Signal coherent watermarking in video

Collusion approaches in general, and block replacement attacks in particular, have been demonstrated to be a major threat against the security of frame-by-frame embedding strategies in video watermarking. These attacks exploit the redundancy of the host signal to replace each signal block with another perceptually similar one taken from another location. Such an attacking approach can be enforced both at a frame level and at a block level. Two countermeasures will consequently be introduced in this paper to combat this threat. The basic idea consists in forcing the watermark to exhibit the same spatio-temporal self-similarities as the host signal either by taking into account the camera motion or by considering the neighborhood characteristics for each sample.