Video watermarking retrievable from MPEG data without specifying the location of embedding

A new method of blind watermarking for video is proposed for retrieving watermarks from MPEG bit streams directly without decoding. This method is based on Sakazawa's method which also aims at retrieving from MPEG bit streams, but realizes embedding and retrieval without specifying the location of the DCT blocks to be embedded. This method does not care about the location of the embedded block. Moreover, since our method embeds more redundantly than Sakazawa's method, the watermarks are less visible, even if the video is enlarged. Also the proposed method is robust, especially against cropping attacks, since the location of the embedded block is not fixed. Computer simulations show the high performance of this method and also show the robustness to various types of attacks.