Image Processing Reference
In-Depth Information
Fig. 8.5
Plots of Bhattacharyya coefficient between the resultant image from the fusion of entire
data and the resultant images for various values of
κ
for the urban and the moffett
2
data using
the bilateral filtering-based fusion technique over the subsets selected using the output-based band
selection scheme
for the output-based band selection scheme
can be provided by Eq. (
8.3
) which is quite similar to the requirement for the input-
The computational requirement
W
(κ)
W
(κ)
=
γ B(κ)
+
c
E
,
(8.3)
where
, and
c
E
denotes
the amount of computation for the evaluation of conditional entropies of the succes-
sive image bands. The
B(κ)
is the number of bands selected for a given threshold
κ
factor represents the proportionality factor to account for
computational requirements of a chosen fusion technique. The only difference lies in
the fact that an intermediate fusion has been carried out first before the computation
of the conditional entropies, unlike in the input-based scheme wherein fusion fol-
lows the entropy computation. This inherently assumes that one uses an incremental
fusion technique for the output-based scheme. The incremental fusion makes use
of the resultant of the previous intermediate fusion for carrying out the next fusion
process. Thus, if
F
γ
is the intermediate fused image from fusion of upto
(
k
−
1
)
(
k
−
1
)
bands from the data using the fusion technique
F
(the actual number of bands being
fused might be much less than
), and
I
k
be
the next input band to be selected for fusion, then the output of the fusion, i.e.,
F
k
,is
obtained as
F
k
≡
F(
(
k
−
1
)
depending upon the values of
κ
. Unfortunately, the incremental fusion methods may
suffer from numerical inaccuracies, and one may have to recompute the fusion of all
selected bands after every additional selection of a band. In this scenario, the compu-
tation increases after every selected band and one may not have any computational
advantage over the normal scheme of fusion of all bands.
F
k
−
1
,
I
k
)