An Improved Spatial Spread-Spectrum Video Watermarking

Although the spatial SS (spread-spectrum) video watermarking algorithm had been proposed by many years, it is still a hot topic in watermarking community due to its good robustness. For watermarking algorithms based on SS, the embedding strength is always the key factor of robustness. Guided by the JND (Just Noticeable Distortion) model in DCT domain and the characteristics of DCT, a more reasonable embedding strength is proposed. First, all DC coefficients of DCT blocks in one frame are collected and then segmented into some sub-sequences, which makes the distribution of watermark energy more even and consequently improves its robustness. Second, the JND value sequences of the segmented DC coefficients are computed, and DCT is applied to these sequences. Third, the resulting DC coefficients are used as the embedding strengths so that higher correlation values will be obtained during the detection. At the same time, in order to resist collusion attack, the proposed algorithm will segment the video into scenes. Compared with the algorithm of using JND as the embedding strength directly, the proposed method improves the robustness effectively. In fact, it also has good robustness to several attacks, such as video compression attacks at different bit-rate, frame dropping and slightly geometric transformation and so on.