中文摘要：

在传统的奇异值分解（SVD）算法的基础上，提出了一种改进型的基于分块奇异值分解（Block-SVD）的新型数字水印算法。该算法不仅鲁棒性更强，可以直接处理长方阵和提取出质量更高的水印，并且在大图像中嵌入和提取水印的时间很短，可以有效克服传统SVD算法提取出的水印在对角线上有明显失真等缺点。给出了鲁棒性测试的结果及SVD与Block-SVD的时间对比分析，并将其与传统的Cox扩频方法进行比较。测试结果表明该算法的鲁棒性更好、速度更快、实用性更强。

英文摘要：

Digital watermarking is a potential technology, which can solve the problem of multimedia copyright protection etc. An algorithm based on SVD is robust against image distortion, But based on SVD, along with the enlarging of its size, more time will be spent in computing the SVs, and experiments show that the distortion in the diagonal direction is conspicuous, due to the calculation error, so we present a watermarking algorithm based on Block-SVD. In this method, the original image is divided into several blocks and we improve on watermark to add distortion compensation. Experimental results show that the watermarking algorithm performs well in robustness. At last the method based on Block-SVD is compared to the traditional Cox method, which, also shows that the new method is very robust.