Feature Extraction from Hyperspectral Data Using ICA
213
a
f
g
h
j
Fig.8.8a-j. Components produced by the urCA-FE algorithm. a roads/towers; b soybean
clean/mowed grass; c pasture; d hay, alfalfa; e trees; f wheat
random vector u, the mutual information of its components is defined as:
I
p(u) 1
I(u!>"',U/II) =E log m
•
np(Ui)
i=l
(8.18)