论文部分内容阅读
基于图像多小波域低频系数子块的相似性,利用神经网络的学习特性,提出了新的盲数字水印算法.将宿主图像变化为多小波域,把水印加入到宿主图像多小波变化后的低频系数中.通过后向传播算法的神经网络训练出宿主图像与嵌入的水印信号之间的关系特征,利用神经网络具有学习和自适应的特性,训练后的神经网络能够完全恢复嵌入到宿主图像中的水印信息.仿真实验表明,该算法针对各种攻击具有很好的鲁棒性,特别是在水印检测时不需要原始图像.
Based on the similarity of sub-blocks of low-frequency coefficients in multi-wavelet domain of image, a new blind digital watermarking algorithm is proposed based on the learning characteristics of neural networks. The host image is transformed into a multi-wavelet domain, and the watermark is added to the multi- Coefficient.The neural network of back propagation algorithm is used to train the relationship between the host image and the embedded watermark signal.The neural network has the characteristics of learning and adaptation and the trained neural network can be completely restored to the host image The simulation results show that the proposed algorithm is robust against all kinds of attacks, especially when the original image is not needed in watermark detection.