Cryptography Reference
In-Depth Information
where coe
cients α
0
, γ
d
,andγ
v
are positive. These coe
cients were esti-
mated based on subjective evaluation, using standard video samples with five
different strengths of embedded WMs. The watermarked pictures had been
encoded using MPEG-2. The results are shown in Table 7.1; the values of γ
d
and γ
v
depend on the bit rate. They increased with a decreasing bit rate. We
used these values in our detailed evaluation (see Sect. 7.3.4).
Table 7.1.
Values of α
0
, γ
d
,andγ
V
at various bit rates.
2 Mbps
4 Mbps
6 Mbps
8 Mbps
α
0
1.07
1.12
1.02
1.41
γ
d
7.4210
−4
6.4210
−4
5.3710
−4
3.2510
−4
γ
V
0.102
0.085
0.080
0.054
7.3.3 Motion-Adaptive Embedding
Overall Structure
Our motion-adaptive embedding technique is most effective when imple-
mented in MPEG encoders because they already have the necessary motion
estimation functions. As shown in Fig. 7.5(a), the watermarking process em-
beds copyright information into each frame by using the motion information
estimated by the MPEG encoder. The watermarked frames are then encoded.
As shown in Fig. 7.5(b), the watermarking process is comprised of an embed-
ding process and an embedding control process.
The motion-adaptive embedding method would be most effective when
used in MPEG encoders as they already have the necessary motion estima-
tion functions. Figure 7.5 shows the overall system structure of the motion-
adaptive method done using a MPEG encoder. As shown in Fig. 7.5(a), the
watermarking process embeds copyright information into each frame by using
the motion information obtained from the motion estimation of the MPEG
encoder. The watermarked frames are then encoded.
Figure 7.5(b) shows the structure of the watermarking process, which has
a embedding process and an embedding control process:
Embedding process: Embeds into each frame WMs representing the copyright
information. The WM strength is determined using the r
i,j
calculated in
the embedding control process.
Embedding control process: Calculates r
i,j
, WMIP of each pixel, from the
original frames. This process consists of two subprocesses:
(a) Intra-frame analysis: Calculates s
i,j
for each frame.
(b) r
i,j
-operation: Calculates r
i,j
using s
i,j
, v,andd.
