中文摘要：

提出了一种基于联想记忆神经网络Restricted Boltzmann Machine（RBM）的数字水印算法.算法利用神经网络的记忆性能,把水印图像记忆在神经网络中,即使对含有水印的图像进行高强度的攻击,该网络依然能够清晰地对水印图像进行提取,仿真显示RBM的记忆具有较强的鲁棒性,可以适用于多种数字水印算法.

英文摘要：

A kind of digital watermarking scheme based on associative memory neural network Restricted Boltzmann Machine（RBM） is proposed.In the scheme,watermarking image is memorized in neural network by using the memory performance of RBM,watermarking image can be clearly restored in RBM,even if watermarked image is strongly attacked.Simulations show that the memory performance of RBM has better robustness,and can be used in various of watermarking scheme.