In this paper, a robust watermarking algorithm in integer wavelet domain using downsampling is proposed. The innovations of this paper can be summarized as follows. First, after downsampling both the host image and watermark, each sub-watermark is embedded into the corresponding sub-host image. Second, four sub-watermarks are encrypted by pixel scrambling and fractional random transform, and the right principal component of watermark is embedded in the host image to avoid false positive problem. Third, two-level singular value decomposition and block cosine transform are performed on host image and watermark respectively, and the dual embedding of watermark is realized. Fourth, guided dynamic particle swarm optimization is used to optimize the embedding factors. The simulation results show that the proposed watermarking algorithm satisfies the requirements of robust watermarking very well. It has large capacity and strong robustness to common attacks. Moreover, compared with the existing related algorithms, this algorithm has obvious advantages.

提出了一种基于整数小波域下采样的鲁棒水印算法。本文的创新之处可以总结如下。首先，在对主机图像和水印都进行下采样之后，将每个子水印嵌入到相应的子主机图像中。其次，通过像素加扰和分数随机变换对四个子水印进行加密，并将水印的正确主成分嵌入宿主图像中，以避免误报问题。第三，分别对主图像和水印进行两级奇异值分解和块余弦变换，实现水印的双重嵌入。第四，采用导引动态粒子群算法优化嵌入因子。仿真结果表明，提出的水印算法很好地满足了鲁棒水印的要求。它具有大容量和强大的鲁棒性，可以抵御常见的攻击。而且，与现有的相关算法相比，该算法具有明显的优势。