【作者】 单大伟；

【导师】 李生红；

【作者基本信息】 上海交通大学 ， 通信与信息系统， 2007， 硕士

【摘要】 数字水印技术的研究已引起广泛关注,而多重水印问题却很少有人提及,多重水印技术是用来解决多著作权问题以及数字产品在发布,销售,使用等不同阶段的版权认证问题。多重水印信号在生成,嵌入算法上均比单个水印信号算法有更多的要求,如水印信号必须满足生成简单,数量众多且多个水印信号之间应满足互不相关,多个水印信号在嵌入和检测时应互不干扰等,水印信号的透明性和鲁帮性的问题都更为复杂。其中最为重要的问题是,如何保证嵌入到同一载体的多个水印信号之间互不影响。考虑到数字产品的迅速发展和版权保护的需求日益增加,有必要研究多重水印的嵌入算法以及先后嵌入的水印相互之间的影响等多方面的问题。BMP图片是当前一种常用的图像编码方式,本文选取BMP图片作为水印载体进行多重水印嵌入研究,具有典型的代表意义。论文利用正交序列将多个水印信号融合成一个水印信号嵌入到载体图像中,由于正交序列是一种自相关很大而互相关性为零的序列,因此这种多重水印信号的生成有一定保密性和容错性;然后利用交织器对多重水印信号进行编码和交织,增强水印序列的鲁棒性;随后论文对水印载体所能嵌入的水印容量以及容错性进行了讨论,若利用交织矩阵的长宽作为密钥,攻击者在不知道加密密钥的情况下对多重水印进行分解就会遇到大数分解的难度,因此经过交织器的水印信号几乎不可能破译;最后将各水印信号叠加在图像小波分解低频系数上,从而做到各水印信号在嵌入和提取时互不干扰,也可以防止水印信号被恶意移走,这样就增强了水印信号的安全性;为进一步提高性能,论文提出了一种分组量化逐次逼近的水印嵌入方法来增强水印信号的鲁棒性,以有效地实现水印信号的透明性和鲁棒性的均衡,由于不需要判决的先决条件,因此可以实现多重水印信号的盲检测提取。本文依托上海市科委重点攻关项目,提出了基于正交序列融合多个水印信息成为多重水印的算法。在此基础上,将本算法应用在移动通信的票证安全系统,其主要应用在两种特殊图片:龙贝码图片和PDF417条形码图片,都取得了良好的实验效果。更多还原

【Abstract】 In recent years, Digital Watermarking technology research has aroused wide concern. However, few people have mentioned Multiple Watermark. Multiple Watermarking technology is used to solve many copyright issues and digital products in the publication, sale, and different using stages of the certification rights. Compared to single watermark, Multiple Watermarking signal’s generation and the algorithms to embed watermark signal has more requests. For example, the watermark signal must meet the formation of simple, and the large number of Multiple Watermarking signals should unrelated, Moreover, how to keep the Multiple Watermark signal transparency and its robustness is more complex. Among all these questions, the most important one is how to ensure several watermark signals embedded into the same carrier has no impact on each other. Taking into account the rapid development of digital products and the growing demand of copyright protection, it is necessary to study multiple watermark embedded algorithm and how the embedded watermark signals have mutual influence between each other. BMP image is a popular image coding method used epidemically recently. Hence, it’s a typical representative of significance for this paper to select BMP images as the carrier of multiple watermarking.In this paper, we propose a useful method to embedded Multiple Watermark into the carrier. Firstly, we use orthogonal sequence to merger several watermark signal into one signal, then we embedded the combination signal into the image carrier. Since the orthogonal sequence is a great autocorrelation and zero cross correlation sequence, we can assure the confidentiality and fault-tolerance of the multiple watermark signals; The next step is to use interleave to make the multiple watermark signal intertwined, in this stage, how many watermark signals the image carrier can carry and fault-tolerance were discussed; If we use the intertwined matrix as a key dimensions, the attacker would not aware of encryption keys under the right circumstances for Multiple Watermarking decomposition will encounter the Factorization difficulty, so the conclusion is that the intertwined watermark signal almost impossible deciphered; Finally, the watermark signal superimposed on the image wavelet decomposition of low-frequency coefficients, so do the watermark embedded in the signal extraction and Noninterference can also prevent malicious watermark signals are removed, thus, the security of the watermark signal were enhanced. To further enhance performance and considering the localization character of wavelet transform in spatial domain, a block-based rounding method is adopted to achieve the purpose of precise embedding. As to the detecting algorithm, a novel difference detecting algorithm is created to improve the robustness of system. As a prerequisite for no judgment, we can achieve multiple watermark detection signals blindly.The research work and results of this paper were relying on a major key project in Shanghai Science and Technology Commission, we proposed an algorithm to integrate multiple orthogonal sequence information into multiple watermarking. On this base, we apply the algorithm to mobile security coupons system, the main application were two special photos: Long-pei pictures and PDF417 photo, and the experiment has produced good results.更多还原