Biomedical Engineering Reference
In-Depth Information
4. Carry out Bayesian decision based on the above estimated parameters in
each point in the narrow band to decide adding or removing the point from
the associated region (
IterateF unc
).
5. Mark the front points to show contours in 2D. In case of 3D, regions can
be visualized by means of binary values.
6. If the functions do not reach the steady state, go to step 2; otherwise, exit.
Pseudo Code for
InitF unc
Set all level set functions to zeros at every point in the domain.
Divide the image into sub-blocks (1
...M
) equal in size based on a given
window size.
For j=1:M
Calculate the average gray level
µ
j
.
Calculate the index of the nearest mean
indx
= arg min
k
=1
,
2
,...,K
(
|
µ
j
−
µ
k
|
2
).
Set
φ
index
to 1 at every point in block
j
.
End
Pseudo Code for
P aramF unc
Set
µ
k
=0,
σ
k
=0, and
π
k
=0
∀
k
∈
[1
,K
]
For k=1:K
Set
µ
k
to the average gray level of pixels in the positive region of
φ
k
.
Set
σ
k
to the variance of pixels in the positive region of
φ
k
.
Set
π
k
to the number of pixels in the positive region of
φ
k
.
End
π
k
i
=1
π
i
Normalize priors to have the sum of one
π
k
=
.
Pseudo Code for
NarrowBandF unc
(
φ
k
)
Set the narrow-band function
NB
k
to zero at every point in space.
Mark the front points
E
.
∀
B
and
B
is the set of neighbor
points of
e
(usually, we take the horizontal and vertical ones).
e
∈
E
, set
NB
k
(
f
)=1where
f
∈
Pseudo Code for
IterateF unc
