Multiplicative video watermarking with semi-blind maximum likelihood decoding for copyright protection

This paper presents a new watermarking technique for uncompressed video. The video signal is first divided into non-overlapping pixel cubes. The spatial and temporal characteristic of each cube are analyzed and used as a criterion for data insertion. The 2-D wavelet transform is applied on each plane of the selected cubes. Data are multiplicatively embedded in the low band wavelet coefficients. To extract the watermark, a semi-blind maximum likelihood decoder is employed. For this purpose, the statistics of the cube's first principal component are used at the decoder as side information. Results show that this scheme is robust against linear collusion, frame swapping, frame dropping, noise insertion, median filtering, and several compression attacks.