基于内容自适应的数字图像水印技术的研究

【作者】 王玮；

【导师】 樊则宾；

【作者基本信息】 昆明理工大学 ， 物理电子学， 2008， 硕士

【摘要】 数字水印技术是目前信息安全技术领域研究的一个新方向,它是一种可以在开放的网络环境下保护版权和认证来源及信息完整性的新技术。创作者的创作信息和个人标志通过数字水印系统以人们不可直接感知的水印形式嵌入在数字产品中,只有专用的软件才可以检测出隐藏的数字水印。数字水印技术是实现版权保护的有效方法,目前己成为多媒体信息安全研究领域的一个热点,也是信息隐藏技术研究领域的重要分支。小波分析由于具有多分辨率和多尺度分析特性,在图像处理方面得到了广泛应用。利用小波分析可以更好地控制水印信息在图像中的分布,从而协调解决稳健性和不可见性之间的矛盾。提升小波变换继承了第一代小波变换的多分辨率特性,并且具有不依赖于傅立叶变换、结构简单、运算速度快等诸多优点,而受到联合图像专家组的大力推荐。本文在对采用灰度水印的图像数字水印技术进行详细研究的基础上,提出了一种基于人类视觉特性和图像纹理特征的灰度水印自适应嵌入与提取新方案。该方案能够在保证水印不可见性的同时,使水印具有较高的鲁棒性,其主要特点有:(1)实现了水印的自适应嵌入。即在充分考虑人眼的亮度感觉及对比灵敏度阈值等人类视觉系统特性的基础上,通过计算各个局部块(用小波树表征)的水印嵌入强度因子来嵌入水印。水印嵌入强度因子能够根据局部块的亮度、纹理复杂度以及所处的频带位置来动态地进行调整,并能兼顾水印的不可见性和鲁棒性。(2)实现了水印的检测。水印的检测是嵌入的逆过程,在检测过程中,需知道嵌入过程使用小波树的顺序,如果不知道载体图像的完全树顺序,即使知道算法,攻击者也很难知道水印嵌入的位置,这就增加了水印的安全性,改善了水印系统性能。(3)实现了该算法抗攻击能力的评价。该算法对诸如裁剪、运动模糊、椒盐噪声污染、压缩等常见攻击都具有较强的抵抗能力。更多还原

【Abstract】 Digital watermark is a new subject in the information security field. It’s a new technology which is used to protect copyright and authentication source. The author’s copyright and personal information can be embedded in the products. But we can’t percept it, only special software can pick it up. Digital watermark technique is a effective method for copyright protection, which has become a hot study field in multimedia security researchs and a key branch of hiding technology researchs .Wavelets be used widely in image processing due to its multiresolution and multiscale analysis property. Using discrete wavelet transform(DWT), We can manage the distribution and location of the watermark properly, and then get a robust and invisible marked image. Lifting scheme is a new methods to wavelet transform. Use this method, we can compute the wavelet transform simply,fastly without the use of fourie transform. So it is strongly recommended by the joint picture experts group(JPEG).Following the instructions of the current watermark technology,a new watermark technology which adaptive to specifics of the image is presented. The details of the algorithm are following:(1) After computing the wavelet trees, a properly intesive watermark will be embeded in the properly location. where the watermark be added to is decided by the complication of the original image, and the intensibility of the watermark is decided by the human vision system (HVS), the new method can guarantee the invisibility as well as the robustness of the watermark.(2)The watermark can also be detected adaptive to the image. The process of detecting is inverse to the embedding. if the order of the wavelet trees is not be known ,the watermark will not be detected easily.(3)Evaluated the algorithm. After the experiments and attack test, we can get that the algorithm has good tolerance to attacks such as cut, additive noise, and compress attacks.更多还原