基于压缩感知的数字水印算法研究

发布时间：2018-03-13 16:33

本文选题：压缩感知　切入点：数字水印　出处：《华北电力大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着互联网与多媒体技术的普及,数字产品的存储、复制和传播越来越便捷,但同时多媒体数据的安全性也遭受着越来越多的威胁。输入输出设备的发展和高精度产品的出现使得伪造更加容易；通信技术的发展与普及,数字作品、电子出版物的传播与交易变得越来越便捷,侵权盗版活动也日益猖獗；互联网的普及与黑客技术的出现使得机要信息的发布、传递、管理更加容易引起攻击者的注意,这样的情况让加密数据与未加密数据几乎一样危险。传统的加密措施已经无法解决现今面临的各种问题,数字水印技术作为信息隐藏技术的一个分支,在版权保护、完整性认证和保密通信方面具有不可小觑的作用。现有数字水印算法的研究成果大多集中在变换域,此类方法嵌入水印相对于空域而言数据量较大,安全性较高,鲁棒性较强,但是其计算复杂且抗攻击能力和防提取能力仍然有限。压缩感知(CRPSUHVVLYH6HQVLQJ, C6)是最早由CDQGHV等人提出的优于传统香农采样理论的新理论,它针对稀疏或者可压缩信号,能够在信号采样的同时对数据进行压缩,且其采样率远远低于lyTXLVW采样定理限制下的采样率,该理论能够通过一定的优化算法由少量采样数据来完美重建信号。C6理论有三个核心过程,分别为：信号的稀疏化、感知矩阵的构建和信号的重建,该理论具有比D：7、DC7以及小波变换所无法企及的优势,因此本文将压缩感知应用于数字水印算法的研究,能够得到更为鲁棒、更为安全的数字水印。本文首先深入的研究了先前的数字水印算法和C6理论及其重建方法,然后结合人眼视觉系统(HXPDQ9LVXDO6yVWHP, H96)的特性,提出一种较高性能的数字水印算法,该算法主要利用块不均匀度来选取容量大的子块,采用量化的方法自适应的选择量化步长来嵌入经过置乱处理后的数字水印,然后详细的介绍该算法实现步骤。经实验验证,该算法可保证数字水印具有更好的鲁棒性、隐蔽性和安全性,并且提高含水印图像的抗攻击能力和防提取能力,而且提取水印不需原图像的参与,能够大大降低存储成本。
[Abstract]:With the popularity of the Internet and multimedia technology, the storage, replication and dissemination of digital products are becoming more and more convenient. But at the same time, the security of multimedia data is also being threatened by more and more. The development of input and output equipment and the appearance of high-precision products make it easier to forge, the development and popularization of communication technology, digital works, The dissemination and transaction of electronic publications have become more and more convenient, and piracy activities have become increasingly rampant. With the popularity of the Internet and the emergence of hacker technology, the publication, transmission, and management of confidential information are more likely to attract the attention of attackers. This situation makes encrypted data almost as dangerous as unencrypted data. Traditional encryption measures can no longer solve the problems faced today. Digital watermarking technology, as a branch of information hiding technology, is protected by copyright. The research results of the existing digital watermarking algorithms are mostly focused on the transform domain. Compared with the spatial domain, this method has a large amount of data, high security, strong robustness. However, its computational complexity and its ability to resist attack and extraction are still limited. CRPSU HVVLYH6HQVLQJ (C6) is a new theory proposed by CDQGHV et al. It is superior to the traditional Shannon sampling theory. It can compress the data while sampling the signal for sparse or compressible signals. The sampling rate is far lower than the sampling rate limited by the lyTXLVW sampling theorem. The theory can reconstruct the signal perfectly by a small amount of sampling data through a certain optimization algorithm. There are three core processes of the theory: sparse signal. The theory of perception matrix construction and signal reconstruction has more advantages than that of D7 / DC7 and wavelet transform. Therefore, this paper applies compressed sensing to the research of digital watermarking algorithm and can obtain more robust and secure digital watermarking. In this paper, the previous digital watermarking algorithm, the C6 theory and the reconstruction method are studied in depth, and then a high performance digital watermarking algorithm is proposed based on the characteristics of the human visual system (HXPDQ9LVXDO6yVWHP, H96). The algorithm mainly uses block inhomogeneity to select large capacity sub-blocks, adopts quantization method to select quantization step size adaptively to embed the scrambled digital watermark, and then introduces the implementation steps of the algorithm in detail. This algorithm can guarantee better robustness, concealment and security of digital watermarking, and improve the ability of resisting attack and anti-extraction of watermark image. Moreover, the watermark extraction does not need the participation of the original image, so the storage cost can be greatly reduced.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP309.7;TN911.7

【参考文献】