Information Technology Reference
In-Depth Information
k
ω
=
1.
(7)
i
i
=
1
[
]
Results are reported in Fig. 24.
If we virtually cut the dendrogram by the cophenetic similarity of 0.0043, we ob-
tain the groups G1={neu4; neu4_r_180; neu1; neu3}, G2={neu2}, G=3{parrot},
G4={gan128}, G5={monroe}, and G6={cell}.
The images neu4 and the image neu4_r_180 rotated by 180 degree are grouped
into the same group, although they have completely different segmentation parame-
ters. We obtained the best result for neu4 with the parameter-set a=0.75, b= 2 and
T=1, while for neu4_r180 the best segmentation was obtained with the parameter-set
a=1, b=0.5 and T=1. By using this latter parameter-set for neu4, we would get an
undersegmented result (compare Fig. 26(c)). Overall we observed that not all the
images having the same image segmentation parameters are grouped into one group
such as neu4 and gan128.
ω
=
1/
k
1,
,
k
.
We use
∀∈
i
i
Fig. 24. Dendrogram for image description based on texture and statistical features
Search WWH ::




Custom Search