Abstract: Against the defects of the traditional difference expansion reversible algorithm, an reversible watermarking algorithm with Integer Wavelet Transform（IWT） and general difference expansion is proposed. The host image is divided into 4×4 non-overlapping blocks whose smooth degree is computed, which is unchanged before and after watermark embedding. According to the smooth degree, all blocks are sorted. The selected block of integer wavelet transform, watermark information is embedded in the low frequency part by using general difference expansion method and auxiliary information is embedded in the high frequency part by using watermark data compression. The experimental results show that the algorithm can increase the amount of information embedded, at the same time it can also reduce the image distortion. In addition, the proposed method can embed more information than common methods at the same [PNSR] and also has certain robustness.

Key words: reversible watermarking, difference expansion, Integer Wavelet Transform（IWT）, data compression, embedded information capacity

摘要： 针对传统差值扩展可逆水印算法的缺陷，提出了一种基于整数小波变换（IWT）和广义差值扩展的可逆水印算法。它将原始图像划分成4×4大小的互不重叠的像素块；计算各分块的近似平滑度值，该值在水印嵌入前后保持不变；将所有分块按平滑度值进行排序；根据水印信息量，依序选取适当数目的分块；对所选取分块进行整数小波变换，在变换后的低频部分利用广义差值扩展法嵌入水印信息，在高频部分通过数据压缩嵌入水印辅助信息。实验结果表明，该算法在减小图像失真的同时，提高了信息嵌入量，另外在同等水印嵌入量下，该算法的[PNSR]高于常见的一些算法。

关键词: 可逆水印, 差值扩展, 整数小波变换（IWT）, 数据压缩, 信息嵌入量