摘要： 为提高水印算法的抗攻击性能，提出一种基于SVD和几何矫正的鲁棒水印算法。通过量化图像每个子块中的第1个奇异值嵌入水印，利用遗传算法搜索量化步长的最优解。在提取水印时，引入几何矩和DFT变换估计几何变换参数，从而矫正测试图像。实验结果表明，该算法在抵抗普通攻击与几何攻击方面均具有较好的鲁棒性。

关键词: 数字水印, 奇异值分解, 离散傅立叶变换, 几何不变矩

Abstract: In order to improve the performance against attack of watermark algorithm, this paper presents a robust watermark algorithm based on Singular Value Decomposition(SVD) and geometrical correction. Watermark bits are embedded in first singular value of each block through quantization, the optimization of quantization step is acquired by genetic algorithm. While extracting watermark, Directe Fouriet Transformer(DFT) and geometrical invariant moment are introduced to estimate geometrical transform and rectify test image. Experimental results show the algorithm is robust to resist normal and geometrical attack.

Key words: digital watermark, Singular Value Decomposition(SVD), Directe Fouriet Transformer(DFT), geometrical invariant moment

中图分类号: 

• TP391