Biomedical Engineering Reference
In-Depth Information
(a)
(b)
(c)
Figure 7.
(a) Original MR image. The white line in (b) is the propagated front obtained by
a classical level set speed term. It can easily leak beyond the true boundary due to the blurry
boundary. (c) Result obtained by introducing the relaxation factor into the speed term.
By introducing this relaxation factor, when the front propagates to places
where the gradient of the object boundary are close to
M
1
, but not exactly
M
1
,
the speed
F
drops to zero and the evolving curve will be stopped appropriately.
Figure 7c shows the result of solving this problem with
δ
=0
.
23.
4.2. Image Content Items
The visual content of an image includes the color, shape, texture, and spatial
layout, reflecting the human visual response directly. The image/object feature
extraction techniques have been fully developed based on these items. In typical
level set methods, the construction of a speed function mainly uses the gradient
information. However, as medical imaging techniques advance, complex content
in the image imposes new challenges to level set-based segmentation methods.
Various image content-based speed items have been introduced in order to tackle
different application problems. Here we derive a simple yet general model to
incorporate image features into the speed function, so as to endow the speed
function with more variability and better performance.
The form of the speed function can be written as [9]:
F
=
F
A
+
F
G
.By
adding the control items or another image force, it can be expressed in a general
but simple form:
F
=
A
(
F
A
+
F
G
)+
B,
(31)
A
=1
/
(1 + dist(
C
img
,C
front
))
,
(32)
−−→
dist(
C
img
,C
front
)
.
B
=
(33)
The dist() can be any appropriate distance function. Its purpose is to measure
the difference of image features between the propagating front and the propagated
areas.
B
stands for the additional forces that arise from the image content.
C
