A Novel Scheme of Robust and Blind Video Watermarking

A novel video watermarking scheme is proposed, in which the watermark is embedded into DCT coefficients of luminance component of the video. Firstly, geometric features of one frame is extracted by Harris detection, then the watermark composed of information part and symbol part is embedded to the subsequent video frames at the same time. Information watermark is divided into several segments which were embedded to frames separately. Every segment was embedded into direct current coefficients of certain blocks of each frame; while the symbol watermark was embedded to low frequency coefficients of the other blocks. To extract the watermark, the geometric attacks are rectified firstly with the aid of geometric features, then the frames with watermark were found by detecting the symbol watermark, so that, the information watermark will be extracted correctly. Experimental results show that the watermark is invisible and the scheme is robust to various attacks such as rotation, scaling, frame dropping, insertion, regrouping and MPEG, The scheme can satisfy real-time requirement and the watermark can be extracted blindly.