摘要
为了有效地保护和验证数字媒体作品的版权信息,提出一种区别于传统嵌入式鲁棒水印的数字水印算法.算法中所提取的特征是由区域亮度和纹理特性所组成的综合特征,其中亮度特征的提取利用了扩展盒中脑神经网络(gBSB)的聚类功能,对于区域的纹理特征,则是通过分析其直方图统计矩来提取的.仿真结果表明,该水印方案对于常规的信号处理及几何变换攻击具有良好的鲁棒性,是一种实用的、有效的数字图像版权保护水印算法.
In order to protect and validate the ownership of digital multimedia works, a novel copyright authentication watermarking scheme, contrasted with the conventional robust watermarking scheme, is proposed. The joint feature extracted in proposed scheme consists of the luminance feature and the texture feature, wherein, the clustering function of generalized brain-state-in-a-box neural network （gBSB） and histogram statistical moment are employed to describe luminance and texture respectively. Experimental results show that the proposed scheme is adequately resilient to some malicious attacks, and it is practicable and reliable to be applied to the application in copyright protection.
出处
《小型微型计算机系统》
CSCD
北大核心
2011年第4期626-631,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60873117)资助
