利用矩阵奇异值分解的单向非对称性，在研究和分析奇异值分解的正交矩阵第一列系数性质的基础上，首先给出了一种基于图像小波域奇异值分解的分块自适应水印方案，将水印嵌入其中，然后通过实验分析系数的稳定性来选择最佳的修改系数，最后利用图像的局部统计特征自适应地修改阈值，并且用两个阈值严格控制系数的修改程度，使算法达到透明性和鲁棒性之间的最优平衡。大量的实验结果表明，该算法对常见的图像处理操作均有较好的鲁棒性。

SVD preserves both one-way and non-symmetric properties. Based on the research and analysis of the first column of the orthogonal matrix after SVD, we propose a block-based image adaptive watermarking scheme based on the SVD of wavelet domain, embedding the watermark in the first column of orthogonal matrix, and selecting the best embedding coefficients by analysing the stability of those coefficients. The threshold, which is used to modify the selected coefficients is decided by the statistical properties of local blocks. The modification of coeffients is controlled by these two thresholds so as to obtain the optimal tradeoff between the transparency and the robustness. A large amount of experiments show that the proposed scheme is effective and robust to common image processing operations, especially to JPEG compression.