Fast Video Watermarking Scheme Based on Adaptive Tensor Block Selection
Shiqin Zhong, Jun Zhang · 2023
The traditional watermarking method regards the video as a series of video frames, and carries out the same watermark embedding operation for each frame, without considering the temporal correlation between video frames. The traditional method is always based on embedding watermark in a fixed number of frames, without considering the relationship between the contents of video segments, which has high computational complexity and large memory consumption. Therefore, we propose a video watermarking algorithm based on fast selection of embedded blocks. Firstly, we use AKAZE feature points and mean square error to segment the video adaptively, and then select the texture block as the video block to embed the watermark based on the feature block, and realize the watermark embedding by quantifying the core tensor. Experiments show that the algorithm not only improves the speed, but also has strong robustness to various attacks.