抗几何攻击的数字图像水印算法研究-智能与分布计算实验室

抗几何攻击的数字图像水印算法研究

姓名	胡建华
论文答辩日期	2007.01.26
论文提交日期	2007.02.02
论文级别	硕士
中文题名	抗几何攻击的数字图像水印算法研究
英文题名	Research on digital watermarking algorithms based on Geometrical Attacks Invariant
导师1	卢正鼎
导师2
中文关键词	数字水印;几何攻击;几何不变域;离散傅立叶变换;拉顿变换
英文关键词	Digital watermarking;geometrically attacks;Geometrically Invariant domain;Discrete Fourier Transformation;Radon Transformation
中文文摘	在数字多媒体技术以及计算机网络飞速发展的今天，数字作品的非法侵权盗版和恶意篡改变得十分容易，这一系列数字化技术本身带来的负面效果，已经成为信息产业持续发展的一大障碍。数字水印被认为是一种极具潜力的数字作品版权保护技术，其基本思想是将版权管理信息作为水印信号，嵌入到图像、文本、视频和音频等数字作品中，并且在需要时能够通过一定的技术检测手段抽取出水印，以此作为判断数字作品的版权归属和跟踪起诉非法侵权的证据。数字水印是数字多媒体版权保护的新技术，是一门新兴的交叉学科，对于该技术的某些研究还处于起步阶段。虽然鲁棒水印技术的研究与应用取得了很大的进展，但是在如何抵抗几何攻击仍然是水印领域所面临的最大困难，成为当前数字水印产品商业化最大的障碍，也是研究的热点之一。另一方面，数字水印的各种应用中又对数字水印的不可见性有较高的要求。水印的鲁棒性和不可见性之间的矛盾要求数字水印技术必须要取得二者合理的平衡。围绕数字水印的应用要求，重点分析了抗几何攻击图像水印算法的基本概念和常用方法，对人类视觉系统特性进行了深入的研究和分析。重点研究了当前数字图像水印鲁棒性研究中的热点领域，即基于离散傅立叶变换和拉顿变换的图像水印算法。基于人类视觉系统的基本性质和特点，给出了一种基于几何不变域的抗几何攻击图像数字水印算法。该算法在对载体图像的亮度矩阵做拉顿变换和离散傅立叶的基础上，并通过限定水印嵌入位置等一些措施来构建了一个可以抵抗旋转和尺度变换等几何攻击的几何不变域，通过在该不变域的中高频域中嵌入水印序列位来实现水印的嵌入和提取。这样既保证水印有足够的鲁棒性也充分保证了水印的不可感知性。
英文文摘	With the rapid development of the digital multimedia technology and computer networks, it is easy to duplicate, modify and transmit the digital works. The negative affect coming after the development has greatly balked the advancement of the multimedia technology. Yet the digital watermark technology has great advantage to satisfy the need of the digital works right protect technology. The base is insert the right information as watermarking into the digital works, such as picture, text , video and audio. After a series of operation, we can still retrieve the watermarking from the works which has been embedded before. And the watermarking should also be the witness of the right pirate action. Digital watermarking is a new multimedia right protect technology without enough advance research, which covers many fields. The researchers has made some progress in the study of robustness, while we still have to face the problem of geometrical distortions, which has became the biggest obstacle in the business. Most application requires great in unnoticeable which is conflict to robustness, so the digital watermarking must balance the two considerately. Focuses are the principium of geometrically Invariant Watermarking and methods most used by image geometrically Invariant Watermarking, while the Human Visual System (HVS) is also present. Great attention has been paid to the common image algorithms using Discrete Fourier Transformation and Radon Transformation, which has seized great passion from many researchers. A new image geometrically Invariant domain algorithm, basing on the Human Visual System model, brings the property of insensitivity to geometrical distortions. The embedding and detection scheme is bound with Geometrically Invariant domain constructed by Radon Transformation and Discrete Fourier Transformation. After the construct of the Geometrically Invariant domain, the watermarking is embedded and detected in some fixed position. Such measure can provided the algorithm robustness to geometrical distortions.