基于NSST变换与粒子群优化的鲁棒水印算法

为增强水印图像的视觉隐秘性并抵御几何失真能力,设计NSST变换耦合粒子群优化的鲁棒水印算法。提取图像对应的Y(亮度)、U(色度)、V(浓度)成分;引入非下采样Shearlet变换(nonsubsampled Shearlet transform,NSST),对Y、U和V实施分解,形成对应的低频子带,将其分割为若干个子块;建立粒子群算法的适应度函数,优化子块的嵌入强度;构建水印嵌入方法,将水印信息隐藏到NSST系数中;构建训练图像集,计算这些样本的多元极谐变换(quaternion polar harmonic transorm,QPHT)矩,从中确定8个QPHT系数;训练支持向量机,预测攻击参数;构建水印检测方法,从校正图像中提取水印。测试数据表明,较已有的鲁棒水印技术而言,所提方案具有更高的不可感知性与抵御几何攻击的能力。

Robust watermarking algorithm based on NSST transform and particle swarm optimization

To enhance the visual concealment and geometric attack ability of watermark images,a robust watermarking algorithm based on non-subsampled Shearlet transform and particle swarm optimization was proposed.The corresponding Y(brightness),U(chroma),V(concentration)components of the image were extracted.By introducing non-subsampled Shearlet transform,the low-frequency subbands of Y,U and V component were obtained,and they were divided into several sub-blocks.The fitness function of PSO was established to optimize the embedding strength of each subblock.The watermarking embedding method was constructed to hide the watermarking information into these NSST coefficients.The training image set was constructed,and the quaternion polar harmonic transformation moments of these samples were calculated to determine 8 low-order QPHT coef-ficients.The support vector machines were trained to predict the attacked parameters.The watermark detection method was constructed to extract the watermark from the corrected watermarking image.The test data show that the proposed scheme has higher imperceptibility and resistance to geometric attacks than the existing synchronous correction watermarking techniques.