摘要
数字水印技术是多媒体安全领域的研究热点。与传统的加密/解密技术不同,它具有广泛的诸如版权保护、权利认证以及秘密通信的能力。提出了一种新的利用神经网络分类的数字水印技术。在所提出的方法中,原始图像首先被分成互不重叠的8×8块,然后利用神经网络按照纹理特征与人类视觉特性对图像块进行分类,不同的类采用不同的水印嵌入强度,以此来提高水印的鲁棒性。检测采用相关运算,不需要原始图像。实验结果表明,所提出的方法对多种攻击具有很好的鲁棒性。
Digital watermarking technology is a new research field of multimedia security. It is different from traditional encrypt/decrypt system. It has the advantages of easy implementation and capability of providing wide security services such as copyright protection, authentication and secret communication. In this paper, an intelligence robust digital watermarking algorithm using neural network is proposed. In the proposed scheme, the original image first is divided into some 8 ×8 sub-blocks,then different embedding strengths are decided by neural networks classification according to different textural features of every sub-block. The watermark detection adopts correlation technology. If the correlative value exceeds to the given limit, then there exists a watermark in the image. This new algorithm does not require the original image for watermark detection. Extensive experimental results demonstrate the high robustness of the watermarking algorithm to watermarked image degradations as additive noise, low-pass filtering, high-pass filtering and JPEG compression.
出处
《电子测量技术》
2012年第10期68-72,共5页
Electronic Measurement Technology
基金
教育部人文社会科学研究规划基金项目(No.11YJAZH040)
武汉市科技攻关计划(No.201210121023)
关键词
数字水印
神经网络
分类
纹理特征
嵌入强度
digital watermarking
neural networks
classification
textural features
embedding strength
作者简介
刘金安。男，1953年出生，学士学位，实验员，主要研究方向为信息分析、文本分析等。E—mail：liujinan@mail．ccnu．edu．cn金聪，女，1960年出生，工学博士，教授，主要研究方向数字图像处理、信息安全等。E—mail：jincong@mail．ccnu．edu．cn
