保护监控视频隐私的漂移失真免疫算法研究

项目名称： 保护监控视频隐私的漂移失真免疫算法研究

项目编号： No.61202302

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 马晓静

作者单位： 华中科技大学

项目金额： 22万元

中文摘要： 当前监控视频隐私泄露问题严重。目前相关研究大多集中在像素域隐私保护领域。而我国监控视频广泛采用H.264压缩格式存储和传输，因此，需研究压缩域隐私保护技术。本项目以H.264压缩视频为隐私保护对象，旨在探索压缩域的对漂移失真免疫的隐私保护算法，加快隐私保护速度。拟基于预研成果，探索用本团队创新性提出的帧内漂移失真免疫理论与压缩域补偿算法、帧间漂移失真免疫理论与压缩域补偿算法、帧内帧间漂移失真均免疫的压缩域受控安全扰乱算法，解决压缩域隐私保护的帧内漂移失真、帧间漂移失真、以及无漂移失真的压缩域扰乱等三个方面的关键科学问题。研究成果可为H.264视频快速隐私保护提供理论依据和技术方法，有利于视频监控应用的实时响应。

中文关键词： 监控视频隐私保护；可逆隐私区域保护；帧内失真漂移；帧间失真漂移；

英文摘要： Digital Intelligent Video Surveillance has attracted a wide spread attention, and has been widely applied in every walk of life. However, the lack of centralized control and unified management makes the leakage of privacy a very serious problem in current video surveillance. Since magnanimous compressed videos have to be analyzed and responded in real-time, thus, Real-time Privacy Protection become a vital issue in video surveillance. However, existing pixel-domain based privacy protection approaches which have to decode the compressed video, protect the privacy-sensitive region, and then re-compress the video, do not support real-time response for their high complexity and long time delay. This research is based on the H.264 video coding standard which is widely used in current video surveillance. According to the real-time response requirement of magnanimous compressed videos, this research aims to present a novel transform-domain privacy protection approach without the distortion drift in privacy-insensitive region which are introduced by the scrambling of privacy-sensitive region. This novel approach can be applied on compressed video directly, while avoiding the decoding and re-encoding process and without distortion drift, thus can provide real-time privacy protection for video surveillance. The main resea

英文关键词： Video surveillance privacy protection；Lossless privacy protection；Intra-frame distortion drift；Inter-frame distortion drift；