Abstract: For the problem of high false-positive detection rate of a class of Singular Value Decomposition（SVD） based image watermarking scheme, an improved reference watermarking algorithm hybrid Discrete Wavelet Transform（DWT） and SVD is proposed. The host image is applied n-level DWT and selects any or all sub-bands to make reference image, and then performs SVD transform on the reference image. The gray watermarking is preprocessed by using Arnold transform and is embedded into the singular value matrix of SVD of reference image. Experimental results show that the proposed algorithm is secure and has good transparency. Compared with other schemes, it indicates that the algorithm can solve the problem of false-positive detection and robust against most common attacks.

Key words: reference watermarking, Discrete Wavelet Transform, Singular Value Decomposition, false alarm errors, robustness

摘要： 针对一类SVD域图像水印方案存在过高虚警概率的问题，提出了一种改进的混合DWT和SVD的图像参考水印算法。算法先对载体图像进行[n]层的离散小波变换，然后随机选取第[n]层的部分或全部子带形成参考子带，并对参考子带进行SVD分解；将Arnold置乱处理后的水印嵌入到SVD分解后的奇异值矩阵中。实验表明，提出的算法具有较好的透明性和安全性；与其他方案相比，解决了虚警概率问题，且对于大部分的攻击，具有更好的鲁棒性能。

关键词: 参考水印, 离散小波变换, 奇异值分解, 虚警错误, 鲁棒性