【作者】 张修浪；

【导师】 江泽涛；

【作者基本信息】 南昌航空大学 ， 计算机应用技术， 2008， 硕士

【摘要】 数字水印是近十年来出现的一种有效的数字产品版权保护技术。它作为加密技术的补充,克服了传统密码学技术中加密的内容在解密之后不再安全的问题,因此它作为知识产权保护和数字多媒体防伪的有效手段而被广泛研究和应用。但作为一个新的研究领域,数字水印仍有许多理论和技术问题亟待解决。找到一种复杂度低,安全性强,不可见性和鲁棒性好的水印算法成为数字水印技术应用的关键。本文在图像水印和视频水印方面提出了一些新的算法。本文在查阅了大量国内外有关数字图像水印和数字视频水印文献的基础上,总结了数字水印技术的国内外研究现状之后提出了一种基于位纠错的DCT域双彩色图像水印算法和一种基于DCT域的MPEG4扩频数字视频盲水印算法,并在windows平台下实现了这两种算法,结合实验结果重点分析了算法的不可见性和鲁棒性。在基于位纠错的DCT域双彩色图像水印算法中,彩色水印图像首先借助了基于DCT的JPEG图像压缩编码算法的思想进行了压缩。在YCbCr彩色空间中每一分量分成8×8块后进行二维DCT,每一块再进行ZigZag排序后提取前15个系数,其他系数被赋值为0。逆ZigZag排序后再进行反离散余弦变换得到待嵌入的彩色水印图像。按块在分量中从行到列的顺序提取它们的前15个DCT系数转化为有符号的二进制数以后再采用扩频通信机制,得到的这个序列经过互不重叠周期延拓8次后与一个等长的伪随机序列进行异或操作产生待嵌入的水印序列。宿主彩色图像经过DCT后,低频部分包含主要能量,高频部分是图像的边缘信息。水印嵌入到低频部分会影响它的不可见性,而嵌入到高频部分水印的鲁棒性不够好,因此本文结合人类视觉系统模型把水印序列嵌入到宿主彩色图像纹理子块DCT后的中低频系数中,该算法简单易实现。在基于DCT域的MPEG4扩频盲数字视频水印算法中,二值水印图像首先经过Arnold置乱变换,得到的序列经过互不重叠周期延拓8次后再与一个等长的伪随机序列进行异或操作得到水印序列。MPEG4视频文件经过XVID解码得到的视频帧转换到YCbCr彩色空间中,Y分量经过DCT后再进行ZigZag排序,根据水印位修改Y分量的中低频DCT系数的大小,嵌入水印后再通过XVID编码得到含水印的MPEG4视频文件。提取水印时,不需要原始视频文件的参与,根据含水印视频帧的DCT系数除以嵌入强度得到的余数范围来提取水印。本文提出的两种水印算法均在C++Builder平台下编程实现并演示了嵌入和提取水印的全过程,实现结果表明它们都具有良好的不可见性和鲁棒性。更多还原

【Abstract】 Digital Watermarking has become an effective method of copyright protection for digital products in the last ten years. As the complement of encryption, it has solved the problem of traditional cryptography that encrypted contents are not secure when have been decrypted. Therefore it has been studied and used widely as an effective method of knowledge copyright protection and preventing from sham for digital multimedia. However, there are still plenty of theoretical and practical difficulties to be solved in a new field of digital watermarking. The key point of applicable digital watermarking technique is to find out an algorithm which is low complexity, good security, invisibility and robustness. Some original algorithms in digital image and video watermarking field are proposed in this paper.Referring to a large number of domestic and foreign literature of digital image and video watermarking, a dual color images watermarking method based on error-correcting bits in discrete cosine translation domain and a blind MPEG4 video watermarking algorithm based on spread spectrum technology in discrete cosine translation domain are proposed after summing up those algorithm which have been proposed in previous authors’papers and they are implemented successfully in windows operating system, analyzing mainly their invisibility and robustness according to the experimental results. The color watermarking image is compressed in viewpoint of JPEG algorithm based on DCT in the dual color images watermarking method. Each component of the color watermarking image in the YCbCr color space is blocked 8×8 and done discrete cosine transform, and then extracting the former 15 DCT coefficients of each block after ordering of ZigZag, the other DCT coefficients assignment 0.The embedded color watermarking image is formed after ordering of inverse ZigZag and then IDCT. Extracting the former 15 DCT coefficients of each block from Row to Column in each component and then they are translated into symbols binary which they are Spread Spectrum to form a sequence, which is extended not overlap each other through 8 cycle to form another sequence. This sequence is XOR-ed with a pseudo random sequence with the same length as it. The host color image after DCT, the low-frequency is main energy of it and the high-frequency is the marginal information of it. The watermark information embedded in the low-frequency of DCT will affect the invisibility and it embedded in the high-frequency of DCT will affect the robustness. The watermark sequence is embedded in the middle-frequency and low-frequency of DCT in texture block of the host image combining human visual system. In the blind MPEG4 video watermarking algorithm based on spread spectrum technology in discrete cosine translation domain, a binary image transformed with Arnold scrambling is extended with not overlapping each other through 8 cycles to form another sequence, which is XOR-ed with a pseudo random sequence with the same length as it. The video frames are obtained from XVID codec and they are inverted to YCbCr color space, the video files are compressed by MPEG4. The Y components are DCT and then ordered by ZigZag, altering those DCT coefficients according to the watermark sequence and then the MPEG4 video embedded the watermark is obtained from the XVID codec. Extract the watermarking under their scope of the remainder which these DCT Coefficients of the MPEG4 video embedded the watermark divided the embedded strength and needn’t have the original MPEG4 video.The two digital watermarking algorithms proposed in this paper are programmed in the C++ Builder 6.0, the whole process of watermark embedded and the whole process of watermark extracted are demonstrated. Experimental results demonstrate that the watermarking is good invisibility and robustness.更多还原