Image Processing Reference
In-Depth Information
Table 4.6.
Examples of distance functions on 3D space ( [ ] means the ceiling
function).
Neighborhood
d
c
(P
,
Q;
β
) = Distance between P and Q, where P = (
i, j, k
)
,
sequence
β
Q=(
l, m, n
)
,
and
β
= neighborhood sequence
1
{
6
}
d
1
=
|i − l|
+
|j − m|
+
|k − n|
2
{
18
}
d
2
= max(max(
|i − l|, |j − m|, |k − n|
)
,
[(
|
i
−
l
|
+
|
j
−
m
|
+
|
k
−
n
|
+1)
/
2])
3
{
26
}
d
3
=max(
|i − l|, |j − m|, |k − n|
)
4
{
6
,
18
}
d
4
= max(max(
|i − l|, |j − m|, |k − n|
)
,
[2(
|i − l|
+
|j − m|
+
|k − n|
+1)
/
3])
5
{
18
,
26
}
d
5
= max(max(
|
i
−
l
|
,
|
j
−
m
|
,
|
k
−
n
|
)
,
+2)
/
5])
d
6
= max(max(
|i − l|, |j − m|, |k − n|
)
,
[(
|i − l|
+
|j − m|
+
|k − n|
+2)
/
4]+
[2(
|
i
−
l
|
+
|
j
−
m
|
+
|
k
−
n
|
6
{
6
,
26
}
[(
|i − l|
+
|j − m|
+
|k − n|
+3)
/
4]
,
[2
{
max(
|i − l|
+
|j − m|, |i − l|
+
|k − n|, |j − m|
+
|k − n|
)+1
}/
3])
d
7
= max(max(
|i − l|, |j − m|, |k − n|
)
,
[(
|i − l|
+
|j − m|
+
|k − n|
+2)
/
3]+
7
{
6
,
18
,
26
}
[(
|i − l|
+
|j − m|
+
|k − n|
+4)
/
6]
,
[3
{
max(
|i − l|
+
|j − m|, |i − l|
+
|k − n|, |j − m|
+
|k − n|
)+4
}/
5])
Remark 4.18.
The length of the minimal path for a variable neighborhood
path stated in Def. 4.13 becomes the distance measure if an appropriate neigh-
borhood sequence is adopted. In 2D images, for example, the lengths of the
minimal paths for the 4-connected, the 8-connected, and the octagonal path
reduce to the 4-neighbor, the 8-neighbor, and the octagonal distance, respec-
tively. Then what condition should a neighborhood sequence satisfy in order
that the corresponding minimal path length becomes a distance measure?
The following two theorems provide solutions to this problem for a 2D image
[Yamashita84].
Theorem 4.6.
If a neighborhood sequence
β
M
=
{
b
0
,b
1
,...,b
M−
1
}
satisfies
the equation
b
i
b
i
+1
,
N
⊆N
∀
i, i
=
1
,
2
,...,M
−
1
,
(4.66)
then the length of the minimal path for the variable neighborhood path with
the above neighborhood sequence becomes a distance measure.
Theorem 4.7.
Given the neighborhood sequence
β
M
b
0
,b
1
,...,b
M−
1
}
consisting of the 4- and the 8-neighborhood, that is,
b
i
=
4
or
8
for all
i
s,
consider an arbitrary cyclic permutation of the sequence
β
M
,thatis,
=
{
β
M
=
b
K
,b
K−
1
,...,b
M−
1
,b
0
,b
K−
1
.
(4.67)
Search WWH ::
Custom Search