N FOR
ATION
ages used in
nent, RMSE
arameters.
The fact that
istortions on
s can distort
oring out the
rmation. The
y about 20%.
)-F software.
thods also in
al images.
lation process
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
Noise reduction and type of filters used are very important in
pre-processing step. Original image could be adequately
enhanced by filtering images, but in such way to not disturb
information of original aerial image. Method which has been
tested is shown on fig. 2.
original images
Y
filtering |g choise of
a filter
filtered image
choise of the
mathematical
operation
images
association
> result image
Fig. 2. Scheme of method of image enhancement in spatial
domain
Usage of addition operation for two images allows to eliminate
weaknesses on original image and to bring out details
simultaneously. Addition of images is made by summing of
optical density values of adequate pixels of each image, and
than conducting normalization.
The other method of filtration used in object space was adaptive
filtering. The result of this transformation may be used directly
to further digital photogrammetric process. :
A Ci i aaa Ad dl)
Dra un Wee Ab Saw fep
588720512 +4+AL
Bom ET
Fig. 3. A part of aerial image, on left - original image, on
right - filtered with adaptive filter
2.2. Filtering in frequency domain
Filtering in frequency domain allows eliminate noise and
regular errors on digital data.
Aerial images filtering in frequency space do not suit to further
using in automatic aerial triangulation process. This possibility
exists indirectly as only adequate association filtered image with
original one gives an effect that is going to be used in further
photogrammetric process. Elaborated and tested method is
shown on fig. 4.
original image
filtering |< choise of
a filter
filtered image
choise of the
mathematical
operation
result image
Fig.4. Scheme of the method of image enhancement in
frequency space
The example of such association may be a picture arisen in the
result of multiplication of an original image with an image
filtered with bandpass filter in Fourier space or enhancement
(by addition in spatial domain) of an original image with an
image filtered by highpass filter in Fourier space. On the figure
6 presenting detail of terrain is shown. We can see sharp
enhancement of edges without loosing of any information from
the original image.
549
