项目名称: 视频内容帧间篡改模式认知的关键技术研究
项目编号: No.61272249
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 孙锬锋
作者单位: 上海交通大学
项目金额: 81万元
中文摘要: 视频篡改模式认知技术,属多媒体信息处理和模式识别领域的应用基础研究。准确高效的视频篡改模式认知算法对视频内容篡改取证、篡改追踪与恢复等具有重要的应用前景。本课题针对视频篡改被动和主动认知方法,展开对视频帧间篡改模式认知得到新算法研究。对于被动认知方法,分别提出了Slice结构特征、宏块参数统计特征等新的获取方法,以及双模态特征交叉验证匹配模型;并结合多层过滤结构,实现对不同视频间篡改模式的被动认知。对于主动认知方法,提出新的微结构水印和新的嵌入域实现认知信息嵌入,并研究失真漂移消除的补偿算法,实现篡改模式的主动认知。课题提出的视频帧间篡改模式认知方法,有效地融合被动认知和主动认知技术的各自优势,可提高对视频帧间篡改模式认知的准确性和适用范围,在面向司法鉴定等要求视频真实性和可追溯性明晰的领域更加有效。研究成果对滥用视频、版权侵权、恶意制造虚假信息牟利等违法行为甄别,具有重要的社会意义。
中文关键词: 视频被动取证;视频主动取证;篡改模式认知;帧间篡改;机器学习方法
英文摘要: The video forgery mode cognition technique, which is a branch of Multimedia Information Process and Pattern Recognition, has significant prospect in video tempering authentication, tracking and recovery. The research of inter frame forgery mode cognition algorithm based on novel passive and active cognition approaches will be delved. For passive cognition approach, new features, such as Slice structural feature and Macroblock statistical feature, and bimodal feature cross validation model are proposed. Furthermore, multilayer filtering framework is adopted. For active cognition approach, new micro-structure watermark, embedding domain and drift compensation algorithm are proposed. The inter frame forgery mode cognition approach can effectively combine the dominances of distinct key technologies, resulting in higher accuracy and larger application scope, especially for judicial appraisal field which calls for the truism and traceability of digital video. The research result has vital social significance, such as the restraint of the abuse of video technique and manufacture of false information.
英文关键词: passive video forensics;positive video forensics;forgery model cognition;inter-frame forgery;machine learning