Feature-based image watermarking is a well-developed area of digital image watermarking. The key advantage of such an approach is the ability to guard against the local desynchronization attacks. However, research in this area usually suffers from the following limitations: (1) The feature points focus too much on the high contrast region and are distributed unevenly. (2) The feature point detector is sensitive to the texture region, and some noise feature points are always detected in the texture region. Based on superpixel image segmentation and quaternion wavelet transform, we propose a digital image watermarking approach which is highly robust against common image processing operations and local geometric transformations. First, the entropy rate superpixel segmentation is used to segment the host image into several homogeneous regions. Then, for each segment region, image feature points are extracted using the SIFER detector, and the affine invariant local regions are constructed adaptively. Finally, the digital watermark is embedded into the local regions by modulating the invariant modulus coefficients. Experimental results are provided to illustrate the efficiency of the proposed image watermarking, especially for noise attacks and local desynchronization attacks.

中文翻译：

---

基于超像素分割的四元小波域鲁棒图像水印

基于特征的图像水印是数字图像水印的一个发展良好的领域。这种方法的主要优点是能够防范本地去同步攻击。然而，该领域的研究通常存在以下局限性：（1）特征点过于集中在高对比度区域，分布不均匀。(2)特征点检测器对纹理区域敏感，总是在纹理区域检测到一些噪声特征点。基于超像素图像分割和四元小波变换，我们提出了一种对常见图像处理操作和局部几何变换具有高度鲁棒性的数字图像水印方法。首先，熵率超像素分割用于将宿主图像分割成几个均匀的区域。然后，对于每个分割区域，使用 SIFER 检测器提取图像特征点，并自适应地构建仿射不变局部区域。最后，通过调制不变模系数将数字水印嵌入到局部区域中。提供了实验结果来说明所提出的图像水印的效率，特别是对于噪声攻击和局部去同步攻击。