复
深度学习
卷积神经网络
自编码网络
对抗生成网络
image inpainting
deep learning
convolutional neural network
auto encoder network
generative adversarial network
