一种基于小波变换和神经网络的数字水印算法

Novel Digital Watermark Scheme Based on Wavelet Transform and Neural Network

提出了一种新的基于小波变换和神经网络的数字水印算法。该算法首先把原始图像分成8×8像素的小块,根据Logistic映射产生的混沌序列选择相应的小块构造一个原始图像的子图,接着把DWT变换作用在这个子图上,得到两个子带LH1和HL1,然后把水印信息嵌入在这两个子带上,接着通过IDWT变换重构子图,最后按构成子图的顺序把每一个8×8像素的小块放回到原图中相应位置,从而得到一个嵌入了水印信息的图像。在水印的提取过程中使用了经典的BP神经网络。实验结果表明,通过该算法嵌入的水印具有较好的鲁棒性、安全性和不可感知性。

A novel digital watermark algorithm based on wavelet transform and neural network was proposed.Firstly,the original image was divided into some small blocks（8×8 pixels）.The sub-image consists of some small blocks that are selected throuth logistic map.Two sub-bands LH1 and HL1 were generated by acting DWT on the sub-image.Se-condly,the watermark was embedded into the sub-band＇s wavelet coefficients.The watermarked sub-image was reconstructed using IDWT.Finally,we put every small block（8×8 pixels） of the watermarked sub-image into the initial position of the original image.Thus,we got the watermarked image.Experiments show that the watermark embedded by this scheme has good robustness,imperceptibility and security.