基于排序和去噪的广义差值扩展可逆水印

Reversible Watermarking Algorithm of General Difference Expansion Based on Sort and De-noise

针对可逆水印中常用的位置图的两点不足:占用大量空间和压缩处理成本高,提出了一种不使用位置图的大容量可逆水印算法.首先将宿主图像分块;然后计算每个像素块的平滑度近似值,该值在水印信息嵌入前后保持不变;依据平滑度近似值对像素块进行稳定的升序排序;最后,根据信息隐藏量的大小,按照排序选取适当数量的像素块,滤除像素块中的差值噪声,利用广义差值扩展嵌入水印.实验结果表明在相同PNSR下,该算法的信息隐藏量高于同类算法.

The location map used in reversible watermarking has two disadvantages. Firstly, the size of the compressed location map is large. Secondly, compression of location map consumes lots of computation. A reversible watermarking scheme of high payload is proposed without location map. The host image is divided into blocks, the smooth degree of each block is computed, which is unchanged after data embedding, according to the smooth degree all blocks are sorted, noised blocks are filtered, watermark is embedded into the selected blocks with the general difference expansion. The experimental results show the proposed method can conceal more information than other methods at the same PNSR.