论文部分内容阅读
为了进一步提高基于支持向量机(SVM)水印算法的鲁棒性,提出了一种基于复Contourlet域的SVM和Krawtchouk矩的双水印算法。首先从RGB宿主图像中提取B分量和G分量,并且充分利用Krawtchouk矩不变量的平移、旋转、缩放不变性和Krawtchouk矩良好的局部重构特性,计算B分量图像的Krawtchouk低阶矩不变量,由此构造鲁棒水印;然后对G分量图像进行两级复Contourlet分解,在其低频分量中,利用SVM建立图像尺度内的局部相关性训练模型,并根据预测结果自适应地实现数字水印图像的嵌入和提取。大量实验结果表明,本文算法不仅具有较好的不可感知性,而且对中值滤波、加性噪声和JPEG压缩之类的常规图像处理,以及缩放、旋转和剪切等几何攻击,均具有较好的鲁棒性能,其性能优于基于小波域的SVM和基于Contourlet域的SVM水印算法。
In order to further improve the robustness of the watermarking algorithm based on Support Vector Machine (SVM), a double watermarking algorithm based on complex Contourlet domain and Krawtchouk moment is proposed. Firstly, the B component and the G component are extracted from the RGB host image, and the Krawtchouk low moment invariants of the B component image are calculated by making full use of the good local reconstruction properties of Krawtchouk moment invariants, such as translation, rotation, scaling invariance and Krawtchouk moment, Then, a robust Contourlet decomposition of G component image is carried out. In its low frequency component, a local correlation training model in image scale is established by using SVM, and the digital watermarking image is adaptively implemented according to the prediction result Embed and extract. A large number of experimental results show that the proposed algorithm not only has better imperceptibility, but also has better performance on conventional image processing such as median filtering, additive noise and JPEG compression, as well as geometric attacks such as scaling, rotation and cropping The performance of which is better than SVM based on wavelet domain and SVM watermarking algorithm based on Contourlet domain.