Digital watermarking for object-based compressed video

This paper describes a watermarking algorithm for MPEG4 object-based compressed video. To be compatible to conventional 8/spl times/8 DCT, MPEG-4 proposed several graylevel padding algorithms for transparent background pixels of object boundary blocks. Though conventional block-based watermarking schemes can be also applicable to these padded boundary blocks, we would like to devise a method capable of embedding data in both the interior and boundary blocks with equal robustness. In this paper, we have considered the so-called re-padding and re-quantization attacks for segmented objects. Experiments show that our algorithm obtains good robustness and is promising for integration into the MPEG-4 codec for versatile applications.