基于平衡多小波与神经网络的图像水印算法研究  

Study of Image Watermarking Algorithm Based on Balance Multi-wavelet and Neural Network

作　　者：胡社教[1] 谢锦生[1] 江萍[2] 马红杰[2] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]合肥工业大学电气与自动化工程学院。安徽合肥230009

出　　处：《系统仿真学报》2008年第11期2995-2998,共4页Journal of System Simulation

摘　　要：利用图像经过CARDBAL多小波变换后,能量汇聚且平均分摊在最低分辨率子图像4个分量上的特点,提出一种改进的多小波域盲水印方法。采用比较法嵌入一幅水印图案,建立含水印图像与水印之间的关系模型,通过神经网络训练来提高水印检测的正确率。实验表明,该方法具有良好的鲁棒性,在JPEG压缩和一些图像处理下可检测到水印。Taking advantage of the feature that the energy of the image would gather and spread averagely on four components in the sub-image of the lowest resolution, an improved blind watermarking scheme in multi-wavelet domain was proposed after CARDBAL multi-wavelet transform. Using a comparative method to embed a watermark image, then establishing the relationship model between the watermarked image and the watermark, the neural network training was used to enhance the accuracy of the watermark detection. Experiments show that the scheme is robust, and the watermark can still be detected under JPEG compression and some common image processing.

关 键 词：信息隐藏 图像水印 CARDBAL2多小波 BP神经网络 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]