基于主分量分析的矢量量化数字水印算法

针对矢量量化过程中码书训练复杂的缺点，基于主分量分析方法对图像进行降维，根据降维后各主分量熵和标准差的差异性对其进行分类，采用非均匀矢量量化方法生成码书。在水印嵌入过程中，将水印图像嵌入降维后主分量能量适中的码书中以提高水印图像质量，利用EENNS算法代替完全搜索算法缩短编码时间。实验结果表明，该算法在提高码书质量的同时，能有效减少码书训练时间，对JPEG压缩、剪切、旋转等图像攻击也具有较强的鲁棒性。

Vector Quantization Digital Watermark Algorithm Based on Principal Component Analysis

This paper reduces the dimensionality of the images using Principal Component Analysis(PCA) to decrease computation, and the principal components classified by the entropy and the standard deviation are quantized separately. Additionally, the watermark image is embedded in the sub-codebook which the energy of principal components is medium for enhancing images quality. The EENNS algorithm is substitutes for full search algorithm because of falling encoding time. Experimental results show that the proposed algorithm not only improves codebook quality evidently, but also decreases codebook generation time obviously. In addition, it is robust to common image processing operations, such as JPEG compression, cropping, and rotation and so on.