鲁棒性数字图像与视频水印算法研究

发布时间：2018-01-24 03:34

本文关键词： 数字水印鲁棒性几何攻击共谋攻击空时攻击　出处：《北京邮电大学》2013年博士论文　论文类型：学位论文

【摘要】：计算机和网络技术的快速发展极大地便利了数字多媒体数据的生成、存储和传播,提高了信息利用的效率,但与此同时也带来了知识产权保护方面的隐患。因此,迫切需要一种能有效保护个人权益的技术。在这种背景下,数字水印技术凸显出其重要作用。数字水印技术是信息隐藏技术一个极其重要的分支,通过在多媒体内容如图像、视频等中嵌入特定信息达到产权保护的目的,而且已经被扩展应用于多个领域。数字水印技术的技术要求较多,其中鲁棒性是非常关键的指标,然而,目前鲁棒性数字图像与视频水印技术还未得到充分的研究,有很多技术方面的难题亟待解决,给研究工作带来了挑战和机遇。本文主要围绕鲁棒性数字图像与视频水印算法的关键技术进行展开。文中针对抗信号处理攻击的鲁棒性数字图像水印算法、抗几何攻击的鲁棒性数字图像水印算法、抗共谋攻击的鲁棒性数字视频水印算法以及抗空时攻击的鲁棒性数字视频水印算法等进行阐述和研究。本文的研究工作及主要贡献如下： (1)提出了基于Contourlet变换和SVD的联合域抗信号处理攻击的鲁棒性图像水印算法 Contourlet变换具有良好的多方向性和多尺度性,可以稀疏表达纹理和亮度信息。传统的Contourlet域水印算法没有很好的结合宿主载体的自身特性,影响了算法整体性能。而图像的奇异值分解(Singular Value Decomposition, SVD)体现了其内容的内在稳定性不随一般信号处理的操作而发生重大变化,有利于增强水印系统的抗信号处理攻击能力。根据Contourlet变换系数的能量值区分宿主的纹理性,并选取方向子带作为水印嵌入区域。图像在Arnold置乱后,由其中最大的若干个Contourlet系数构建矩阵,并对其进行SVD。结合人类视觉系统(Human Visual System, HVS)特性,采用自适应的嵌入强度,水印被嵌入到其奇异值中。水印信息在Contourlet逆变换中影响到低频子带,使水印分布到整个载体的高频和低频区域。实验结果表明,该算法获得了良好的视觉不可感知性和抵抗滤波、JPEG压缩等信号处理攻击的鲁棒性。 (2)提出了基于Harris特征点的抗几何攻击的鲁棒性图像水印算法基于空间特征点可以增强图像水印的鲁棒性的思想,提出了一种基于离散余弦变换(Discrete Cosine Transform, DCT)域Harris特征点的图像水印算法。检测Harris特征点作为待选嵌入参考点,搜索其邻域确定具有最大响应值的特征点作为局部最稳定的Harris特征点,并形成方形特征区域,选取具有最大响应值的非重叠特征区域作为嵌入区域。利用改进的奇偶量化方法将水印嵌入到ZigZag扫描后的中频系数中。实验结果表明,本算法确保了较高的峰值信噪比(Peak Signal Noise Ratio, PSNR),而且对旋转、缩放等几何攻击具有鲁棒性。 (3)提出了基于空时域HVS和STDM的抗共谋攻击的鲁棒性视频水印算法帧内共谋攻击是视频水印系统经常遇到的攻击类型。传统的扩展变换抖动调制(Spread Transform Dither Modulation, STDM)基于固定量化步长,不能充分利用视频内容特征,抵抗共谋攻击的能力较差。空时联合域的掩蔽效应可以表征人眼在时间和空间方面的感知冗余,对于低于掩蔽值的视频内容,人眼通常不敏感。空时掩蔽模型不仅采用了空间对比度敏感函数、亮度掩蔽函数、对比度掩蔽函数,而且采用了时间的掩蔽函数来解释人眼对快速运动和低速运动对象的感知能力。建立了空时域的HVS模型,确定了模型参数,进而确定了自适应的STDM量化步长。实验结果表明,本算法不仅保证了较高的PSNR,而且改善了算法抵抗时间共谋攻击的鲁棒性。 (4)提出了基于3维SIFP的抗空时攻击的鲁棒性视频水印算法空时特征点代表了视频中重大的变化,在空间和时间攻击下不会发生很大改变,因此,可以用于改善视频水印算法的鲁棒性。提出了基于尺度不变特征点(Scale Invariant Feature Points, SIFP)的视频水印算法。构造了3维高斯金字塔体,并依据扩展的Hessian矩阵,提出了3维SIFP的检测方法,检测的特征点表征了三维体内的尺度不变点,而且体现了空时联合域内重大变化的内在特征。以该特征点为中心生成方形区域,该区域被转换到DCT-SVD联合域内,通过修改最大奇异值的范数将水印嵌入到分段的中频系数中。实验结果表明,该算法没有明显降低视觉质量,而且在抵抗空时域内各种攻击方面更加鲁棒,包括时间域的帧操作以及空域的缩放、JPEG压缩等。综上所述,本文以提高数字图像与视频水印算法的鲁棒性为核心,系统地研究了数字图像与视频水印中抗信号处理攻击的鲁棒性、抗几何攻击的鲁棒性、抗共谋攻击的鲁棒性、抗空时攻击的鲁棒性等关键问题。通过鲁棒性水印算法关键问题的理论建模及分析,提出了相应的解决方案,并通过仿真验证了本文算法的性能。基于以上解决方案,可以有效地改善现有数字图像与视频水印算法的鲁棒性问题。本文的研究工作对于进一步开展面向不同业务需求的鲁棒性数字图像与视频水印算法研究提供了基础,对于数字水印技术的研究发展和实际应用具有积极的推动作用。
[Abstract]:The rapid development of computer and network technology greatly facilitates the generation , storage and propagation of digital multimedia data , improves the efficiency of information utilization , but at the same time brings hidden troubles in the protection of intellectual property rights . Therefore , there is an urgent need for a technology that can effectively protect individual rights and interests . In this context , digital watermarking technology highlights its important role . Digital watermarking technology is an extremely important branch of information hiding technology . It can embed specific information in multimedia content such as image , video and so on to achieve the purpose of property protection and has been extended to many fields . This paper mainly focuses