摘要
为了更有效地提取数字漫画的边缘轮廓特征,提出一种基于图像归一化和非下采样剪切波变换(NSST,non-subsampled shearlet transform)的零水印算法.首先,将数字漫画图像进行归一化处理.然后,采用NSST和离散余弦变换(DCT)提取图像的低频分量,再采用非负矩阵分解(NMF,non-negative matrix factorization)提取特征向量作为图像的特征信息;处理水印图像时,采用基于完全互补码的扩频技术和混沌置乱技术以提高算法的鲁棒性和安全性.最后,对数字漫画图像的特征信息与水印图像进行异或逻辑运算得到零水印.实验结果表明,本文算法在抵抗高斯噪声、椒盐噪声、几何变换、JPEG压缩、邻域滤波等攻击时都具有很强的鲁棒性,同时能够抵抗一定程度的剪切攻击.
A zero-watermark algorithm based on image normalization and non-subsampled shearlet transform （NSST） is purposed to extract the edge contour feature of digital cartoon more effectively. In this method, the image is processed with the method of image normalization firstly. Then, the low-frequency component is extracted with NSST and DCT. Next, the feature vector is extracted with NMF and put it as the feature information. For the watermark image, the spread spectrum technology based on complete complementary codes and chaotic scrambling technology are used to improve the robustness and security of the algorithm. Finally, the zero-watermark is derived by making the feature information of the cartoon image and the watermark image do XOR operation. The experimental results show that the method is very robust against Gaussian noise, salt pepper noise, geometric transformation, JPEG compression and neighborhood filtering attacks, and can resist on cropping attacks to some extent.
出处
《延边大学学报（自然科学版）》
CAS
2017年第1期43-50,共8页
Journal of Yanbian University（Natural Science Edition）
基金
吉林省教育厅"十三五"科学技术研究项目(吉教科合字[2016]第249号)
国家自然科学基金资助项目(61262090)
