基于分形技术与神经网络的数字水印算法

Image Digital Watermarking Algorithm Based on Fractal and Neural Networks

提出了一种基于神经网络和分形技术的频率域数字图像水印新算法。将含有版权信息的灰度图像作为水印,利用分形技术对其进行特征提取,再利用对称加密算法对特征值进行加密。利用人类视觉系统和神经网络的特性,实现了水印在原始宿主图像频率域中的自适应嵌入,从而兼顾了水印的不可见性和鲁棒性。大量仿真结果证明了该算法的有效性。

A new image digital watermarking algorithm in frequency domain based on neural network and fractal technique is proposed. The gray level image which contains the copyright information is used as watermarking. By using fractal technique, the characteristic data is obtained, and then is encrypted by utilizing the symmetric encryption algorithm. Then, by exploiting the characteristic of neural networks and Human Visual System（HVS）, a Just Noticeable Differences（JND） threshold controller is designed to ensure the adaptive embedding of watermarking in frequency domain of original host image, and thus realizes the visibility and robustness. The experimental results prove the validity of the proposed algorithm.