Information Technology Reference
In-Depth Information
Region 1
Region 2
Region 3
Region 4
Region 5
D
1
xo
1
yo
1
m
1
a
1
b
1
…
D
5
xo
5
yo
5
m
5
a
5
b
5
features of the texture of
region 1
features of the texture of
region 5
Fig. 7.4.
Segmentation regions and the feature vector
This rule indicates the importance of the average gray level in region 2 (
a
2
)
to identify images from angiogram category. The rule shows that, the mean of
a
2
in images of angiogram (0.1) is different from the mean of the same feature
for the remaining images (0.43). The standard deviation of
a
2
in images from
angiogram is small (0.07) comparing with the standard deviation of the values
of
a
2
for the remaining images (0.18). These values indicates that the feature
a
2
(region 2 gray level average) has a particular behavior in images of angiogram,
contrasting with its behavior in images of other categories, evidencing that it is
a relevant feature to distinguish images from angiograms.
Table 7.3 shows the attributes selected by StARMiner. The attributes selected
by StARMiner are underlined. The attributes that are not underlined should be
eliminated from the feature vector.
The results indicate that the fractal dimension
D
feature has a very low con-
tribution to distinguish the images for this dataset (see Table 7.3). The results
also indicate that features of average gray level
a
are the most relevant to dis-
tinguish the images in categories, because none feature of this type was removed
by the feature selection process.
Table 7.3.
Attributes selected by StARMiner. The attributes selected by StARMiner
are underlined. The attributes that are not underlined should be eliminated from the
feature vector.
Region 1 Region 2 Region 3 Region 4 Region 5
D
D
D
D
D
xo
xo
xo
x0
xo
yo
yo
yo
yo
yo
m
m
m
m
m
a
a
a
a
a
b
b
b
b
b
