392 
The correction of the digital levels of the image, 
due to atmospheric effects, especially by water 
vapor, was realized using the subtraction method 
of the dark pixel (CHAVEZ 1975). 
b. Selection of variables: 
According to the experimental spectrograms for 
hydroxide phases such as caolinite and illite 
(muscovite-sericite), absorption and maximal 
reflectance TM values were registered in the 
bands 1, 4, 5, and 7 (Fig. 2). For the case of 
limonite, the useful bands are 1, 3, 4 and 5 (Fig. 
3). 
c. Calculation of the principal components of 
the images 
H defines the principal component. Those 
bands, where theoretically the effect of 
hydroxiles within the registered spectral values 
is known, are used as input variables. The 
component F is defined in an analogous manner 
with the limonite bands. The results of the 
calculation of the principal components FI and F 
are shown in Table 2 and 3. 
Table 2: Covariances of the eigenvector for 
the principal components of hydroxides 
PCI 
PC2 
PC3 
PC4 
Band 1 
0.206 
0.204 
0.881 
-0.373 
Band 4 
0.317 
-0.904 
0.216 
0.191 
Band 5 
0.813 
0.103 
-0.391 
-0.419 
Band 7 
0.444 
0.362 
0.153 
0.805 
Table 3: Covariances of the eigenvector for 
the principal components of limonite (Fe) 
PCI 
PC2 
PC3 
PC4 
Band 1 
0.223 
0.309 
-0.573 
-0.726 
Band 3 
0.261 
0.361 
-0.574 
0.687 
Band 4 
0.359 
-0.863 
-0.353 
0.022 
Band 5 
0.868 
0.17 
0.466 
-0.029 
d. Interpretation of the principal component 
information 
The covariance data of the eigenvectors for the 
four calculated principal components gives a 
scale for the distribution of the information for 
each principal component. For example, the CPI 
is formed by a combination of covariances of the 
original bands 1, 4, 5 and 7, given by the values 
in Table 2. Analysing the extreme variations 
(positive and negative) for each component, it 
can be shown that the major absolute difference 
occurs in CP4. In terms of images processing, 
the positive values indicate pixels with maximal 
brightness and the negatives indicate very dark 
pixels. 
Since the combination of the selected tracks was 
that one, in which theoretically the extreme 
values for absorbance and emittance of the 
hydroxides were achieved, the CP4 from the 
multaviaried analysis is strongly influenced 
from the precence of these mineral group. 
For the case of limonite, the maximal absolute 
difference also appeared in CP4, but it has to be 
remembered the variation within its input bands. 
It can be concluded, that the presence of 
hydroxide minerals is emphasized in the fourth 
principal component of the bands 1, 4, 5, and 7, 
whereas the presence of limonite is pinted out in 
the same component, but calculated with the 
bands 1, 3, 4 and 5. 
e. Construction of an hydroxyl and limonite 
image 
The described calculation were realized for each 
of the image pixels (1200 x 1200). Thus, a 
virtual image of the H and F components can be 
created, formed by the principal fourth 
component of the described bands. 
Based on the calculation, two images were 
created containing information about hydroxiles 
and limonite, respectively. In theory, a new 
principal component can be formed between 
both obtained virtual images. Its result will be 
the presence of both, hydroxiles and limonites, 
shown by high pixel values. With this 
assumption, a new principal component was 
calculated, were the input variables were the 
virtual tracks H and F. Table 4 shows the 
covariance values of the eigenvectors, where the 
second principal component presents the the 
major absolute difference. According to the 
covariance values of the eigenvectors from CP2, 
the high concentrations of hydroxile and 
limonite will produce brillant colores, and low 
conce 
miner 
Table 
the p 
limon 
Fe 
H 
It can 
were < 
other 1 
combi 
them j 
an eff 
follow 
Red = 
Green 
compc 
Blue = 
As a 
abund; 
illite i 
redish 
The pi 
all hyd 
in the 
zones: 
especii 
A segi 
disting 
hydrot] 
recogn 
notable 
zones; 
along 
which 
fields v 
CONC 
1. Thi 
hyc 
of :