摘要
将水印嵌入到宿主图像的小波变换域的低频分量;利用BP神经网络的自学习、自适应的特性和一段已知序列训练神经网络,根据确定的神经网络模型可实现水印的盲提取;在神经网络的输入信号计算上提出选择邻域窗口为3*3方形窗口比十字窗口具有更好的实验效果。仿真实验结果表明该算法对常用的图像处理如JPEG压缩、剪切、加噪和滤波等攻击具有较好的鲁棒性和不可见性。
The watermark is embedded the host image low-frequency coefficients based on wavelet transform. Using the self-learning and adaptive capabilities of neural network and a segment sequences can train the neural network. While the model of neural network is made certain, the blind watermark is extracted. The calculation of input sign show that selecting the 3^*3 square window is better experimental effects than the cross window for calculating the input signals based on neural network. The experimental results show that the algorithm is robustness and imperceptibility against the original image process attacks such as JPEG compression, cropping, noise and filtering etc.
出处
《计算机工程与设计》
CSCD
北大核心
2006年第20期3858-3860,共3页
Computer Engineering and Design
关键词
数字水印
盲提取
神经网络
小波变换
鲁棒性
不可见性
digital watermark
blind extracting
neural network
wavelet transform
robustness
imperceptibility