The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1121
Figure 3. RADARSAT-1 scene, September 1998
Since geometric distortions are normally introduced to satellite
data during acquisition due to satellite drift and Earth’s rotation,
highly accurate geometric correction is crucial before any
image processing. Therefore, the RADARSAT-1 and the
Landsat TM images were geometrically corrected and
registered. Well distributed control points were interactively
selected on both images. The coordinates of these points were
compared to determine a polynomial equation for adjustment
between them. The images were thus rectified according to the
Universal Transverse Mercator (UTM) projection. Finally, they
resampled at (15 m) resolution using cubic convolution
technique. Once the geometric correction was completed, the
images have been fused using the IHS transformations as
follows; the Landsat TM image is transformed into the IHS
perceptual color space. Since intensity is related mainly to the
brightness of the spectral responses, RADARSAT-1 image
histogram has been stretched to match the variance and average
of this computed intensity. Then, it was directly substituted and
the inverse IHS-to-RGB transformation is performed as shown
in the schematic diagram (Figure 4).
Figure 4. Schematic diagram of the fusion technique
The resulted fused image has high spatial enhanced details and
spectral parameters, because the component of intensity has
high correlation with spatial variation and the components hue
and saturation correspond with the spectral characteristics of the
image, as depicted in the false color composite (Figure 5).
Figure 5. The resulted fused image of the study area
4. RESULTS AND DISCUSSION
The previous procedures led to an excellent spectral
discrimination and interpretation of the study area. However, a
problem of slight color distortion (variation in hue) appears
after the fusion process, since the panchromatic image and the
intensity image are different. Such a problem has been reported
by many authors (Pellemans et al., 1993) and (Wang et al.,
2005). Somewhat better performance can be expected when the
available data-sets have strong correlation (Kalpoma and Kudoh,
2007). The study area is underlain by metamorphic rocks which
are represented by: i) mafic-ultramafic granulite and ii)
orthogneisses (felsic granulite), paragneisses (quartzite, quartz-
feldspar gneiss) and BIF. These rocks are intruded by younger
granitic rocks with sharp intrusive contacts. These rocks are
uncomfortably overlain by the Gilf Formation (Paleozoic), Abu
Ras Formation (Mesozoic) and Quaternary sediments (Khattab,
et al., 2002).
The most important finding of this study is the appearance of
features beneath the sand surface on the fused Landsat TM and
RADARSAT-1 images. These features are not observable at all
on the Landsat TM image of similar spatial resolution (Figure
6). Therefore, new geological and structural information were
achieved with regard to the drainage pattern, lithological and
structural features. The output fused image brought up the
buried drainage pattern (W) in a most revealing manner.
Excavations in this area indicated that the fine-grained sand
dunes in this locality have a thickness of only (0.5-1 m) and the
aridity of the sand and soils permitted radar subsurface
penetration and capturing of the feature in the returned signal.
One major NE trending drainage lines are revealed by
RADARSAT-1 data to lead into the eastern side of the study
area. The fused image also revealed some subsurface
precambrian structures such as foliations, folds (D1 and D2)
and faults (FI and F2) that control the distribution of the
gneissic rocks and associated BIF in the study area. These are