双重功能图像水印算法

An image watermarking algorithm with dual purpose

提出一种能同时实现内容认证和版权保护双重功能的图像水印算法。首先,对原始图像进行分块奇异值分解(SVD),计算所有子块最大奇异值的均值,通过比较各子块的最大奇异值与所有子块最大奇异值的均值间的数值关系产生鲁棒零水印序列。然后,对原始图像进行分块离散余弦变换(DCT),调整图像子块DCT高频系数的数值大小,建立同一子块两个不同DCT系数间数值的线性关系,通过逆DCT(IDCT)得到系数调整后的图像。内容认证时,利用DCT的可逆性,通过判断遭篡改图像每个子块两个相应DCT系数间是否存在所设定的数值大小线性关系实现篡改检测。版权归属鉴别时,计算原始鲁棒零水印序列与从攻击图像提取的鲁棒零水印序列间的相似度作为评价的依据。实验结果表明,本文提出的算法不仅具有很好的不可见性,而且能同时实现内容认证和版权保护。

An image watermarking algorithm with dual-purpose is proposed,which can achieve both content authentication and copyright protection.Firstly,the original image was divided into non-overlapping blocks,and each block was conducted with singular value decomposition（SVD）.A robust zero-watermarking sequence was produced by judging the numerical relationship between each block′s biggest singular value and the average of biggest singular values from all blocks.Afterwards,the original image was divided into non-overlapping blocks,and each block was transformed with discrete cosine transformation（DCT）.Several high frequency coefficients were adjusted to establish linear numerical relationship between two different DCT coefficients from the same block,and the modified image was received after inverse discrete cosine transformation（IDCT）.Content authentication was accomplished by judging whether there was a linear numerical relationship between two relevant DCT coefficients from each block of the tampered image based on DCT＇s reversibility.Copyright was identified by the similarity between the original robust zero-watermarking sequence and the robust zero-watermarking sequence extracted from the attacked image.The experimental results show that the algorithm has good invisibility,and can achieve both content authentication and copyright protection.