With the development of multimedia technology, multi-view image has become an active research topic with an increasingly urgent demand for its copyright protection. Zero-watermarking, which is a novel watermarking method known for its outstanding performance, plays an important role in digital image copyright protection. However, in contrast to the research on applying zero-watermarking to two-view stereo image, very limited work has explicitly considered its application on multi-view image. In addition, most of the algorithms can only resist common image processing attacks and cannot resist geometric attacks effectively. In this project, our main contributions are: 1) A novel polar orthogonal moment is constructed by modifying the radial basis function to tackle the numerical inconsistency problem of radial harmonic Fourier moments and exponent-Fourier moments. 2) A hypercomplex novel polar orthogonal moment is proposed for multi-view imaging system based on hypercomplex theory and the novel polar orthogonal moment, where each viewpoint is depicted as a vector in order to preserve the internal relations among various viewpoints. 3) A multi-view zero-watermarking algorithm is proposed based on the developed hypercomplex novel polar orthogonal moment to achieve multi-view image copyright protection. In this project, resistance to geometric attacks is attained by employing the novel polar orthogonal moment in the proposed method. Additionally, correlations across the viewpoint components of multi-view image are retained by applying hypercomplex theory in order to enhance the robustness of the method. This work provides a theoretical basis and guiding principles for multi-view image copyright protection.
随着多媒体技术的发展,多视点立体图像的版权保护日趋紧迫。图像零水印作为一种性能优良的新型水印方法,在图像版权保护中具有重要的作用。然而目前立体图像的零水印算法主要是针对两视点的立体图像,应用于多视点立体图像的研究寥寥无几。此外,现有算法集中于抵抗常规攻击,而未能有效抵抗几何攻击。本项目首先通过修改圆谐-傅里叶矩和指数-傅里叶矩的径向基函数,构造出一种数值稳定的新型极正交矩;然后基于超复数理论和新型极正交矩提出超复数新型极正交矩,将多视点立体图像的各视点分量作为一个矢量整体进行描述,保留了各视点分量之间的内在联系;最后基于超复数新型极正交矩提出多视点立体图像的零水印算法,实现多视点立体图像的版权保护。本项目使用新型极正交矩使算法具有抗几何攻击的能力,同时使用超复数保留了多视点立体图像各视点分量之间的相关性来增强算法的鲁棒性,为多视点立体图像的版权保护提供了充分的理论依据。
随着因特网技术与人工智能技术的日益普及与飞速发展,多媒体技术正朝着立体化、数字化等方向发展。但是,由于数字作品极容易被复制,由此带来的立体图像的安全保护问题变得日趋紧迫,本项目设计了抗几何攻击的立体图像零水印算法,有效实现了立体图像的安全保护,具有重要的学术意义和长远的市场前景。完成的主要研究工作包括:(1)图像连续正交矩方面,构造了一种新型连续正交矩—极谐-傅里叶矩;基于高斯数值积分和小波数值积分构造了连续正交矩的精确计算方法;将整数阶连续正交矩扩展至分数阶,构造了分数阶连续正交矩。(2)超复数连续正交矩的构造方面,将超复数理论与连续正交矩结合,构造了多种超复数连续正交矩,包括面向彩色图像的三元数伪切比雪夫-傅里叶矩,面向灰度立体图像的三元数分数阶连续正交矩,以及面向彩色立体图像的八元数连续正交矩等。(3)基于连续正交矩的图像零水印技术方面,构造出面向平面图像和立体图像的多种图像零水印技术,提高了零水印算法抵抗各种攻击的能力,实现了图像的无损版权保护。本项目已录用发表期刊论文34篇,其中SCI论文30篇(入选热点论文1篇,高被引论文4篇,IEEE Transactions论文3篇),CCF A类中文期刊论文3篇;发表会议论文2篇;授权发明专利2项,申请发明专利11项;授权软件著作权3项。
{{i.achievement_title}}
数据更新时间:2023-05-31
农超对接模式中利益分配问题研究
正交异性钢桥面板纵肋-面板疲劳开裂的CFRP加固研究
内点最大化与冗余点控制的小型无人机遥感图像配准
基于多模态信息特征融合的犯罪预测算法研究
栓接U肋钢箱梁考虑对接偏差的疲劳性能及改进方法研究
基于正交图像矩的不完全投影数据CT图像重建研究
基于局部四元数极谐变换的鲁棒彩色图像水印方法研究
超复数描述下的彩色超声图像高性能压缩方法研究
四极矩气体修饰构建新型超微孔活性炭吸附剂用于甲烷/氮气高效分离