Image Processing Reference
This chapter extends the discussion on information theoretic selection of specific
bands of hyperspectral image for efficient fusion. Similar to the previous band selec-
tion scheme discussed earlier in this monograph, the present scheme also makes
use of the conditional entropy as the measure of additional information. A band is
selected when it is less correlated with the corresponding intermediate output. Thus,
this scheme aims at the removal of the redundancy in the input hyperspectral data
depending on the contents of the fusion output. Quite naturally, the performance of
this scheme is dependent on the employed fusion technique, unlike in the input-based
band selection technique.
We have also discussed performance evaluation of the output-based band selec-
tion scheme using the bilateral filtering-based fusion technique. The performance
measures of the fusion results over two test datasets substantiate the efficiency of the
band selection scheme. A subjective evaluation of the results of these datasets is also
seen to be in agreement with the quantitative analysis.