基于神经网络的多类水印嵌入技术

Embedding Technique of Multiple Watermarks Based on Neural Network

提出了一种基于神经网络模型的两类水印嵌入算法。两类水印提取均不需要原始图像。采用信息放大技术,加强了水印的嵌入强度。实验结果表明,算法对剪切、旋转、JPEG压缩、噪声污染具有很强鲁棒性。

A neural network algorithm is presented which embeds robust and fragile watermarks into a color image simultaneously, Both of the two kinds of watermarks can be retrieved without knowing the original image. The robust performance of the embedded robust watermark is enhanced by the information amplification technique. Experimental results show that the proposed technique is very robust against some image processing operations and joint photographic experts group （JPEG） lossys compression.