Image Processing Reference

In-Depth Information

1

λ
R

1

(

−

)
=

, is the same as that in

Eq. (
4.6
). Hence, the maximum achievable savings factor in this case is also the same,

i.e.

Corollary 2
:FromEq.(
4.10
), max

r

k

ln

1

−
κ

1

S
=

.ButEq.(
4.10
) does not tell us how tight is the bound.

λ
R

ln
(
1
−
κ)

1
−

Corollary 3
:If
H

(

I
r
)

is high, it can tolerate a proportionally larger perturbation

Δ

.

4.4 Experimental Results

We shall substantiate the effectiveness of the band selection scheme over the two

datasets- the urban data, and the moffett
2
data. We shall demonstrate the results using

the bilateral filtering-based fusion technique discussed in Chap.
3
as the readers are

conversant with the same.

The results show that in both datasets the fused images over the subsets of both of

the test data, selected using the conditional entropy-based technique are comparable

in quality to the resultant images generated over the fusion of entire dataset using

the same fusion technique.

Since the available bands in hyperspectral data are ordered according to wave-

length, we can employ the specific case of the entropy-based selection scheme,

and select the bands using Eq. (
4.4
). Figure
4.2
a shows the result of fusion for the

Hyperion data using the bilateral filtering-based fusion technique where only 27

selected image bands have been fused. This figure may be compared to Fig.
4.2
b

which represents the fusion of the entire hyperspectral data cube. It can be seen

Fig. 4.2
Results of fusion

of the Hyperion data applied

over a subset of bands selected

using the conditional entropy-

based approach.

was set to

0.50 and only 27 bands were

selected.
a
shows the results

of bilateral filtering-based

fusion over the selected bands.

b
shows the corresponding

result of fusion of the entire

data (© ACM 2010, Ref: [87])

κ