Istanbul 2004
—
iverages and
© X-axis
iges taken on
xels in these
, B and L-
ficiency was
values were
was almost
s 0.35% for
for R, 61%
were 0.7%,
ation values
reas, which
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
0%-saturation of Lumiance-channel
e *
0.2 4 es : + + i e
Soro 9o. 4 —4::9: 0:9. Coe
deles beso dell elle eoe etel
Essi essc
03417
Figure 6. 0%-saturation of luminance-channel of CIR-positive
images.
were black. Example of image-detailed-data of 0%-saturation
of luminance-channel can be seen on figure 6.
In general, the statistics of images of single flight were quite
uniform, even though scanning parameters were determined
using only a few images of the flight (Figure 3). From the
image-detailed efficiency plots of all the flights a few flights
with exceptionally low efficiency could be found. This
indicates that tuning of the scanning parameters using statistics
calculated from a few sample images will probably lead to
acceptable scanning statistics of the whole flight.
4.3 Radiometry of enhanced orthophotos
To determine optimally enhanced CIR-image suitable for
visual forestry and agricultural applications, preliminary
investigations of enhanced orthophotos were made using 20
images received from three image producers and image users
(two Forestry Centres). R, G, B, L, I, H and S-histograms and
their statistics were calculated from the images. Different
kinds of areas (fields and various forest objects) of size
100x100 pixels were cropped from the original images and the
same histograms and statistics were calculated. According to
these statistics, all images were considered to have good
radiometric quality, even though in visual analysis every image
appeared different, so also their interpretability varied.
After these studies and response received from co-operative
partners, a decision was made that the optimally enhanhed
CIR-image would not be defined by numerical factors but
rather a radiometrically and visually high quality image would
be created from the scratch. A scanned CIR-image from
southern part of Finland from summer 2003 was chosen as a
basis for this work.
The criteria’s for the radiometric enhancements were:
e To get coniferous and deciduous trees separated from
^ach other and at the same time set the balance of red
and green colours so that the composition of stands of
mixed forest would be interpretable. :
e To seek optimal tone for open rock- and other
inorganic areas.
e To avoid dark forest shadows to get saturated.
Figure 7. Original aerial image and tone-model-image.
Efficiencies
oser
200 —
150
100 4—
| Orig. R Model R Orig. G Model G Orig. B Model B
Figure 8. Efficiencies of original image and tone-model-image.
For justifying the preliminary adjustments a field control was
made with two forest interpretation specialist from the Forestry
Development Centre Tapio. After this final enhancements were
made. The result, tone-model-image, is presented in Figure 7.
For the fine-tuning also larger scale model images of various
object types were extracted.
The usability of the model-image was tested with several CIR-
images taken on various stages of growing season around
Finland. Excluding Lapland, all images were possible to be
adjusted to match visually with the model-image.
Even though the enhancement procedures were made in such a
way that the efficiencies of all histograms were followed, some
loss of information especially in the near-infrared-channel (red)
could not be avoided. The efficiencies of original and final
image are presented in figure 8.
5. DISCUSSION
5.1 Histogram QC
The purpose of the 10096-QC of scanned images is to ensure
the suitability of the images for further use. The most
descriptive statistics for QC are efficiency, 99%-efficiency and
saturation. Efficiencies and saturation monitor the histogram
from opposite directions; if efficiency is at its maximum,
saturation may exceed its tolerance values. The difference
between efficiency and 99%-efficiency is that efficiency is
more easily controlled during the scanning process, but the
99%-efficiency describes the utilization of the histogram
better. Efficiency may contain some insignificant amounts of
DNs that are excluded in 99%-efficiency.
