摘要
为增强图像水印的鲁棒性,改进了一种基于支持向量机的彩色图像水印算法。主要思想是由版权所有者提供的密钥随机选取嵌入位置,依据HVS视觉掩蔽特性选取嵌入因子将Hilbert曲线置乱后的水印信息,嵌入到彩色图像的红绿蓝三个通道中。其中水印信息由参考水印和标识水印组成,参考水印在水印提取过程中只用于获取SVM训练模型,进而利用训练模型进行预测提取标识水印信息。实验结果表明该算法对剪切、JPEG压缩、噪声等图像处理操作具有较好的鲁棒性。
To enhance the robustness of the watermarking,an improved color watermarking algorithm based on Support Vector Machine is proposed in this paper.According to the embedding factor on HVS,the watermark is embedded after scrambling randomly by Hilbert to the host image position which is selected on the key providing by the copyright holder.The watermark is composed of two different parts;one is the logo watermark which is the true watermark in the watermark embedding;the other is the reference watermark which is used for choosing the feature vector to retrieval SVR training model.And the logo watermark is extracted by using a prediction technique based on support vector regression model.The experimental results show that the presented scheme is robust against various processing such as cropping,JPEG compressin,noise adding,etc.
出处
《山东农业大学学报(自然科学版)》
CSCD
北大核心
2010年第3期431-434,共4页
Journal of Shandong Agricultural University:Natural Science Edition