抗强剪切和涂抹攻击零水印算法

A ZERO WATERMARKING ALGORITHM RESISTING STRONG CROPPING AND SMEARING ATTACKS

为了增强现有水印算法抵抗存在信息量丢失的各种攻击(如剪切、涂抹、行列移除等),提出一种基于NMF(Non-negative factorization)和Radon变换不变矩抗强剪切和涂抹攻击零水印算法。算法首先将原始图像V矩阵进行非负矩阵分解(NMF)得到基矩阵W和系数矩阵H。由NMF部分感知全局的特性可知,利用部分V矩阵图像信息和对应的系数矩阵H可以重构出完整的W矩阵;然后计算非负矩阵分解后W矩阵的Radon变换不变矩,最后利用有限个Radon变换不变矩来设计和构建零水印信息。实验结果表明:当剪切和涂抹的面积达到87.5%时,水印检测正确率为100%,同时对于加噪、滤波、JPEG压缩等攻击,该算法也具有良好的鲁棒性。

For improving the capability of existing watermarking algorithms in resisting various information lost attacks such as cropping, smearing and rows and columns moving, etc., a zero watermarking algorithm based on NMF and Radon transform invariant moments is proposed, which can resist strong cropping and smearing attacks. First, the non-negative factorization （ NMF ） is applied for translating original image matrix V to the base matrix W and the coefficient matrix H. It is known from the characteristic of NMF in perceiving the global from the local, the whole matrix W can be reconstructed by using part of image information in matrix V and the corresponding coefficient matrix H. Then, the algorithm calculates Radon transform invariant moments of matrix W derived from NMF. Finally, a finite number of Radon transform invariant moments are used to design and construct the zero watermarking information. Experimental results demonstrate that the CWDP （correct watermark detection probabilities） of the proposed algorithm is 100% when the region of shearing and cropping is 87.5%. Meanwhile, the algorithm is robust enough to some image degradation processes such as adding noise, filtering and JPEG compression.