International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
limitations (Mulder, 1986). Enhancement is used to improve
the image's interpretability. Saturation enhancement, which
is a part of the colour coding, is implemented by using the
linear 3 x 3 matrix transformation on the multispectral data:
R' : R
G'|=|M'A M| \G
p B
ml M] R
m2|-— |. M2 pl u G
7 7
1 I B
The main purpose of M mapping is to transform the RGB
domain into the intensity normalized R^ G^ B' where we can
specify change in colour balance, change in hue and change
in saturation, independent of intensity. ml and m2 are
orthogonal axes in the two-dimensional colour triangle R^ G'
B? derived from the colour space (colour cube). “The colour
space is a linear space in which the mixture of two additive
colours is equivalent to a colour vector which is a linear
interpolation of the mixing colour vectors" (Mulder, 1986)
(see Figure 3).
G
i
€ i
i
i
Ber
8
Figure 3. The colour cube and the colour triangle
A is a matrix which performs an affine transformation of the
data triangle onto the corners of the colour triangle by
bringing the vegetation to green, the soil to red and the water
to blue (Figure 4).
w=water
$= soil
v=vagatation
Figure 4. Affine transformation with A matrix
M'! is the inverse of M which brings back the data from the
MI, M2 I domain into the RGB coordinate system (the colour
cube). Before the execution of the matrix multiplication,
however, parameters for colour enhancement were derived
750
from the scatter diagram on ml, m2. A two-dimensional
scatter diagram of ml and m2 was made for this purpose. The
colour triangle was drawn onto the scatter diagram (Figure
5). In this study, the A matrix implements only a saturation
enhancement (stretching and shifting on the m1, m2 axes).
Mae: and dE |
st ve que sq di
es 1 * mg "T a T T a "t €
Figure 5. A two dimensional scatter diagram of m1 and m2
The saturation enhancement is implemented by using the
resulting matrix as a multiplication factor to the RGB
coordinates of the original image to obtain the new R'G'B'
coordinates (as new feature files) of the colour enhanced
image. Figure 6 shows the scatter diagram after saturation
enhancement.
As
Ware epe pq pterea tique AT ap et
Figure 6. Scatter diagram after saturation enhancement
6. COMBINATION OF PANCHROMATIC AND
MULTISPECTRAL BANDS
Many applications of remote sensing image data require two
or more images of the same area to be represented together
for various purposes (eg, change detection). Having the high
resolution panchromatic 10 m ground resolution and the
multispectral 20 m, the basic idea was to merge them to use
the double data set to get a better image with more details.
Different experiments were carried out to illustrate this
combination:
(1) enhanced panchromatic plus the haze-corrected
multispectral with saturation enhancement of
colours;
Internat
The va
multisp
followe
combin:
Knowin
external
shadow.
intensit:
Was car
band b:
multisp:
scene el
Ip is th
norm
First, fc
which
calculat
panchro
merge t
Figure 7
Figure
multispe
Another
panchro
multispe
The proi
Specialis
Their re
hanced |
multispe