摘要
生成对抗网络作为近年出现的一类深度学习模型,在数字图像智能化修复领域展示出较好的应用前景。通过文献回顾及考究该网络在老照片档案智能化修复中的特点,发现其仿真度好、灵活性强、处理效率高、数据可扩增。在具体技术实现机理及过程中,也验证了其在老照片档案智能化去模糊、修补及修补效率等方面效果良好,值得推广,但同时也要防范其中的“深度伪造”问题。
Generative adversarial networks,as a type of deep learning model that has emerged in recent years,have shown good application prospects in the field of intelligent digital image restoration.Through literature review and research on the characteristics of this network in intelligent restoration of old photo archives,it was found that it has good simulation,strong flexibility,high processing efficiency,and data augmentation.In the specific technical implementation mechanism and process,it has also been verified that has good effects in intelligent deblurring,repair,and repair efficiency of old photo archives,which is worth promoting.However,it is also necessary to prevent the problem of"deep forgery"in it.
作者
徐辛酉
杨晓芳
XU Xinyou;YANG Xiaofang(College of Society and History,Fujian Normal University,Fuzhou 350117,China;Hengzhou City People's Court,Nanning 530300,China)
出处
《档案学通讯》
CSSCI
北大核心
2024年第6期115-121,共7页
Archives Science Bulletin
基金
国家社会科学基金一般项目“新《档案法》规制下档案数字监管模式研究”(22BTQ091)。
关键词
老照片档案
生成对抗网络
图像修复
智能化修复
Old photo archives
Generative adversarial networks
Image inpainting
Intelligent repair