融合视觉认知机制的新型量化水印算法

发布时间：2018-04-25 02:42

本文选题：数字水印 + 扩展变换抖动调制　；参考：《山东大学》2017年硕士论文

【摘要】：信息时代,互联网日益普及,信息的交流和共享达到了空前广泛和深入的程度,信息的形式也更加繁多,从传统的纸质演变成数字多媒体(数字图像、音频、视频等)形式,人们的工作和生活方式日渐数字化和网络化。数字多媒体的出现带来了很多便利,信息的存储、复制、修改、传播等变得简单快捷,然而随之而来的盗版、侵权、恶意篡改等行为以及日益增多,合法用户的权益受到严重的损害。数字水印的出现为解决这一问题提供了十分有效的方案。作为信息隐藏领域的一个重要分类,其利用在多媒体数据中隐秘嵌入的信息,实现对多媒体内容的保护和对版权的认证,具有广阔的应用前景,也因此得到了各领域学者的广泛关注。量化水印算法作为一类经典的算法,在计算复杂度和有效性方面展现出了良好的性能。然而其局限性在于对所有载体系数采用均匀量化器,忽视了人眼视觉的掩蔽特性。为了进一步提高载体图像视觉质量和水印鲁棒性,许多研究者在水印算法中引入了视觉特性。本论文在深入分析和理解扩展变换抖动调制(STDM)水印算法的基础上,结合一种鲁棒的视觉JND亮度模型,并改进了模型中平均亮度的计算方式,提出了基于JND亮度视觉模型的STDM水印算法。在此基础上,更加全面地考虑了视觉关注特性,用视觉关注因子调制JND亮度模型,显著改善了含水印图像的视觉质量和水印的鲁棒性。论文的主要创新点和研究成果包括:1、提出了基于JND亮度模型的STDM算法(STDM-LmJND)。在STDM算法的基础上,采用一个改进的JND亮度模型,利用该模型计算出水印嵌入端和检测端的视觉冗余信息,以此决定水印嵌入的量化步长。并且通过改进该模型中平均亮度的计算方式,使该模型前后计算出的视觉冗余一致,从而保证水印的正确提取。实验结果表明,所提出的基于JND亮度模型的STDM水印方法,显著提高了含水印载体的视觉质量,并且在对抗高斯噪声、JPEG压缩、亮度倍增等信号处理攻击时表现出较好的鲁棒性。2、提出了视觉关注非线性调制的STDM-LmJND算法。在基于JND亮度模型的STDM算法基础上,更加全面地考虑人眼视觉的关注特性,利用融合先验信息的图像显著性检测方法(SDSP),计算得到图像的"关注"区域和"非关注"区域,并采用不同的调制因子调制对应的JND系数,得到更贴近人类视觉系统特征的视觉冗余信息,据此修改水印嵌入的量化步长。实验结果表明,载体图像的视觉质量有显著提升,水印的鲁棒性得到了进一步提高。3、设计并实现了基于安卓系统的互联网信息安全服务平台。在安卓平台设计并实现了一个针对图像内容保护的应用程序,实现了对图像的版权保护和完整性校验。同时调用微信、QQ、微博等社交应用的接口,便于将已保护的图像进行分享和传播。应用程序中采用了经典的STDM算法,因大量的改进算法都是在STDM算法的基础上进行,因此实现STDM算法有助于后期应用程序的扩展和更新。移动终端作为被广泛使用的信息传播平台,在其上开发此水印应用程序,对互联网资源的安全保护有直接有效的作用,对数字水印技术的推广和应用也具有极大的意义。
[Abstract]:In the information age, the Internet is becoming more and more popular. The communication and sharing of information has reached an unprecedented level of extensive and in-depth information. The form of information is also more and more. From the traditional paper to digital multimedia (digital image, audio, video and so on), people's work and living formula are becoming more and more digital and networked. The appearance of digital multimedia brings about the emergence of digital multimedia. A lot of convenience, information storage, copy, modification, transmission and so on become simple and quick, but the following piracy, infringement, malicious tampering and other behavior and increasing, the rights of legitimate users are seriously damaged. Digital watermarking has provided a very effective solution to this problem. As a information hiding field The important classification, which uses the hidden information embedded in the multimedia data, realizes the protection of the multimedia content and the authentication of the copyright, and has a broad application prospect. Therefore, it has been widely concerned by the scholars in various fields. As a class of classical algorithms, the quantization watermark algorithm has shown good results in computing complexity and effectiveness. However, the limitation lies in the use of uniform quantizer for all carrier coefficients and neglecting the masking characteristics of human eye vision. In order to further improve the visual quality and watermark robustness of the carrier image, many researchers have introduced the visual characteristics in the watermarking algorithm. In this paper, the extended transform dithering modulation (STDM) water is deeply analyzed and understood in this paper. On the basis of the printing algorithm, combining a robust visual JND luminance model and improving the calculation of average luminance in the model, a STDM watermarking algorithm based on JND luminance visual model is proposed. On this basis, the visual attention characteristics are considered more comprehensively, and the visual attention factor is used to modulate the JND luminance model, and the watermark image is significantly improved. The main innovation and research results of the image quality and watermark are as follows: 1, the STDM algorithm (STDM-LmJND) based on the JND luminance model is proposed. On the basis of the STDM algorithm, an improved JND luminance model is used to calculate the visual redundancy of the embedded end and the detection end of the watermark by using the model to determine the watermark. By improving the calculation method of the average brightness in the model, the visual redundancy calculated before and after the model is consistent, which ensures the correct extraction of the watermark. The experimental results show that the proposed STDM watermarking method based on the JND luminance model significantly improves the visual quality of the watermark carrier and has a high antagonism. When noise, JPEG compression, luminance multiplication and other signal processing attacks are better robust.2, a STDM-LmJND algorithm for visual attention nonlinear modulation is proposed. On the basis of STDM algorithm based on the JND luminance model, the attention characteristic of human eye vision is considered more comprehensively, and the image saliency detection method (SDSP) is used to fuse the prior information. The "attention" area and the "non concern" region of the image are calculated, and the JND coefficients corresponding to different modulation factors are used to obtain the visual redundancy information which is closer to the characteristics of the human visual system. Accordingly, the quantization step of the watermark is modified. The experimental results show that the visual quality of the carrier image is significantly improved and the robustness of the watermark is obtained. To further improve.3, the Internet information security service platform based on Android system is designed and implemented. An application program for image content protection is designed and implemented on the Android platform. The copyright protection and integrity verification of images are realized. Meanwhile, the interface of social applications such as WeChat, QQ and micro-blog is invoked to facilitate the protection. The application program uses the classic STDM algorithm, because a large number of improved algorithms are based on the STDM algorithm, so the implementation of the STDM algorithm helps to expand and update the later application. As a widely used information communication platform, the mobile terminal is used to develop the watermark application. The security protection of Internet resources has a direct and effective effect. It also has great significance for the popularization and application of digital watermarking technology.

【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309.7

【参考文献】