一种基于神经网络和多小波的水印新算法

A Novel Watermarking Algorithm Based on Neural Network and Multiwavelet

文章提出一种基于平衡多小波和模糊CMAC神经网络的数字水印新算法,充分利用了平衡正交多小波的正交性和对称性,不需要进行预滤波处理;而结合SISO模糊CMAC对存贮器和训练数据需求量少等优点,有效地减小了运算量。实验结果表明,该水印算法具有较强的鲁棒性和透明性。

In this paper a novel watermarking algorithm based on balanced multiwavelet transform and fuzzy CMAC neural network is proposed. In this scheme, the original image is firstly decomposed by multiwavelet transformation and the relation among subblocks is learned by FCMAC neural network. Finally a watermark is embedded into the multiwavelet domain by adjusting the relation among these subblocks. The multiwavelet transform achieves simultaneous orthogonality and symmetry without requiring any input prefiltering. Therefore, considerable reduction in computa- tional complexity is possible. The SISO fuzzy CMAC neural network can lower the demand of memory and training data, but it still maintains the same performance. The experimental results show the watermark robustness and imperceptibility of this method.