Robust Video Watermarking Scheme in Transform Domains

Due to large amount of frames, similarity between frames and temporal attacks (frame dropping, frame averaging, frame swapping etc.), video watermarking process is more difficult than image watermarking. Current image watermarking methods are not adequate to solve these difficulties. We propose a novel uncompressed video watermarking system based on Hidden Markov Model (HMM) and Artificial Neural Network (ANN). The proposed watermarking scheme splits the video sequences into Group of Pictures (GOP) with HMM. Portions of the binary watermark will be embedded into each GOP with a selected transformation domain watermarking algorithm. For each GOP, ANN produces the optimal transformation algorithm. The embedding process is the standard additive algorithm in low and high frequencies in different transformation domains. This novel system increases the robustness against geometric and temporal attacks, and increases the quality of the watermarked video. Keywords—Watermarking, Hidden Markov Model, Neural Network, Frequency Domain, Group of Pictures.