Biomedical Engineering Reference
In-Depth Information
Out-of-Core Segmentation Algorithm:
(1) Compute Object Characteristic Function
.Traverse interval tree to find the list
L
of active meta-cells;
.While
L
is not NULL
. Read M active meta-cells to main memory.
. Take a
me t ac e l l
. Given a grid node
p
∈
me t ac e l l
:
if
I
(
p
)
∈
[
I
1
,
I
2
] then
χ
(
p
)
=
1
(2) Extract isosurfaces.
(3) If needed, increase grid resolution. Go to step (1)
(4) Find a seed and insert it into
processing list
(5) Begin T-Surfaces model;
.While the
processing list
is not empty:
. Pop a point
p
from
processing list
. Find the corresponding meta-cell( p)
. If meta-cell( p) is not in memory, read it
. Find
I
(
p
) and
∇
I
(
p
)
. Update
p
according to Eq. (7.14)
. Call
insert neighbor s
(
p
)
.Update function
χ
.Reparameterization of T-Surfaces (Section 7.2.3)
.If the termination condition is not reached, go to (4).
We shall observe that when the grid resolution of T-surfaces is (locally)
increased in step (3), the list
L
of active meta-cells remains unchanged and the
procedure to define the Object Characteristic Function does not change. Also,
we must observe that the isosurfaces are taken over the object characteristic
function field. Thus, there are no I /O operations in step (2).
7.8 Convergence of Deformable Models
and Diffusion Methods
Despite the capabilities of the segmentation approach in Section 7.5, the pro-
jection of T-surfaces can lower the precision of the final result. Following [49],
when T-surfaces stops, we can discard the grid and evolve the model without it
avoiding errors due to the projections.
