Global resynchronization-based image watermarking resilient to geometric attacks

摘要

How to resist geometric attacks effectively and improve the watermark embedding capacity are still challenging tasks. An algorithm with better capabilities to resist cropping and combined attacks, as well as having a larger embedding capacity, is proposed. Feature points from the attacked image are obtained with the Speeded-Up Robust Features (SURF) algorithm. Then, they are matched with a few feature points obtained from the watermarked image. The matching point pairs are used to estimate the affine matrix, and then the geometric attacks are corrected by the inverse affine transform. Some positioning watermarks are embedded in the spatial domain to improve the accuracy of the resynchronization. Finally, the watermark is encoded by the fountain code such that the anti-cropping performance of the algorithm can be improved. The experimental results show that the proposed algorithm not only has a larger embedding capacity but also is resistant to many kinds of attacks.