一种DCT域的最优均值量化盲图像水印算法

针对常规的图像水印算法不能很好地实现水印不可感知性和鲁棒性折中的问题,给出了一种最优均值量化盲水印算法.该算法首先对水印图像进行Arnold置乱,提高了水印的安全性；然后将载体图像进行8×8分块DCT(离散余弦变换),并根据均值量化原理将置乱后的水印嵌入到局部量化噪声值最小的图像子块中,提高了水印的不可感知性；最后水印提取时可以实现盲提取.实验结果表明,该算法对加噪、剪切、JPEG压缩、滤波等多种攻击都具有很强的鲁棒性.

Optimal mean quantization blind image watermarking algorithmbased on DCT domain

The conventional image watermarking algorithm cannot make compromises between watermark imperceptibility and robustness easily, so this paper gave an optimal mean quantization blind watermarking algorithm. Firstly, the algorithm Arnold scrambled the watermark image to increase its security, secondly, it divided the carrier image to 8×8 blocks and transferred each blocks by DCT. According to the method of mean quantization, the algorithm embedded the scrambled watermark image in the blocks with minimal quantization noise. Finally it could extract the watermark blindly. Experimental results show that the algorithm is robust against lots of attacks, such as noise adding, cropping, JPEG compression, filtering.