An Image Watermarking Algorithm Resistant to Geometric Distortion Matched Based on SIFT Feature Points

We propose a robust watermarking algorithm based on scale invariant feature transformation (SIFT) and pseudo-Zernike moment against geometric attacks. First, the SIFT algorithm is used to extract stable feature points from the carrier image. Then, the size and direction of each local feature region are adaptively determined according to the feature scale and direction. Finally, the feature region with large feature scale non-overlapping is selected from it, and the watermark is embedded into each local feature region by means of quantization modulation pseudo-Zernike moment amplitude method. The simulation results show that the proposed algorithm not only has good transparency, but also has a strong ability to resist conventional signal processing and geometric attacks.