【作者】 黄达；

【导师】 车生兵；

【作者基本信息】 中南林业科技大学 ， 计算机应用技术， 2007， 硕士

【摘要】 数字水印技术是近年来国际信息安全界兴起的一个前沿研究课题,是一种十分贴近实际应用的信息隐藏技术。数字水印技术通过把水印信息嵌入到数字图像、音频、视频等多媒体数字信息中的方法,用以证明权益拥有者对其作品的所有权。同时,通过对水印信息的检测和分析保证数字信息的真实性,完整性及可靠性,从而成为保护多媒体信息版权和进行多媒体信息认证的有效手段。人眼视觉系统(Human Visual System,简称HVS)是一个非常复杂的系统。虽然至今我们仍不能完全了解其所有的特性,但是合理的利用某些人眼视觉特性已经可以有效的解决很多实际问题。为了求得最佳的视觉效果,许多数字水印算法也不可避免的要利用人眼视觉特性。如何合理有效的结合人眼视觉特性建立视觉特性模型是提高数字水印算法性能的重要环节。现存主要半脆弱水印算法均不同程度的利用人眼视觉特性来嵌入水印,但存在着透明性差、鲁棒性不够的问题。为了更好地提高半脆弱水印的透明性和鲁棒性,本文通过研究人眼视觉特性、分析目前主要的半脆弱水印算法,建立了小波变换域中人眼视觉特性模型,并在此模型基础上对待嵌入水印图像进行分类；提出了根据攻击特性嵌入水印的思想和根据实验得到的最佳水印恢复概率的概念；提出了用于参数调整的量化中心极限定理,使得在动态量化过程中嵌入的半脆弱水印能够达到最大的鲁棒性。此外,给出了逆小波变换后系数调整的方法,使得提取水印时对载体图像进行小波变换后得到的系数和当初嵌入调整时的系数保持一致,从而解决了逆小波变换后小数部分的舍入问题。实验表明,本文算法生成的载体图像透明性好,对JPEG压缩、噪声迭加和平滑滤波等常见图像处理操作,具有较强的鲁棒性,而且可嵌入的水印信息比特数达到了原始图像像素个数的1/4,并能准确定位恶意攻击的位置。更多还原

【Abstract】 Digital watermarking technology has recently been a leading research subject in international information security field. It’s an information hiding technology which is highly close to practical application. Through the method that embedding the mark into digital multimedia such as images, audios and videos, it can prove the ownership of content owners. Meanwhile, through examining and analyzing the watermarking, it can guarantee the facticity, integrality and reliability of the digital data. Digital watermarking has provided a valid solution to protect copyright and the authentication of multimedia.HVS (Human visual system) is a very complicated system. Although we are still not fully know all the features so far, rational using of certain visual features can solve many practical problems effectively. In order to pursue the best invisibility, many digital watermarking algorithms make use of the human visual features inevitably. How to integrate the human visual features to build human visual model is an important step to improve the digital watermarking algorithm. Many existing semi-fragile watermarking algorithms embed the mark based on human visual features, but they have bad invisibility and robustness.In order to improve the invisibility and the robustness of semi-fragile watermarking, through studying the human visual features and analyzing the popular semi-fragile watermarking algorithms, the paper puts forward a human visual features model in digital wavelet transform (DWT) domain and classifies images based on this model, and brings up the concept of the best restore probability of the pixel value which is from the idea of embedding watermark based on the characteristic of the attacks, and it has been adjusted in experiments. Meanwhile, it brings forward the quantized central limit theorem, which is to adjust the coefficients. These all make the embedded semi-fragile watermarking achieve the greatest robustness in the process of dynamic quantization. Besides, it gives a coefficient redressal algorithm that is how to round the decimal fraction of the coefficients after IDWT. This makes the obtained coefficients after DWT when fetch the mark and the modified one when embed the mark consistent with each other. Experimental results suggest that this algorithm leads up to a better invisibility of the carrier image, a better robustness to the image processing, such as JPEG compression, noise adding, filtering, and the other embedded information. And the bits of embeddable watermarking information account for one fourth of the number of pixel in the original image. Moreover, it can also ascertain the exact position for vicious attacks.更多还原