基于PU块划分模式和DCT系数共生矩阵的HEVC视频重压缩检测算法

发布时间：2018-04-22 13:41

  本文选题：数字视频取证 + HEVC/H.265　； 参考：《北京交通大学》2017年硕士论文

【摘要】：随着网络应用的普及和多媒体信息产业的快速发展,数字视频通过互联网和智能手机融入到人们的生活中,并成为司法证据的重要组成部分。数字视频内容真实性问题日益严峻,各种功能强大且操作简单的多媒体编辑软件使得人们能够轻易地对图像视频进行恶意的编辑和修改,影响司法公正和社会安定,使得数字视频真伪的鉴别具有重要的实用价值和广阔的发展前景。视频篡改的过程必然要对视频进行重压缩,这使得视频重压缩检测技术成为了视频真实性取证的一个重要技术手段。HEVC作为最新一代的国际视频编码标准,将广泛地应用于在高清、超高清视频和流媒体服务领域。本文从HEVC视频重压缩引起的帧图像内容变化的角度出发,提出了两种视频重压缩检测算法。1.针对不同比特率下HEVC视频重压缩检测问题,提出了基于PU块划分模式的HEVC视频重压缩检测算法。利用双重压缩HEVC视频4×4PU块的数目变化趋势与单次压缩视频不一致的特性,统计分析视频在单次压缩和重压缩下4×4PU块的数目并做归一化处理,提取出数目变化曲线,利用数目变化曲线中的凸起现象对视频进行重压缩检测。实验结果表明,该算法平均检测率达到了 80%以上,能够有效地对HEVC重压缩视频进行鉴别。在目前针对HEVC重压缩检测研究成果较少的情况下,该算法能够针对HEVC特有的PU语法元素进行特征提取,实现对HEVC重压缩视频的检测,进一步推动了视频重压缩取证算法的发展。2.为了进一步提高对HEVC重压缩视频的检测率,在基于PU块划分模式检测算法的基础上,利用双重压缩HEVC视频Ⅰ帧量化DCT系数与单次压缩视频不同的特性,通过共生矩阵来表征这种差异,提出了基于PU块划分模式和DCT系数共生矩阵的HEVC视频重压缩检测算法。将HEVC视频中Ⅰ帧预测单元PU(Prediction Unit)划分类型的块数目特征和量化DCT系数的共生矩阵特征结合构建联合特征,全面地反映重压缩对HEVC视频数据的影响。实验结果表明,所提算法能在基于PU块划分模式检测算法的基础上大幅提高检测率,有效地区分HEVC单次压缩视频和HEVC双重压缩视频。
[Abstract]:With the popularity of network applications and the rapid development of multimedia information industry, digital video has been integrated into people's lives through the Internet and smart phones, and has become an important part of judicial evidence. The authenticity of digital video content is becoming more and more serious. All kinds of powerful and simple multimedia editing software make it easy for people to edit and modify image and video maliciously, which affects judicial justice and social stability. The identification of digital video authenticity has important practical value and broad development prospects. The process of video tampering is bound to recompress the video, which makes the video recompression detection technology become an important technical means of video authenticity forensics. HEVC as the latest generation of international video coding standards, Will be widely used in HD, HD video and streaming media services. From the point of view of the change of frame image content caused by HEVC video recompression, this paper proposes two video recompression detection algorithms. Aiming at the problem of HEVC video recompression detection with different bit rates, a new HEVC video recompression detection algorithm based on pu block partition mode is proposed. Taking advantage of the fact that the number of 4 脳 4PU blocks of a double compressed HEVC video is not consistent with that of a single compressed video, the number of 4 脳 4PU blocks in a single compression or a recompression is analyzed and normalized to extract the number variation curve. The recompression detection of video is carried out by using the convex phenomenon in the number changing curve. Experimental results show that the average detection rate of the algorithm is over 80%, which can effectively identify the HEVC recompressed video. Under the condition that the research results of HEVC recompression detection are few, the algorithm can extract the feature of pu syntax element of HEVC, realize the detection of HEVC recompressed video, and further promote the development of video recompression forensics algorithm. 2. In order to further improve the detection rate of HEVC recompressed video, based on the detection algorithm of pu block partition mode, the characteristics of quantization DCT coefficient of double compressed HEVC video frame I are different from that of single compressed video. This difference is represented by co-occurrence matrix, and a HEVC video recompression detection algorithm based on pu block partitioning mode and DCT coefficient co-occurrence matrix is proposed. The block number features of the division type of PU(Prediction unit and the co-occurrence matrix feature of quantized DCT coefficients in HEVC video are combined to form a joint feature, which reflects the effect of recompression on HEVC video data. Experimental results show that the proposed algorithm can greatly improve the detection rate on the basis of pu block partition pattern detection algorithm, and can effectively distinguish HEVC single compressed video from HEVC dual compressed video.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【参考文献】

中国期刊全文数据库 前10条

1 张珍珍;侯建军;李赵红;郭胜;;基于MSSIM商一致性的视频插帧和删帧篡改检测[J];北京邮电大学学报;2015年04期

2 黄美玲;王让定;徐健;李倩;徐达文;;基于DCT系数统计特性的HEVC视频双压缩检测算法[J];光电子·激光;2015年04期

3 张旭;黎智辉;王鑫;彭思龙;许小京;王世君;;视频取证技术研究进展[J];刑事技术;2015年02期

4 杜振龙;焦丽鑫;李晓丽;郭延文;杨小健;;帧内复制粘贴视频伪造的盲检测[J];中国图象图形学报;2014年06期

5 王琬;蒋兴浩;孙锬锋;;基于首位数字特征的双重MPEG压缩检测算法[J];电子与信息学报;2012年12期

6 黄添强;吴铁浩;袁秀娟;陈智文;;利用模式噪声聚类分析的视频非同源篡改检测[J];计算机科学与探索;2011年10期

7 胡永健;刘t2贝;贺前华;;数字多媒体取证技术综述[J];计算机应用;2010年03期

8 王俊文;刘光杰;张湛;王执铨;戴跃伟;;基于模式噪声的数字视频篡改取证[J];东南大学学报(自然科学版);2008年S2期

9 刘连山,李人厚,高琦;视频数字水印技术综述[J];计算机辅助设计与图形学学报;2005年03期

10 叶登攀,戴跃伟,王执铨;视频水印技术研究综述[J];计算机工程与应用;2005年01期

中国博士学位论文全文数据库 前1条

1 徐俊瑜;数字视频被动取证技术研究[D];天津大学;2013年

中国硕士学位论文全文数据库 前3条

1 郭胜;基于视频特征QoMSSIM的帧间篡改取证算法研究[D];北京交通大学;2016年

2 董琼;基于统计特征的数字视频被动取证技术研究[D];湖南大学;2012年

3 刘洁;基于MPEG-2运动补偿边缘效应的视频篡改检测研究[D];天津大学;2010年

，

本文编号：1787465