A robust video watermarking based on feature regions and crowdsourcing

摘要

Video watermarking technique aims at resolving insecurity problems. Recently, many approaches have been proposed in order to satisfy the new constraints of video applications such as robustness to collusion attacks, high level of security and signature invisibility. In this paper, a new video watermarking approach based on feature regions is proposed. The originality of this approach is to use crowdsourcing technique in order to detect feature regions. First, video summary is generated. This summary is then used to detect the first type of feature regions based on crowdsourcing technique. On the other hand, mosaic is generated from original video to detect the second type of feature region browsed by the moving objects. Finally, the signature is embedded into the mosaic generated after merging these two types of feature regions using multi-frequential watermarking scheme. Experimental results have shown a high level of invisibility thanks to the efficient choice of the embedded target. Moreover, the proposed approach is robust against several attacks especially to collusion attacks.