基于神经网络的自适应彩色图像盲水印算法

An Adaptive Blind Watermarking Algorithm Based on Neural Networks for Color Images

文中提出的算法是自适应地将水印嵌入,并利用神经网络实现水印的盲提取。首先将二值水印嵌入到YIQ色彩空间的Y分量小波分解的低频系数中,为达到平衡水印图像的鲁棒性和不可见性,嵌入的强度是根据人类视觉特性自适应调整至最佳;利用BP神经网络的学习和自适应的特性和一段已知序列训练神经网络,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具有较好的鲁棒性和不可见性。

In this paper, a watermaking algorithm that embedding adaptively the watermark and blindly extracting based on neural network is proposed. First, a binary watermark is embedded into low- frequeney coefficients, from which Y component in YIQ color space is decomposad, by the wavelet transform. Second, the tradeoff of the watermarked image between robusmess and imperceptibility would be available; the embedding factor is adjusted adaptively to the best value by human visual characteristics. Third, use the learning and adaptive capabilities of neural network and a segment sequences to train the neural network for the blind watermark extracting. Finally, 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 robusmess and imperceptibility against the original image process attacks.