Cryptography Reference
In-Depth Information
where Q is the embedding quantity for frame y
(f )4
. This means that
the Q for a frame is distributed among its pixels in proportion to
r
i,j
.
Step E5: Generate watermarked frame y
′
(f )
by adding WM pattern m (=
m
i,j
∈−1, +1, 1≤i≤W, 1≤j≤Hwhich comprises
a pseudorandom array of1s) to the original frame y
(f )
. This is
dependent on embedding bit b:
m
i,j
y
(f )
i,j
+ µ
(f )
i,j
m
i,j
,
if b =1;
′
(f )
i,j
y
=
(7.7)
y
(f )
i,j
−µ
(f )
i,j
m
i,j
,
if b =0.
An example process flow of the corresponding WM detection comprises
the following steps:
Step D1: Do the following steps over the range f =1, 2,.
Step D2: Calculate the correlation value c by correlating the WM pattern m
using the watermarked image y
′
(f )
.Thatis,
1
HW
1
HW
′
(f )
m
i,j
(y
(f )
µ
(f )
c =
m
i,j
y
i,j
=
i,j
m
i,j
)
i,j
i,j
i,j
1
HW
1
HW
m
i,j
y
(f )
µ
(f )
=
i,j
.
(7.8)
i,j
i,j
i,j
Step D3: Determine the bit-value b by comparing c with the threshold value
T (> 0):
⎧
⎨
1,
if c≥T ;
0,
if c≤−T ;
b =
(7.9)
⎩
not detected,
if−T<c<T.
7.3.4 Experimental Evaluation
Subjective Evaluation of the Watermarked Picture Quality
Procedure
We evaluated the MPEG quality by using five standard motion pictures [15],
each of which had 450 frames of 720480 pixels. We used the motion-adaptive
technique and a conventional method to embed WMs into luminance set y
(f )
of the motion pictures. After MPEG-2 encoding using a software codec, we
compared the visual qualities of the resulting pictures using standard subjec-
tive evaluation [16].
4
Q is set by the system designer and is controlled by the users taking account of
the intended use.