Video watermarking using quantized coefficients and motion vectors analysis

Most of the compressed domain watermarking algorithms embed the watermark in I-frames since these frames provide higher capacity. In contrast, P-frames are highly compressed and therefore have less capacity. However, P-frames occur more frequently; thus, for applicable video watermarking it is desirable to exploit P-frames too. This paper explores the capacity of P-frames for watermark embedding utilizing just the syntactic elements of the compressed video. The limited level of robustness and quality degradations are some of the main challenges in embedding the watermark in P-frames. In the proposed method, the motion information of the video sequences is analyzed to find appropriate selections; then, nonzero quantized residuals are utilized for watermark embedding. Our results show that embedding a limited number of bits with a higher degree of robustness can be performed by analyzing motion information.