【作者】 张渊；

【导师】 罗小华； 丁勇；

【作者基本信息】 浙江大学 ， 电气工程， 2012， 硕士

【摘要】 近年来,随着互联网技术、多媒体处理技术的不断发展,视频图像等视觉产品的生产、拷贝和传播变得越来越容易。如何有效地对视觉产品进行版权保护就成为了一个亟待解决的问题。数字水印作为版权保护的一种有效手段,受到了研究者越来越多的关注。通常来说,利用数字水印技术向视觉产品中注入的版权信息必须具有一定的鲁棒性,并且要求嵌入的信息对原始视觉产品的质量有较小的损耗,即通常来说一个高性能的数字水印系统要同时兼备高鲁棒性和高透明性。一般来说,如何在鲁棒性和透明性之间达到一个良好的平衡点是一个难点。本文针对该问题,提出了一种在平衡多小波域(Balanced Multiwavelet Domain)基于恰好可感知失真(Just Noticeable Distortion, JND)模型的数字水印算法。首先我们在平衡多小波域构建了一个多分辨率的JND模型,该模型包含了空间对比度敏感函数(Contrast Sensitivity Function, CSF),亮度自适应掩蔽和区分强边缘和纹理区域的对比度掩蔽。然后基于此JND模型,水印信息被自适应地嵌入到亮度层和色度层的视觉冗余区域中。实验表明该算法不仅具有较高的鲁棒性,而且水印的嵌入对图像的质量损耗也较小另外,针对传统图像质量评价方法(如PSNR)对评价数字水印系统透明性的不足,我们提出了一种基于人类视觉系统(Human Visual System, HVS)特点的全参考图像质量评价方法,作为评价视频图像处理系统性能的关键指标之一。在该算法中,我们首先利用Gabor小波对图像进行线性滤波,随后利用DNT(Divisive Normalization Transform)对其提取的特征量进行非线性变换,以此模仿HVS中相邻简单视细胞对其中心视细胞的增益抑制效应。最后我们计算参考图像与失真图像视觉能量信息之间的互信息,并以此作为图像质量的客观评价指标。实验表明,该算法所预测的图像质量与真实的图像主观质量具有很高的相关性。更多还原

【Abstract】 In the past few years, with the development of the internet and multimedia technology, the production, copy and transmission of the visual content has become more and more easily. So the copyright protection of the visual content has become a critical issue in the recent years. The digital watermarking technology which has been considered as an efficient solution to the copyright protection problem is getting more and more attention by the researchers. Generally speaking, the copyright information which to be embedded into the visual content must have high robustness. Additionally, the procedure of the information embedding should keep the quality of visual content.In general, it is a conflicting issue to make a good balance between the robustness and the transparency of the digital watermarking system. In order to solve this problem, we proposed a novel digital watermarking algorithm based on the JND (Just Noticeable Distortion) model in balanced multiwavelet domain. First, according to the properties of the human visual system, a novel multiresolution JND model is constructed in balanced multiwavelet domain. This model incorporates the spatial CSF (Contrast Sensitivity Function), the luminance self-adaptation masking, and the contrast masking effect via separating the sharp edge and the texture. Then, based on this model, the watermark is adaptively inserted into the visual redundancy areas of the luminance and chrominance components without introducing the perceivable distortions. Experimental results show that the proposed watermarking scheme is transparent and has a high robustness to various attacks.In addition, to solve the problem of the low powerful of the traditional image quality metric such as PNSR to evaluate the transparency of the digital watermarking system, we developed a novel full reference image quality assessment method using the properties of the human visual system. In the proposed scheme, the Gabor wavelet transform is first applied to the image, then DNT (Divisive Normalization Transform) is applied to the Gabor features to mimic the suppression effect of the neighboring simple cells to the center simple cell with respect to the visual response. Finally, the mutual information between the visual energy information of the reference and distorted image is used as an indicator to predict the visual quality of the image. Experiment results show that the objective quality of the images which predicted by our algorithm has a high correlation with the subjective quality of the images.更多还原