Robust Video Watermarking Based on Residual Synchronization in the DCT Domain

Designing an efficient synchronization method for video watermarking under complex hybrid attacks is an urgent task due to the development of social networks and multimedia software. However, few existing watermarking methods can resist various complex attacks on the social network, such as scaling, frame rate conversion, and other strong hybrid attacks. To address this issue, we first analyze the relationship between the spatial and discrete cosine transform (DCT) domain and propose a robust video watermarking based on residual synchronization (RS) mechanism. In the RS mechanism, the spread-spectrum encoded RS unit and sync frames are generated according to the change rules obtained from the analysis to achieve robustness. On the other hand, the high-frequency energy average distribution (HFEAD) mechanism is designed by analyzing the effects of distortion drifts and texture features for watermark embedding to achieve good visual quality. Extensive experimental results show that the RS mechanism can effectively synchronize watermarks under various attacks, e.g., asymmetric cropping, resizing, scaling, etc., hybridized with strong recompression attacks. Moreover, the proposed scheme is resistant to screenshots and social platform transcoding.