Information Technology Reference
In-Depth Information
images. The segmentation accuracy is measured for each method using the ground
truths. The proposed method is compared with other 3 alternatives which can be
represented as follows: A1: Active appearance method described in [
32
], A2: Level
sets method described in [
33
], and A3: Shape based level sets method described
in [
29
].
The segmentation results and the average accuracy on the ESP (when the ini-
tialization is optimal) are shown in Figs.
18
and
19
, respectively. In this test, the
initial point is chosen at the center of the object of interest. The elapsed time
1
to
segment 12 ESP images is 136.2 s for A1, 194 s for A2 (with 30-pixel radius seed
size), 248 s for A3 (with 30-pixel radius seed size) and 12.8 s for the proposed
method (without the detection and VB separation parts). It should be noted that the
all results are obtained until each method reaches its possible convergence stage.
Fig. 18 Good Initialization: Segmentation comparison under (a) no noise, noise variances
(b)
r
= 0.5. (i) Initialization. The results of (ii)
A
1, (iii)
A
2, (iv)
A
3
and (v) the proposed method. (The red and yellow colors show the contour of the ground truths and
segmented regions, respectively.)
2
n
2
n
2
n
= 0.1, (c)
r
= 0.25, and (d)
r
