基于支持向量机与结构相似度的图像数字水印算法*

摘要

A watermark algorithm based on SVM and structural similarity is proposed in this paper to improve robustness of spatial domain watermark technology. Structural similarity is used to calculate the largest embedding strength for different image blocks and SVM is used to build model which was able to predict the relationship between image block and embedding strength. The proposed algorithm embeds and extracts watermark by modifying central pixel of blocks selected from image. The experimental results show that the proposed algorithm is robust enough to resist multiple watermark attack including JPEG compress, noise add and median iflter.%针对空域数字水印算法鲁棒性差、难以抵抗较强的攻击的问题，本文提出一种基于支持向量机SVM与结构相似度SSIM的自适应图像数字水印算法。本文利用结构相似度算法SSIM计算不同图像子块的最大水印嵌入强度，通过回归性支持向量机建立不同图像子块与最大水印嵌入强度的相关性模型，实现了根据不同图像子块预测水印嵌入强度。本文在现有基于图像邻域像素之间相关性的时空域数字水印算法的基础上，选取图像中的图像子块进行水印嵌入，通过修改子块中心位置像素值，进行水印嵌入与提取。本文提出的算法在确保水印算法具有较好的透明性的基础上，提高了水印算法的鲁棒性。实验结果表明，该算法在保持较好透明性的基础上，对于JPEG压缩、噪声、中值滤波等攻击具有较强的抵抗能力。