一种基于顺序QR分解和AdaBoost分类器的盲水印算法

New blind watermarking algorithm based on sorted QR decomposition and Ada Boost classifier

提出了一种基于顺序QR分解和Ada Boost分类器的智能水印算法。对顺序QR分解并提取R矩阵的对角线元素来构造样本特征向量并嵌入水印。在水印提取过程中,利用Ada Boost分类器对待检测的样本向量进行分类。实验结果表明,提出的方法不仅能够抵抗诸如噪声、滤波和压缩的类噪声攻击,也可以抵抗类似裁剪和旋转的几何失真。

This paper presented a novel intelligent watermarking algorithm based on sorted QR decomposition and AdaBoost classifier. For embedding watermark, this algorithm decomposed a matrix by sorted QR decomposition and extracted diagonal el- ements of R matrix to construct feature vectors of samples. For extracting watermark, it used AdaBoost classifier to classify the feature vectors of detected samples. Experimental results shows that the proposed solution can not only resist noise-like attacks such as noise ,filtering and compression, but also resist geometric distortion such as cutting and rotating.