利用视觉感知模型宏块分类的视频安全研究
发布时间：2019-03-14 20:46
【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.
[Abstract]:Most existing H. 264 encryption algorithms only consider the characteristics of video coding and ignore that the human eye is the receiver of video signal processing. An entropy-coded video encryption algorithm based on human visual perception is proposed. According to the visual perceptual attention model, all macroblocks are divided into I frame macroblocks, central area macroblocks, attention larger macroblocks and small attention macroblocks. Chaotic system is used to generate encryption keys, and then four kinds of macroblocks are encrypted by different encryption methods. First, the DCT coefficients of frame I macroblocks are flipped and then entropy coded encryption is carried out. Then the visual attention weight distribution w is introduced, which indicates the attention intensity of macroblocks in human eyes. When wij=1, the macroblocks are in the central region, and the entropy coding process of MV and DCT coefficients is encrypted. When wij 鈮，
本文编号：2440355
[Abstract]:Most existing H. 264 encryption algorithms only consider the characteristics of video coding and ignore that the human eye is the receiver of video signal processing. An entropy-coded video encryption algorithm based on human visual perception is proposed. According to the visual perceptual attention model, all macroblocks are divided into I frame macroblocks, central area macroblocks, attention larger macroblocks and small attention macroblocks. Chaotic system is used to generate encryption keys, and then four kinds of macroblocks are encrypted by different encryption methods. First, the DCT coefficients of frame I macroblocks are flipped and then entropy coded encryption is carried out. Then the visual attention weight distribution w is introduced, which indicates the attention intensity of macroblocks in human eyes. When wij=1, the macroblocks are in the central region, and the entropy coding process of MV and DCT coefficients is encrypted. When wij 鈮，
本文编号：2440355
本文链接：https://www.wllwen.com/kejilunwen/wltx/2440355.html
