Image Processing Reference
for the AVIRIS data compared to the Hyperion data lies in the fact that the image
size for the AVIRIS is much larger in our study.
This chapter discusses an information theoretic scheme for the selection of specific
image bands for fast visualization of hyperspectral data. The conditional entropy-
based selection scheme selects a subset of bands from the input which are mutually
less correlated. Thus, this subset retains most of the information contents in the
data, and assures a minimal degradation in the quality of the result of the fusion. As
only a fraction of the entire dataset undergoes fusion, the band-selection approach
facilitates a faster and computationally efficient fusion. The correlation pattern among
the adjacent bands in a spectrally ordered hyperspectral data enables us to accelerate
the process of band selection. A band is evaluated for the purpose of fusion by
calculating its additional information content against only the last selected band.
A performance evaluation of the band-selection technique in terms of the quality
of the subsequent results of fusion using the bilateral filtering-based fusion technique
has also been presented in support of the efficiency of the band-selection technique.