奇异值分解是一种特殊的矩阵变换，并具有良好的性质。本文充分利用奇异值分解的特性，提出了一种新的基于块奇异值分解的量化水印算法和一种新的基于块奇异值分解的扩频水印算法。这两个算法都是通过对各个数据块的最大奇异值进行修改来嵌入水印，都可以根据待嵌入的水印信息量来调整分块的大小，算法的复杂度较低。其中的量化水印算法是含边信息的嵌入方法，可以实现盲检测。实验结果证明，基于块奇异值分解的水印算法对常规的图像处理攻击具有很好的鲁棒性，尤其是其中的量化水印算法。

Singular Value Decomposition (SVD) is a special matrix transform with very good properties. This paper makes full use of the properties of SVD, and proposes a new quantization watermarking algorithm based on block SVD and a new spread spectrum watermarking algorithm based on block SVD. Both of the two algorithms embed watermark by altering the largest singular value of each data block, thus they can adjust the block size according to the amount of the watermark to be embedded, and their complexities are very low. The quantization watermarking algorithm is a method that makes use of the side information, and it’s a blind algorithm. The experiment results show that watermarking algorithms based on block SVD are very robust to usual image manipulation attacks, especially the quantization watermarking algorithm.