Identify regions of interest(ROI) for video watermark embedment with principle component analysis

The temporal redundancy of video provides a greater space than images for information hiding at the expense of invitation towards many forms of spatial and temporal attacks, such as frame dropping, frame averaging that are not common in images. With video, the active change of watermark placement location serves as an effective counterattack measure. In this paper, we utilize principal components of joint feature observation of video frames to robustly determine the location of watermark embedment. The approach eliminates the need of storing original sequence and is robust against common attacks.