International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
hydrogeological unit and the polje, with a mean annual 
discharge of 5m?/sec. The water is coming out from a large and 
imposing cave with a tube like shape (karstic conduit). This 
conduit connects the shallow holes and the spring. There are 
two other springs (Fig.2,7, S2, S3) at the elevation between 
600-800m with mean annual discharge «llit/sec (Dimadi, 
1988). 
4. DEM GENERATION AND ORTHO-RECTIFICATION 
OF THE IMAGE 
For the creation of DTM, two topographic maps, at scale 
1:50,000 and in UTM projection, were used (Tsakiri, et al., 
2003). The image was orthorectified by using the generated 
DTM and 25 control points. The accuracy of the 
orthorectification was RMS=0,66 pixels. 
5. SPECTRAL CONTENT INFORMATION 
The six Landsats TM bands cover the visible region of 
spectrum, VIS (TMI, TM2 and TM3) the near infrared, NIR 
(TM4) and the middle infrared, MIR (TM5, TM7). In these 
regions of spectrum the rocks have different spectral 
characteristics. The MIR is characterized with high reflectance 
values for most of the rocks at the spectral range covered by 
TMS and by strong absorption features for clays and micas at 
the TM7. Even if the TM bands do not allow precise 
discrimination of the mineral content in soils or rocks, it may be 
possible to detect some groups of minerals. The differences in 
the reflection can be duc to the different lithological 
composition of bared soils that is created for example from the 
surface humidification of granite or micas chist slate. It is 
obvious that the soil on micas chist contains more micas and 
clays than soils on the granite with consequently the intense 
absorption (Yésou, et al. 1993; Tsakiri et al., 2003). 
The comparison of the NIR and VIS provides the bare soils and 
vegetation discrimination. In the study area, the rocks are 
marbles, which are constituted from calcite that approaches the 
90%, dolomitic limestone with calcite in 8% and dolomite in 
92% and granodiorites mainly with feldspars (intense 
kaolinited) and quartz. As a result they have different spectral 
signatures. It is known that the spectral signatures allow the 
discrimination of the objects. The comparison of the bands in 
the near infrared and in the visible region provides the 
discrimination between the bare soil and the vegetation, as it is 
well known. 
6. VISUALIZATION OF BAND COMBINATION 
The six TM bands provide 20 possible combinations per three, 
for color display. To get an optimal color composite, the three 
bands used, must have the minimum redundancy that is to have 
the minimum correlation coefficient. Different ways exist to 
determinate of most optimal combination, which can be or 
empirical or statistical. 
6.1 Empirical procedure 
Three bands are selected based on their spectral characteristics. 
It is the most commonly used technique. This technique 
decreases the volume of possible combinations of bands. An 
optical study based on the spectral parameters and on the 
radiometric contrast of different color composition allows the 
determination of best combinations in respect of information. 
The combination TM5-4-2 allows the better perception of 
geological structures and the TM7-4-3 provides good 
discrimination between bare soil and vegetation. According to 
the international bibliography, the combinations TM7-5-1, 
TM7-5-4, TM4-3-1 and TM4-3-2 have poor contrast and do not 
facilitate the discrimination of rocks (Yésou, et al., 1993). 
These combinations for the imagery that was used in the present 
work, gave images suitable for interpretation that proves that 
the combinations arc not submitted in rules but they depend on 
local characteristics (geology, geomorphology, vegetation). In 
this study were used combinations from empiric methodology 
depending on the area. For example, for the localization of the 
granodiorite was used the combination TM7-5-1. More 
frequency was used for the combination TM4-5-2 (Fig. 3). 
  
  
  
  
  
Figure 3. 3D Visualization of Landsat TM image, over the 
DEM, where: R=band4, G=band5 and B=band2. 
gr: granodiorite, S1: spring Aggitis, S: shallowholes area, 
WI: well 
6.2 Statistical Procedure 
The factor of most optimal index (Optimum Index Factor, OIF) 
is computed which is based on the variance and the correlation 
of bands. The Factor OIF is computed for each possible 
combination of three bands. The band combination, that has the 
bigger value in factor disposes, has the more information, and it 
is calculated according to the function (1). In Table 2 the OIF is 
presented for all the combinations of bands. In the third column 
is presented the degree where smallest corresponds in the bigger 
OIF and biggest in smallest. 
202 
  
Internatio 
EE 
where 
  
BA 
  
  
Table 2. T 
7. 
The princi 
PC PC. 
PC, gets 1 
regions (T 
from the 1 
(TMS$-7). | 
near and n 
soil (conti 
informatio 
be said tha 
The first 
compressi 
imagery L 
express le: 
used the th 
It is know 
multispect 
NDVI ind