Image Processing Reference
In-Depth Information
C3
C2
C1
C1 (brain)
C2 (ventricles + CSF)
C3 (ALD)
Figure 7.3.
The different decision zones depending on the values of
m
2
(
C
3
)
and the decision image, by taking the maximum belief over the simple hypotheses
m
1
(
C
2
)
and
C
2
), rela-
tive to
m
1
(
C
2
) and
m
2
(
C
3
), has shown that the conflict calculation does not present
a systematic evolution towards an increase or a decrease of conflict. For example, if
m
2
(
C
3
) is small, the conflict is reduced and the weight attributed to
m
1
(
C
2
) increases
(in this case, the two sources are in better agreement over
C
2
, i.e. the ventricle class).
The calculation of the decision areas (as in Figures 7.2 and 7.3) shows that if the
weights of
m
2
(
C
3
) and
m
1
(
C
2
) increase, the decision area in favor of
C
1
decreases
(when the decisions are made based on simple hypotheses). This can also be observed
in decision images: Figure 7.4 shows a close-up of the decision images in which the
ventricles and the CSF are better detected if the weight for
m
2
(
C
3
) and
m
1
(
C
2
)
increases. Differences are also apparent on the small ALD branches which are bet-
ter detected as well. It is very important to note that the areas that are different are
classified as favoring the
C
1
∪
Studying the influence of the weighting of
m
1
(
C
1
∪
C
3
) and
m
2
(
C
1
∪
C
2
hypothesis (i.e. brain or ventricles and CSF) when
Search WWH ::
Custom Search