Hopfield网络与HVS相结合的小波域图像水印算法

为了实现图像水印作为一种数据认证和版权保护等功能,提出一种Hopfield网络与人类视觉系统相结合的小波域图像水印算法.该算法首先对水印进行存储,将其分成互不重叠的同样大小得子块,然后将子块由二维变换成一位序列,得到Hopfield网络的输入样本集Q,并对网络节点的权值进行初始化及修正,将其模式保存到网络中;然后对水印图像多次矩阵置乱处理,并对原始图像进行3级DWT分解;最后结合人眼视觉特性和Hopfield网络的记忆上限,保证低频部分只嵌入一次水印,高频部分重复嵌入水印.实验结果表明,该算法能抵抗JPEG压缩、加噪、剪切等多种攻击,显著提高了水印图像的质量,与直接提取水印的算法相比,性能有了很大的改善.

Watermarking algorithms of wavelet image for Hopfield network with HVS combination

In order to achieve the image watermark, as a kind of data authentication, copyright protec tion, and other functions, a Hopfield net is proposed with the human visual system with the combination of wavelet domain image watermarking algorithm. The algorithm of watermark is stored in the first place, it is divided into nonoverlapping subblock of the same size, then put the subblock by twodimensional transform into a sequence, get the Hopfield net the input sample set of Q, and initialized weights of net work nodes and correction, to save the pattern to the network;And then the watermark image matrix scrambling processing for many times, and the original image is level 3 DWT decomposition ; Finally, combining the human visual characteristics, and the memory of the Hopfield net cap, guarantee the low frequency part only a watermark embedding, the high frequency part repeat embedding watermark. The experimental results show that the algorithm can resist JPEG compression, adding noise, shearing and other attacks, significantly improving the quality of the watermark image, and compared with directly ex tracted watermark algorithm, the performance has improved dramatically.