利用视觉感知模型宏块分类的视频安全研究  

Video Security Research Based on Visual Perception Model Macroblocks Classification

作　　者：袁春经 张小红[1] 

机构地区：[1]江西理工大学信息工程学院,江西赣州341000

出　　处：《小型微型计算机系统》2014年第12期2749-2753,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(11062002)资助;江西省自然科学基金项目(2010GZS0083;20122BAB211019)资助

摘　　要：已有的大多数H.264加密算法只考虑了视频编码的特性,而忽略了人眼才是处理视频信号的接收端.提出一种基于人眼视觉感知特性的运动注意力的视频熵编码加密算法.将所有宏块根据视觉感知注意力模型分成I帧宏块、中心区域宏块、注意力较大宏块和注意力较小宏块.采用混沌系统产生加密密钥,然后对四类宏块采用不同的加密手段,先对I帧宏块DCT系数翻转然后进行熵编码加密,然后引入标示宏块在人眼中注意力强弱的视觉注意力权重分布w:当wij=1时,宏块处于中心区域,对MV值、DCT系数的熵编码过程加密;当wij≠1时,则计算宏块运动量的加权,并与阈值T比较.注意力较大宏块进行熵编码加密,注意力较小宏块则不做加密操作.本文算法将感知模型引入加密算法中,在同一帧内实现分类的加密手段.实验表明,本算法加密后数据量仅增加了10%-15%,加密所占用额外时间较小,基本对信噪比SNR没有造成影响,从直方图和混淆后的图形分析,加密的效果显著.Most of the H. 264 encryption algorithms only consider the characteristics of the video coding, but the human eyes process the video signal at the receiving end. This paper puts forward a kind of video encryption algorithm based on human visual perception characteristics of motion attention and entropy coding. There are four kings of macroblocks based on visual perception attention model, the I frame macroblocks,center area macroblocks, smaller attention macroblocks and the larger attention macroblocks. Using chaotic system to produce the encryption key, then the four types of macroblocks with different means of encryption： turn the I frame macroblocks DCT coefficient and entropy encoding;Then introduce the visual attention weight distribution w which marking macroblocks attention of the strength in the eyes, when wij = 1, the macroblocks in the central region, encrypt the MV value, DCT coefficient of entropy coding; when wij ≠ 1, calculate the weight of macroblocks MV value, and compared with the threshold T, the larger attention macroblocks entropy coding encryption;smaller attention macroblocks is not do cryptographic operations. In this paper bring the perception model into the encryption algorithm, the same frame within classification encryption methods, the experimental results show that, encrypted data volume increased by 10% - 15%, the algorithm has little influence on SNR, analysis the histogram and promiscuous graphical, the encryption effect is remarkable.

关 键 词：视觉感知 宏块分类 视频安全 熵编码 运动矢量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]