基于小波矩和神经网络检测的鲁棒水印算法

为提高抵抗旋转和剪切攻击等的能力，提出了一种基于小波矩特征调制和神经网检测的图像水印算法。利用水印信息调制载体的低阶小波矩特征，经过二值图像中附加的模板训练的神经网络几乎能够完全恢复嵌入到图像中的水印数据。实验表明该算法具有较好的鲁棒性，能有效地抵抗剪切，旋转攻击。算法利用具有旋转不变的小波矩，提高了抵抗攻击的能力。

Robust watermark algorithm based on the wavelet moment and neural network detection

To enhance the ability of defending attacks such as rotation and cropping,an algorithm based on wavelet moment modulation and neural network detection was presented.This watermarking technique hided an imperceptible watermarking in a host image by modulating the wavelet moments of the host image,then a neural network was established to learn the relations between the host image's moment and the watermarking data,the trained artificial neural network could almost exactly restore the watermarking imbedded in the host image.The experiment results show that the algorithm can effectively resist signal processing attacks,such as filtering,compression,rotation and cropping,and the detection does not need the host image.