A new shot-based video watermarking

This paper presents a method to embed a watermark in a video for recording copyright information to prevent the video from being pirated. Firstly, a video is cut into small shots, and the shots are classified with clustering algorithm in accordance with their similarity. Then, the DWT (Discrete Wavelet Transform) is applied to transform a watermark into wavelet domain, and embed the wavelet coefficients in appropriate shots according to the level of its coefficients. Up to this step, the procedure of watermark embedding is completed. On the other hand, while extracting watermark, it is also necessary to cut a video into shots and extract the frame from each shot to extract the wavelet coefficient embedded. The last step for extracting watermark is collocating all parameters to extract the original watermark by using IDWT (Inverse DWT). With the method proposed, most potion of watermark could be extracted even when only partial video shots are available. It is able to obtain not only higher NC (Normalized Coefficient) value, but also better visual quality compared to directly embedding partial watermark data into shots without adopting DWT.