文本图像认证技术研究

【作者】 杜敏；

【导师】 高宝建；

【作者基本信息】 西北大学 ， 通信与信息系统， 2010， 硕士

【摘要】 20世纪末,数字多媒体的安全问题成为信息安全领域的研究热点。文本图像作为一类特殊的数字多媒体,在社会生活中具有举足轻重的地位。在当前脆弱的互联网安全机制下,某些重要文本图像的使用有很多安全隐患。因此,有必要对文本图像的认证技术进行深入研究,用以保障这些重要文本的安全使用。本文首先全面介绍并分析了当前文本图像认证领域的研究现状,然后分别对二值文本图像认证和鲁棒文本图像认证进行了深入研究,主要成果如下：(1)针对当前二值文本图像认证算法无法有效解决图像均匀区篡改认证问题,提出一种基于短数字签名的二值文本图像认证算法。该算法通过对图像的均匀区分布情况提取长度很短的数字签名,有效的解决了图像均匀区的篡改认证问题；并且,算法在保证良好不可见性的前提下,对非均匀区有75%的理论篡改检测概率。(2)经过观察和分析发现,二值文本图像16模块中心像素点翻转后并不会改变图像行、列扫描向量的游程个数。基于这一特性,设计了一种新的基于数字水印的二值文本图像认证算法,实现了二值文本图像均匀区的认证保护和较高精度的篡改定位。实验表明,算法能有效认证图像均匀区,并具有较高篡改检测率和良好不可见性。(3)针对汉字的骨架结构特性,设计了一种文本图像特征提取的新方法。实验表明,这种提取文本图像特征的方法具有一定的抗噪能力,可用于文本图像的内容认证和篡改定位。(4)改进了一种小波域水印嵌入方法,通过减少修改参数的个数来改善原算法不可见性较差这一缺点。实验表明,改进后嵌入算法的鲁棒性较原算法略有下降,但不可见性远优于原算法,可用于文本图像的版权保护和内容认证。(5)首次分析了对灰度文本图像进行内容认证的必要性,并结合(3)、(4)的结论,提出一种鲁棒文本图像内容认证方案。实验表明,该方案对灰度文本图像有50%的篡改检测概率、并具有较好不可见性和一定的鲁棒性。更多还原

【Abstract】 In the late 20th century, security issues on digital multimedia become a hot research in the information security field. As a special kind of digital media, text image plays a decisive role in social life. In the current fragile internet security mechanism, there are lots of security problems in using some important text images. Therefore, it is necessary to study text image authentication in-depth in order to protect these important texts.Firstly, the current research status about text image authentication is introduced and analyzed in the first two chapters of this paper. Then binary text images authentication and robust image authentication is studied in-depth respectively in the last three chapters. The main results to the research are as follows:(1) Because current binary text image authentication algorithm cannot effectively solve the problem that tampering authentication to the uniformity of image, a new binary text image authentication algorithm based on short digital signature is proposed. In this new algorithm, the distribution of the uniform area in the image is extracted to a short signature, which can effectively solve the issue tampering authentication to the uniformity of image. With good invisibility, the algorithm has the theoretical probability of tamper detection 75% to the non-uniform area of the image.(2) Through observing and analyzing, it is found that turning the centre pixel of The-16-Blocks does not change the number of runs to the row or column scan vectors of the binary text image. Based on this feature, a new binary text image authentication algorithm based on watermark is designed, by which the uniform area of the image can be certified and high accuracy positioning can be achieved. The experiments show that the algorithm can effectively certificate the uniform areas of the image, with a higher tampering detection rate and good invisibility.(3) Considering the characteristics of the character structure, a new text image feature extracting method is designed. Experimental results show that this image feature extracting method has a certain anti-noise ability and it can be used for text image content authenticating and tamper locating. (4) A wavelet domain watermark embedding method is improved. The poor invisibility of the original algorithm is highly improved by reducing the number of parameters to modify. Experimental results show that the robustness of the improved embedding method is slightly worse than the original one, but the invisibility is much better than the original one. The improved algorithm can be used for protecting the copyright and authenticating the content of the text image.(5) The necessity of authenticating the content of the gray text image is proposed in the paper. Combing with the conclusions of (3) and (4), a robust text image content authentication algorithm is proposed. Experiments show that the algorithm has a tampering detecting probability 50% to the gray text image, and the invisibility and robustness works well.更多还原