基于视觉特性和DCT变换空域视频水印算法.PDFVIP

第5卷 第2期 106 中国科技论文在线 Sciencepaper Online 2010 年 2 月 基于视觉特性和 DCT 变换的空域视频水印算法 刘绍辉，王继广，石 峰，姚鸿勋 （哈尔滨工业大学计算机科学与技术学院, 哈尔滨 150001） 摘 要：结合人眼视觉特性HVS(human visual system) ，提出了一种基于块DCT(discrete cosine transform)直流(DC)系数 视觉特性的视频水印算法。为了满足视频水印的不可见性和鲁棒性要求，该算法利用DC系数的稳定性，首先将帧内所 有块的DC系数组成一维序列，然后将其划分为若干子序列，对每个子序列进行一维DCT变换，并采用扩频方法在变 换后的高频系数中嵌入水印位，水印嵌入强度由DCT域JND(just-noticeable distortion)模型完全地动态自适应调节。实验 结果表明，相对于固定的嵌入强度，采用JND模型可以在保护原有视频序列视觉质量的基础上，提高水印自身的鲁棒 性。该算法对不同视频压缩标准的不同码率下的编码、常见信号处理和几何攻击等，均具有抗攻击的鲁棒性。 关键词：DCT ；JND视觉模型；扩频；视频水印 中图分类号：TP309.7 文献标志码：A 文章编号：1673－7180(2010)02 －0106 －6 Spatial video watermarking algorithm based on human visual system and DCT Liu Shaohui ，Wang Jiguang ，Shi Feng ，Yao Hongxun (School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China) Abstract: We propose a novel video watermark algorithm based on the visual characteristics of block DCT direct current (DC) coefficient which considers the JND model in human visual system (HVS). In order to meet the invisibility and robustness of video watermarking, the proposed method utilizes the stability of DC coefficient. First, we gather all the DC coefficients in luminance blocks to form a sequence, and then divide the sequence into many sub-sequences, on which we use 1D-DCT respectively. Second, the highest frequency coefficient of each sub-sequence is chosen to embed the watermark message by the spread spectrum (SS) method, where the JND model is adopted to dynamically adjust the embedding strength. Experimental results indicate that proposed algorithm makes a good tradeoff be