数字水印是解决数字多媒体作品认证的一个重要途径，是传统加密有效的补充手段，它已经成为信息安全领域的重要研究课题。数字水印技术是在多媒体数字作品中嵌入隐藏信息，只有通过专用的检测器才能提取。随着嵌入式设备的普及，嵌入式设备采集的图像在使用无线传输时，会遭受到节点和图像数据的伪造攻击等，在嵌入式设备上对图像嵌入水印可以达到鉴别节点和数据源的真伪的目的。本文研究基于嵌入式系统的JPEG数字盲水印算法，主要工作包括： 针对嵌入式监控系统中存在的认证安全性和传输数据量大问题，提出了一种基于嵌入式系统的JPEG数字盲水印算法。算法步骤为：第一，使用Arnold变换置乱水印图像；第二，把图像分成不重叠的 的数据块；第三，首先仿三维DCT变换，然后使用QIM的嵌入算法嵌入水印，再进行反一维DCT变换；第四，量化和熵编码形成JPEG的数据。水印的提取算法的步骤为：第一，将JPEG图像进行熵解码和反量化；第二，将得到的数据块进行一维DCT变换，使用QIM的提取算法计算水印的值；第三，将提取出的数据进行周期性Arnold变换。所提出的算法将QIM与JPEG和DCT变换结合，加强了水印的不可见性和鲁棒性。 将上述算法移植到ARM11嵌入系统，该系统使用linux操作系统。采用v4l2技术通过摄像头采集图像，并把采集到的图像用所提出的算法嵌入水印并压缩成JPEG格式，通过无线网络把嵌入水印的JPEG图像发送到计算机。在计算机端可以提取JPEG图像的水印，验证数据源的真伪。计算机端把收到的JPEG数据，转换成AVI视频，提供给用户。实际应用表明，尽管所提出算法虽然对旋转攻击的抵抗性性能一般，但是对高斯噪声、椒盐噪声、裁剪攻击以及滤波攻击都具有良好抵抗性。

Digital watermarking is one of important ways to solve the digital multimedia authentication, it is one of effective complement means of traditional encryption and becomes an important research in the field of information security. Digital watermarking technology refers to embed the hidden information into digital multimedia works embedded hidden information in digital multimedia works can be extracted through only the dedicated detector. With the popularization of embedded devices, they will face nodes and image forgery attacks. When image data are sent by wirless network, image watermarking can certificate nodes and data source on the embedded system. This thesis researches based on embedded system of JPEG blind digital watermarking, the major work can be summarized as following: A blind digital watermarking algorithm of JPEG based on embedded system is proposed for solving the problems of authentication and mass data transmission on the embedded monitoring system. The steps of inserting watermarking are: Firstly, watermarking image is scrambled by Anorld’s transform; Secondely, host image is devided into non-overlap data blocks; Thirdly, blocks data are executed pseudo three-dimension DCT, and inserted watermarking by QIM, and execute inverse one-deminsion DCT; Finally, blocks data are quantizated and enthropy coded to be JPEG data. The steps of extracting watermarking are: Firstly, JPEG data are entropy decoded and inverse quantizated; Secondly, blocks data are executed inverse one-dimension DCT and extracted watermarking by QIM extracting calculation; Thirdly, the extracted watermarking is conducted periodically Arnold’s transform. This algorithm combines JPEG with QIM and DCT to strengthen its invisibility and robustness. The proposed algorithm is transplanted on ARM11 embedded system that uses linux operating system. Camera acquisites image data by v4l2 technique, and are inserted watermarking by proposed algorithm to be JPEG format data that are sent to computer via wireless network. The JPEG data can be extracted watermarking and converted into AVI video on computer, and checked the true of source of data. Practical application testifys that the proposed algorithm is good for the performance of all resistance gaussian noise, salt noise, cropping attack, filtering attacks although it is not excellen for the performance of resistance spinning attack.