Adaptive Embedding and Detection for Improved Video Watermarking - Intelligent Multimedia Data Hiding - page 169

fr

y

f + 1 Frame Region Subregions:

(,,1) fr

y

f

f ,1 f ,2

y Frame

y

(,,) frk

y

fr

y

fR

(,, ) frK

y

y

Region

Fig. 7.9. An example of video partitioning.

′ (f,r,k)

Step E3: Generate each watermarked region y by adding the WM pat-

. This is a pseudo m (k)

i

tern m (k)

N
(RK)

= ∈−1, +11≤i≤

′ (f,r,k) :

random array1s and introduced into the original region y

y (f,r,k) + µ (f,r) m (k) , if b k =1,
y (f,r,k) −µ (f,r) m (k) , if b k =0,

′ (f,r,k)

y = (7.15)

where µ (f,r) is WM strength of the region y (f,r) .
Step E4: Output the watermarked frame y

′ (f ) .

WM detection

K correlation values c k s are calculated by correlating the WM pattern, m (k) ,
with the accumulated subregion.

Step D1: Do the following steps over f 0 =1,F +1, 2F +1,.
Step D2: Input F watermarked frames y

′ (f ) s(f = f 0 ,,f 0 + F−1) and
then divide FRK subregions y ′ (f,r,k) s(f = f 0 ,,f 0 + F−1,r =
1,,R,k =1,,K).
Step D3: For each k, where k =1,,K, accumulate the FR subregions
y . The y (k)

i

′ (f,r,k) s in the subregion y (k)

y (k)
i

N
(RK)

= 1≤i≤

of the accumulated subregion y (k)

is given by

f 0 +F−1

R

1
FR

y (k)
i

′ (f,r,k)
i

= y . (7.16)

f =f 0 r=1

Step D4: Calculate the set consisting of the K correlation values

c =c k 1≤k≤K. (7.17)

Each correlation value c k is obtained by correlating the WM pattern
m (k) with the accumulated subregion y (k) .Thatis,