An ICA and temporal segmentation composite video watermarking scheme

Since the invention of digital video, it is easy to make unlimited copies without quality loss. This leads to problems regarding copyright protection. Digital watermarking is the proposed technique to resolve these problems. An independent component analysis (ICA) and temporal video segmentation composite blind video watermarking scheme is proposed, in which the video is divided into different segments temporally according to a global histogram comparison algorithm. Frames in such segments are taken as the observations of ICA to get independent frame components (IFCs) of the segments. The watermark is embedded into these IFCs extracted by ICA according to the 4-neighboring mean scheme in the wavelet domain. Simulations show that this scheme is robust to common frame-based processing, such as frame resizing, median filtering and salt noising, and common video-based processing, including frame rate changing, frame swapping and frame dropping. In addition, the scheme avoids locating which frames and estimating how many frames are dropped during watermark detection and it is also robust to inter- and intra-collusion attacks.