Cryptography Reference
In-Depth Information
based on a threshold T
DP
. The selection of T
DP
will be discussed later. For
each iā1, 2,,n, suppose the subset C
i
contains k
i
vertices denoted as
v
1
,v
2
,,v
k
i
. It is clear that the first vertex v
1
and the last vertex v
k
i
are two
adjacent feature points in set F and the other vertices are non-feature points
between them. Firstly, two vertices v
1
and v
k
i
are connected to become a line
segment L
i
. Then a distance sequence D
i
=d
2
,,d
i
k
i
ā1
is calculated,
where d
2
,,d
i
k
i
ā1
are distances from every non-feature point v
2
,,v
k
i
ā1
to the line segment L
i
. The local shape of C
i
can be directly related to set D
i
.
The middle point of the segment L
i
(denoted o
i
) is also calculated for further
use. Repeat this process for all subsets C
1
,,C
i
,,C
n
, and obtain a final
data set composed of the distance set D
o
=D
1
,,D
n
. A corresponding
middle point set O =o
1
,,o
n
will be extracted from M
o
. Set D
o
will be
used as the cover data for watermarking. The main advantage is that D
o
is
directly related with the local shapes of the original map M
o
and it is invariant
to map translation or rotation.
Figure 6.7 demonstrates the above procedure using a simple polyline as
the example.
Fig. 6.7.
Cover data extraction based on the Douglas-Peucker algorithm.
As shown in left part of the figure, the detected feature pointsp
1
,p
2
,p
3
,p
4
divide the polyline into three subsetsC
1
,C
2
,C
3
. All feature points are con-
nected in turn to form the line segmentsL
1
,L
2
,L
3
. This is shown as a dotted
line. The middle points of the segments areo
1
,o
2
,o
3
. The right part of the
Fig. 6.7 then indicates the procedure of calculating the distance by taking the
segment L
1
as the example. Suppose there are six non-feature points between
the feature points p
1
and p
2
, the distance from each non-feature point to the
line segment L
1
is calculated as D
1
=d
2
,...,d
7
. The complete calculation
of the whole polyline will obtain a distance sequence D
o
=D
1
,D
2
,D
3
.
Watermark Embedding
The watermark is embedded by changing the distributions of the cover data.
