Cryptography Reference
In-Depth Information
There has been little research on the determination of f , but so far there
is no known way to determine f optimally. We therefore used a two-step func-
tion of motion vector v and a deformation quantity d. The suitability of this
function was demonstrated by our experimental evaluation (see Sect. 7.3.4).
To establish the two-step function specifically, we used the motion estima-
tion results to classify the macro-blocks in each frame into two types. This is
shown in Fig. 7.4.
Static areas: Macro blocks in which the length of motion vector v is less
than threshold value T
v
and deformation quantity d is less than threshold
value T
d
: d<T
d
andv<T
v
. Objects in these areas are static.
Motion areas: Macro-blocks in whichvis not less than T
v
or d is not less
than T
d
. That is, areas where d≥T
d
orv≥T
v
. Objects in these areas
are moving or being deformed.
Deformation quantity,
d
large
Motion areas
α
(,)
ij
ds
v
T
d
,
Static areas
s
Motion vector
length
|V|
ij
small
T
V
small
large
Fig. 7.4.
Classification using motion information.
The WMIP r
i,j
is given by a two-step function:
s
i,j
,
d<
d
,v<T
v
,
r
i,j
= f (d, v,s
i,j
)=
(7.4)
α(d, v)s
i,j
,
otherwise,
where coe
cient α(d, v) is greater than 1 and increases with d or v. T
v
and T
d
were respectively set to 6 and 5000. This is based on the results of
subjective evaluation using standard video samples [15] with five different
strengths of embedded WMs. The value of α(d, v) was then set subjectively
using the T
v
and T
d
values. To set the value of α(d, v), we assume that the
value of α(d, v) is proportional to the deformation quantity, d. The length of
the motion vector,v, is as follows:
α(d, v)=α
0
+ γ
d
d + γ
v
v,
(7.5)
