一种基于DWT抗几何攻击数字水印鲁棒算法

A DWT-based Watermarking Algorithm Robust to Geometric Attacks

针对大部分数字水印算法抗几何攻击能力差这一难题,提出了一种基于小波变换可抗击几何攻击算法。在水印嵌入部分:首先对宿主图像进行小波分解、嵌入水印,再通过余弦变换求得该水印图像的一个特征向量,利用该特征向量和水印信息通过HASH函数生成一个二值逻辑系列,将该序列作为密钥;水印提取部分为:对待测的图像进行小波变换,再通过余弦变换求得待测图像的一个特征向量,利用该特征向量和密钥通过HASH函数提取出水印。文章最后对水印图像进行了旋转、缩放和扭曲等几何攻击试验。试验表明,该算法有理想的抗几何攻击能力。旋转攻击的旋转角度可达40°,缩放攻击的缩放比例可达0．2,旋转扭曲攻击的扭曲角度可达100°。

Most watermarking algorithms are not robust to geometric attacks. This paper presents an algorithm based on the discrete wavelet transform （DWT） to embed and extract watermarks. A watermarking is embedded in four stages. First, the original image is decomposed using DWT. Secondly, the watermarking is embedded into the approaching sub - band image in DWT domain. Thirdly, a characteristic vector of the image is acquired using a discrete cosine transform （DCT）. Finally, a binary logic sequence, or secret key, is produced using the characteristic vector and the watermarking. The watermarking can then be extracted in three stages. First, the test image is decomposed using DWT. Secondly, a characteristic vector is obtained in the DCT domain. Finally, the watermarking is extracted using the secret key and the characteristic vector. The algorithm uses the characteristic vector with a HASH function to embed or extract the watermarking. Results show that the algorithm is robust to rotation, scaling, translation, and bending of images. Robustness against rotation is demonstrated up to 40° and against scaling down to 20% of the images＇s original size. Furthermore, robustness against whidy bend is demonstrated for angles up to 100°.