Image Processing Reference
In-Depth Information
Algorithm 1: CA Evolution Rule
while
(
!
Converged
)
{
∀
p
∈
P
l
t
+
1
p
l
t
p
;
=
(9.1)
t
p
;
Converged
t
+
1
p
θ
=
θ
=
true
;
∀
∈
(
)
For
q
N
p
Attack Rule
(
p
,
q
)
End for
}
where
P
is the set of cells inside user input, which can be changed in the next step.
t
is the time of iteration.
l
t
p
is the current cell label (state set),
t
p
is the property func-
tion (always is strength) of the current cell. The related attack rule can be described
as:
θ
Algorithm 2: CA Attack Rule
q
t
p
If g
(
|
q
−
p
|
)
θ
>
θ
l
t
+
1
p
l
t
q
;
t
+
1
q
;
=
θ
=
g
(
|
q
−
p
|
)
θ
(9.2)
p
Converged
=
false
;
Break
;
Endif
g
is a monotonous decreasing function to guarant the iteration to converge.
g
and
θ
p
[
,
]
are bounded to
0
1
.
|
x
|
g
(
x
)=
1
−
(9.3)
max
(
C
)
where
C
is the image gray level and
is the difference of
p
and
q
. The cell status
will be updated once neighbor cells attacking successfully occurs.
Thus, CA provides a general framework derived from the evolution with simple
interactions, which is applied in medical image segmentation can extract complex
tissues of interest from two-dimensional or volume data. The pixels, image area and
volume data blocks are set as the cells can produce different types of segmentation
mechanism. Such as the proposed Vladimir's CA segmentation [27] can be used
for N-D cases. But the classic method has the shortcoming of high complexity in
computing and interaction.
|
x
|
Search WWH ::
Custom Search