基于新闻人物图像的水印算法

A Watermarking Algorithm Based on News Character Images

现有的图像水印算法在针对新闻人物图像时没有将人物与背景区域区分,而攻击者以盗版为目的,会聚焦到人物区域。为了对人物区域进行更好保护,提出了一种基于新闻人物图像的水印算法。首先,将人物与背景区域区分,由图像各通道的四元数加权系数得到图像的三通道相关分量,对其进行离散小波变换后再得到子带相关分量;其次,使用SIFT算法找到复杂的纹理区域上的特征点,根据特征点进行水印嵌入。经过实验,2张人物图像嵌入水印后的PSNR值都在48 dB以上,人物区域的PSNR值也都超过44 dB。在进行各种常规攻击后,所提算法几乎都能完整提取出水印信息,并且在与其他算法进行对比时,拥有更好的鲁棒性。

The existing image watermarking algorithms do not distinguish the character from the background area when applied to news character images,and attackers focus on the character area for piracy purposes.In order to provide better protection for the character area,a watermarking algorithm based on news character images is proposed.First,distinguish the character from the background area.Obtain the three-channel correlation components of the image from the quaternion weighting coefficients of each channel of the image,and then obtain the subband correlation components after performing discrete wavelet transform on them.Secondly,the feature points on the complex texture area are found according to the SIFT algorithm,and the watermark is embedded according to the feature points.The experimental results indicate that the PSNR values of the four news character images embedded with watermarks are all above 48 dB,and the PSNR values in the character area are all over 44 dB.After various conventional attacks,the proposed algorithm can almost completely extract the watermark information,and has better robustness when compared with other algorithms.