Real-Time Labeling of MPEG-2 Compressed Video

Digital mass recording devices for video data will enter the consumer market soon. Service providers are reluctant to offer services in digital form because of their fears for unrestricted duplication and dissemination. Therefore adequate copy protection systems should be developed, most likely based on labeling techniques. In this paper we propose two different techniques for the real-time labeling of digital video. Both methods embed the label information directly into an MPEG compressed video bitstream. The first method embeds the label by changing variable length codes in the bitstream. The second method discards some of the high frequency DCT-coefficients of the bitstream to embed the label. The first technique is computationally less expensive than the second one, however, the second one is more robust against attacks to remove the label. We show that the label can still be extracted after MPEG re-encoding at a lower bit-rate. A: Top half of an lc-region B: Bottom half of an lc-region b: DCT block number [0..n-1] in an lc-region c: cutoff point [0..63] in an lc-region c A ,c B : cutoff point [0..63] in lc-region A or B c tmp : temporary cutoff point D: energy difference between lc-subregion A and lc-subregion B E A ,E B : energy in lc-subregion A or lc-subregion B EOB: end of block marker L: label bit string L i : i th bit out of the label bitstring l: length of label bit string lc-VLC: label bit carrying variable length code lc-region: label bit carrying region in an I-frame lc-subregion: half of label bit carrying region in an I-frame LSB: least significant bit N: dimension of a DCT block (N=8) n: number of 8x8 DCT blocks in an lc-region S(c): Subset of DCT coefficient numbers dependent on c S(c)  {0..63} T: detection threshold T m : threshold for maximum number of changes per 8x8 DCT block u: DCT-coefficient number after zigzag scanning [0..63]