摘要
清晰人脸受实际场景各种因素制约难以获得,因此模糊人脸的增强与清晰化技术具有重要的实际应用意义,但传统的图像增强算法存在处理类型单一等局限。采用基于深度生成对抗网络的模糊人脸增强方法,通过生成模型实现人脸清晰化。在7种模糊类型上与当前主流的图像到图像生成对抗网络进行对比,结果表明,该算法生成能够生成更加清晰、与原图相似度更高的增强人脸图像。
It is difficult to obtain a clear face due to various factors in the actual scene.Therefore,enhancement of fuzzy faces and clearness techniquehave have important practical application significance.However,traditional image enhancement algorithms have limitations such as a single type of processing.This paper uses a fuzzy face enhancement method based on deep generative adversarial network.The face sharpening was realized by generating model.The current mainstream image-to-image generation adversarial network was compared on the 7 fuzzy types.The results show that the algorithm can generate enhanced facial images with clearer and higher similarity to the original image.
作者
仝宗和
刘钊
Tong Zonghe;Liu Zhao(School of Information Technology and Network Security,People’s Public Security University of China,Beijing 102623,China;Cyberspace Security and Rule of Law Collaborative Innovation Center,People’s Public Security University of China,Beijing 100038,China)
出处
《计算机应用与软件》
北大核心
2020年第9期146-151,193,共7页
Computer Applications and Software
基金
国家重点研发计划项目(2018YFC0809800)
公安部公安理论软科学项目(2018LLYJGADX014)。
关键词
深度学习
生成对抗网络
图像增强
Deep learning
Generative adversarial network
Image enhancement