on the key techniques of robust digital image and video watermark algorithm . The robust digital image watermarking algorithm , robustness digital image watermarking algorithm , robust digital video watermarking algorithm for anti - collusion attack and robust digital video watermarking algorithm for anti - space attack are described . The research work and main contributions of this paper are as follows : ( 1 ) Robust image watermarking algorithm for joint domain anti - signal processing attack based on Contourlet transform and SVD is proposed . Contourlet transform has good multi - directional and multi - scale , can sparse express texture and brightness information . The traditional Contourlet domain watermark algorithm does not have a good combination of the characteristics of the host carrier and affects the overall performance of the algorithm . The image has its inherent stability not to change with the operation of general signal processing . The watermark is embedded in its singular value . The watermark information affects the low - frequency sub - band in the Contourlet inverse transform . The experimental results show that the algorithm achieves good visual nonperceptibility and robustness against signal processing attacks such as filtering , JPEG compression and the like . ( 2 ) A robust image watermarking algorithm based on Harris feature point for anti - geometric attacks is proposed . An image watermarking algorithm based on discrete cosine transform ( DCT ) domain Harris feature points is proposed based on spatial feature points . The Harris feature point is used as the embedded reference point to search for its neighborhood to determine the feature point with the maximum response value as the locally most stable Harris feature point , and a non - overlapping feature region with the maximum response value is selected as the embedded region . ( 3 ) The robust video watermarking algorithm based on the anti - collusion attack of the space - time domain and STDM is proposed . Traditional extended transform Dither Modulation ( STDM ) is based on fixed quantization step size , can ' t make full use of video content features and resist collusion attack . The masking effect of space - time joint domain can be used to characterize the perceptual redundancy of human eyes in time and space . ( 4 ) A robust video watermarking algorithm based on 3 - dimensional SIFP for air - time attack is proposed . This paper presents a video watermarking algorithm based on scale invariant feature points ( SIFP ) . In conclusion , this paper studies the robustness of anti - signal processing attack in digital image and video watermark , robustness of anti - geometric attack , robustness of anti - collusion attack , robustness of anti - space attack robustness and so on . Based on the above solution , this paper provides a foundation for further research on robust digital image and video watermark algorithm for different business needs .

【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TP309.7

【参考文献】