基于水印主成分的小波域数字水印方法

摘要

For the problems of severe false positive error and being difficult to balance robustness and transparency in a kind of watermark algorithm with singular value decomposition (SVD),we proposed a new idea of watermark principal components-based wavelet domain digital watermarking method.The method first acts three-level discrete wavelet transform (DWT)on original carrier image,then applies SVD on three-level approaching subgraph LL3,and selects the product of left singular matrix and singular value matrix of watermark image as the principal components of watermark,finally it determines the optimal watermark embedding coefficient by learning from fruit fly optimisation algorithm (FOA),and embeds the principal components of watermark image.Experimental results showed that this method can achieve the best balance in robustness and transparency of the watermark while eliminating false-positive problem compared with traditional SVD image watermarking algorithm.%针对一类奇异值分解水印算法中存在严重的虚警错误、鲁棒性和透明性难以平衡的问题，提出一种基于水印主成分的小波域数字水印新思路。该方法首先将三级离散小波变换（DWT）作用于原始载体图像，再对三级逼近子图 LL3运用奇异值分解（SVD），选择水印图像左奇异矩阵和奇异值矩阵的乘积作为水印主成分，最后借鉴果蝇优化算法（FOA）确定最优的水印嵌入系数，嵌入水印图像的主成分。实验结果表明，与传统的 SVD 图像水印算法相比，该方法在消除虚警问题的同时，也可使水印的鲁棒性和透明性达到最佳平衡。