基于EOSHHO-SIFT的混合域鲁棒图像水印算法

Hybrid Domain Robust Image Watermarking Algorithm Based on EOSHHO-SIFT

针对含水印信息的图像对几何攻击鲁棒性不强的问题,提出一种基于改进的尺度不变特征变换(Elite opposition-learning strategy based Harris hawks optimization and Scale-invariant feature transform, EOSHHO-SIFT)的图像水印算法。首先,水印预处理:对水印进行基于奇异值分解的预处理,得到需要嵌入奇异矩阵的信息;其次,载体预处理:对载体图像进行HAAR多小波变换,并在变换后的载体图像中使用EOSHHO-SIFT算法进行特征点提取;再次,水印嵌入:上述特征点作为鲁棒水印的嵌入区域,而用EOSHHO-SIFT优化后的特征点进行鲁棒水印的嵌入;最后,水印提取:利用HAAR以及EOSHHO-SIFT提取出含有水印信息的特征点,并利用此特征点所含的奇异值信息对水印进行重组。仿真结果表明,与其它不具有旋转不变特性的图像水印算法相比,算法具有较强的抗噪声、抗滤波以及抗几何攻击的能力。

Aiming at the problem that the image with watermark information is not robust enough against geometric attacks,this paper proposes an image watermarking algorithm based on Elite opposition-learning strategy based Harris hawks optimization and Scale-invariant feature transform(EOSHHO-SIFT).Firstly,the watermark was preprocessed based on singular value decomposition to obtain the information to be embedded.Secondly,the carrier image was processed through HAAR multiwavelet transform,and the EOSHHO-SIFT algorithm was used to extract the feature points on the basis of the transform,and then the robust watermark was embedded according to the specified rules.Thirdly,the feature points optimized by EOSHHO-SIFT were used to embed robust watermarking.Finally,feature points containing watermark information were extracted by HAAR and EOSHHO-SIFT,and the watermark was recombined by using this information.Simulation results show that,compared with other image watermarking algorithms without rotation invariance,this algorithm is robust against noise,filtering and geometric attacks.