Cryptography Reference
In-Depth Information
Only the results for 8 Mbps are shown because of space limitations. It was not
as effective for pictures in which there was little movement and deformation.
This includes Figure 7.7(d), which are the results for Hall. This also applies
when there is much movement or deformation. Figure 7.7(e) shows the results
for Leaves. Because most parts of the frames in these pictures are either static
or moving, it is di
cult to preferentially allocate WMs to parts with motion.
For all the pictures evaluated, the quality of those watermarked using the
motion-adaptive embedding technique was better than that of those water-
marked using the conventional method.
Evaluation of WM Detection Rate
This section describes the results of evaluating the motion-adaptive technique.
It considers improvement in WM detection in comparison with rates obtained
using the motion-adaptive technique and that of the conventional method.
The rates were measured using the same picture quality after the MPEG-2
encoding.
Procedure
We evaluated the detection rate using the five pictures evaluated.
Step 1: Select the level of picture quality, L, at which the evaluation is to
be done. For example, L = 4 means Perceptible but not annoying.
Then obtain the average embedding quantity q
(L)
prop
for the motion-
adaptive technique corresponding to L. This is done by referring to
the relationship between q and picture quality shown in Fig. 7.7. Ob-
tain the corresponding average embedding quantity, q
(L)
prev
, by the con-
ventional method. For Fig. 7.7(a4), Walk at 8 Mbps, q
(4)
prop
and q
(4)
prev
were, respectively, 2.2 and 1.5
5
.
Step 2: Embed the 64-bit information into each of the 450 frames of the pic-
ture using the motion-adaptive and conventional methods. Do this in
such a way that the average embedding quantities are, respectively,
q
(4)
prop
and q
(4)
prev
. This results in two watermarked pictures with the
same quality.
Step 3: Encode and decode the watermarked pictures using an MPEG-2 hard-
ware codec.
Step 4: Using the process based on WM detection described in Sect. 7.3.3,
detect the 64-bit information in each frame of the two watermarked
pictures. In order to clarify the improvement in WM detection of
the motion-adaptive technique, we did the following.The threshold
5
For Fig. 7.7(e), Hall was at 8 Mbps, because the levels of picture quality for the
motion-adaptive and conventional methods were both greater than 4. We set q
(L)
prop
and q
(L)
prev
to 5 which is the highest evaluation value.
