Results and Discussions - Hyperspectral Image Fusion - page 156

(a) (b)

(c) (d)

Fig. 10.3 Results of visualization of the lunar 2 data from the AVIRIS using a the bilateral filtering-
based technique, b the Bayesian technique, c the variational technique, d the optimization-based
technique ( c 2012 IEEE, Ref. [90]), e the three band selection technique, f the piecewise linear

function technique, and g the color matching function technique

We begin with the results of fusion over the geological dataset. The geological data
contain less number of features, similar to that of the lunar 2 hyperspectral image.
For such datasets, obtaining an image with high contrast and higher details is quite
a challenging task. The fusion technique should be able to selectively extract the
necessary details from the data, and appropriately represent them in the resultant
image. The results of fusion using various techniques can be seen in Fig. 10.4 .The
bilateral filtering-based technique identifies the fine textural contents in every band,
and uses them to define the fusion weights. Therefore, the corresponding result can be
seen to have preserved a large number of finer details in Fig. 10.4 a. The result of the
Bayesian method as seen in Fig. 10.4 b appears to be quite appealing through visual
examination. The gray appearance of this image indicates similarity in contents of the
three images each fused from one-third data to form the RGB output. The variational
technique also brings out various minor details in the scene as shown in Fig 10.4 c.