Cryptography Reference
In-Depth Information
d = d(k
0
,l
0
) = min
k,l
d(k, l),
(7.2)
where d is the inter-frame deformation of the object, and v =(k
0
,l
0
)
is the motion vector representing the velocity of the movement of
the object. If more than one (k
0
,l
0
) gives the same minimum value,
d(k
0
,l
0
), the shortest (k
0
,l
0
) vector is then chosen as motion vector
v.
y
(
F
-1 )
y
(
F
)
Search range
Macro-block
(
b
,
b
)
(
b
,
b
)
xy
xxy
d
(
k
,
l
)
(
b
+
k
,
b
+
l
)
(
b
+
k
,
b
+
l
)
x
x
y
y
Reference frame
Current frame
Fig. 7.3.
Block-matching method.
Criteria Using Motion Information
We first define the following terminology.
WMIP r
i,j
of a pixel in a motion picture: It is a measure representing the de-
gree of imperceptibility of a luminance change at a pixel (i, j). If, for ex-
ample, the WMIP of pixel (1,0) is larger than that of pixel (2,0), a change
in luminance at pixel (1,0) is less perceptible than one at pixel (2,0).
As described in the analysis of WM imperceptibility, the element r
i,j
of a
pixel in a motion picture depends on two kinds of motion information and on
the still picture properties. Thus, r
i,j
is a function of d, v,ands
i,j
:
r
i,j
= f (d, v,s
i,j
),
(7.3)
where s
i,j
is the pixel WMIP based only on the still picture properties
3
.
3
Various methods of determining s
i,j
have been presented; we use the one reported
by Echizen and coworkers [12]. It uses the one-dimensional luminance standard
deviation to prevent contour distortion by WMs. See Ref. [12] for details.