<title>Robust and high-quality video watermarking with the use of temporal redundancy</title>

In this paper, we present a new video watermarking technique, which utilize the characteristic of temporal redundancy of video sequence to improve the quality and robustness of the watermarked sequence. The proposed watermarking technique can be combined with many existing 2D image-watermarking algorithm to take advantage of their robustness against various attacks. For every watermark bit a_i, the pseudo random sequence (-1)^ai p is added to the mid-band coefficients of a block and the complementary sequence, i.e., -(-1)^ai p is added to the same block in another frame of the insertion pair. To retrieve the embedded watermark bit, the block in the first frame is substracted by the same block of the second frame, resulting in a watermark with double the magnitude (-1)^ai p-[-(-1)^ai p] = 2(-1)^ai p. Since adjacent video frames are highly correlated and the DCT coefficients are almost the same (especially in non-moving regions), the substraction of the block pair also cancels out most of the interfering DCT coefficients originated from the host signal, such that the interference to the watermark signal from the host signal is minimized during detection. Experiment shows that the probability of receiving an error bit can be reduced or the picture quality is improved while maintaining the same probability of error