3. IMAGE INTERPRETATION
The satellite data, used in this paper, are SPOT Panchromatic
image, dated 2007-10-05, preprocessing level 2A, resolution
10m. The study — image interpretation is done with ERDAS
Imagine 2011 software and it particularly concerns Gallikos and
Axios river delta areas.
Figure 2. SPOT image of the wider study area
Two portions of the above image were selected with the
software, subsetting delta area of Gallikos (figure 2, number 3)
and Axios (figure 2, number 4) rivers.
Visual image interpretation of these subsetting images followed,
on the base of geometric (shape, size, etc.) and radiometric
(tone, texture) characteristics.
Enhancement techniques (digital image processing) were
applied, next, to make the image more interpretable for the
application (Erdas Field GuideTM, 2010).
Particularly, we applied: a) point based operations (operations
dealing with a single pixel) and b) local operations (the output
value at a specific coordinate is dependent on the input values in
the neighbourhood of that same coordinate) (ASPRS, 2004).
In this way, the following were done respectively: a) histogram
equalization (non-linear contrast stretch) and brightness
inversion (opposite contrast of the original image) and b) high-
pass filtering and edge detection with Sobel operator (zero-sum
kernels convolved separately with the original image and then
combined) (Erdas Field Guide™, 2010). Below, we refer to
relevant observations and comments.
Gallikos (figure 2, number 3) is a small river discharging in
Thermaikos Gulf, figure 3. The differentiations of tone at the
mouth of the river, indicates that the water of the sea has come
in the riverbed, or in other words the riverbed at the mouth has
formed a lagoon (figure 3, L). Orthogonal pattern is identified
on the left and right of the riverbed, corresponding to cultivated
land. Tone and texture on both sides differentiate the overall
appearance. Road network is also detected in the area.
After applying point operations (figure 3a, figure 3b), we
observe that the boundaries of the cultivations in relation with
the original image (figure 3) appear more clear. The boundaries
of the lagoon also, as well road network are more discernible,
while in the mouth of the river areas of low vegetation are
detected.
Figure 3b. Gallikos river - brightness inversion
After applying local operations (figure 3c, figure 3d) linear
features are highlighted, as 3x3 high pass filtering serves as
edge enhancer without eliminating other features and the
application of Sobel operator shows points at which the
intensity
boundarie
Axios (1
in north
large bii
4).