Neural Network Based Transformation Selection in Video Watermarking

Because of 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 cannot solve these difficulties. One of the important issue in digital video watermarking is the methodology selection. For one group of pictures, DCT might give better result; however, for some others, DWT might give better results. We propose a novel Artificial Neural Network (ANN) based classification system to select the best transformation method in the embedding process for the Group of Pictures (GOP). Experimental results show that transformation selection based watermarking increases the robustness against geometric attacks, and increases the quality of the watermarked video.