A novel zero-watermarking algorithm based on robust statistical features for natural images

In this paper, we develop a robust image zero-watermarking algorithm based on multiple circular statistical features. First of all, we detect stable feature points from the host image, and construct circular region partitions from the identified feature points, which contain the critical information of the host image. Second, the statistics and differences of circular region partitions are calculated, which contain both the local differences and the global information of the host image. Third, we apply discrete cosine transform (DCT) and singular value decomposition (SVD) in statistics and differences in order to construct the feature image. The stability of DCT and SVD further warrants the robustness of the algorithm. Moreover, Arnold transform is used to scramble the watermark image to enhance the security of the algorithm. Finally, we obtain the zero watermark by performing the exclusive-or (XOR) operation between the feature image and the scrambled watermark image, and the zero watermark is stored in the copyright authentication database for copyright authentication. Moreover, to combat geometric attacks, we develop a blind image correction algorithm without using the original image to accurately correct the attacked image. Numerous experiments and comparisons with the state-of-the-art (SOTA) watermarking algorithms confirm that the proposed algorithm affords good robustness against various attacks, such as filtering attacks, noise attacks, geometric attacks, and even a random combination of the above-stated attacks.