nbul 2004 
ile black 
> filter) to 
y images 
> value of 
> majority 
fer 2000). 
tion, the 
following 
s (in the 
ies in the 
ith multi- 
ous edges 
d smooth 
shrinking 
  
  
  
dges (a), 
cas with 
bution of 
highlight 
eliminate 
ed to be 
c, multi- 
city of 
suggested 
rows by 
in 1956, 
solution) 
ho-image 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV. Part B2. Istanbul 2004 
(5.0m resolution) created from an aerial image captured in 
1999, and 300 rows by 300 columns LANDSAT image (30m 
resolution) captured in 2001. These scenes exhibit significantly 
different geometric and radiometric properties. Straight-line 
segments have been manually digitized in these images. As an 
example, Figure 3 shows the digitized segments in the 1999 
ortho-photo and 1956 aerial image, where 139 lines have been 
digitized in the reference image (1956 aerial) and 183 Lines 
have been digitized in the input image (1999 ortho-photo). 
  
  
Aerial 1956 
  
  
Ortho-photo 1999 
  
Figure 3. Digitized linear features in the 1956 aerial image and 
the 1999 ortho-photo 
2 
A closer look at Figure 3 reveals that there is no complete 
correspondence between the digitized lines in the input and 
reference images. The digitized segments were then 
incorporated in the MIHT strategy to determine the parameters 
involved in the registration transformation function as well as 
the correspondence between conjugate line segments. 
The estimated parameters for affine transformation functions 
and their variance components for the abovementioned datasets 
are listed in Table 1. The estimated variance components, 
which reflect the quality of fit, reveal two facts. First, they 
show good registrations between the involved images (within a 
few pixels). Also, the small variance components signify the 
validity of the affine transformation as the registration 
transformation function. 
  
  
  
  
  
  
  
  
  
  
  
  
  
Affıne Aerial_5 6/ Aerial 56/ Aeria | 56/ 
Aerial 72 Ortho 99 Land O1 
6, (Pixel ?) 2.9524? 24537? 1.8822? 
a, (pixel) -78.9282 -105.8868 41.3663 
a, 1.4148 1.0899 00.1753 
a, -0.0925 -0.0235 00.0493 
b, (pixel) 415.1431 614.5326 83.0329 
bi 00.099324 0.0292 -00.0481 
b, 1.4242 1.0916 00.1754 
  
Table 1. Affine transformation parameters between the involved 
datasets 
In addition to the estimated parameters, the correspondences 
between line segments have been identified. For example, 
Figure 4 depicts established correspondences between the 
digitized primitives in the 1956 aerial image and the 1999 
ortho-photo. A mosaic image covering the northwest part of the 
city is derived by combining the 1999 ortho-photo and the 1956 
aerial image, where every other square patch in the reference 
image has been replaced by the corresponding resampled patch 
in the input image, is shown in Figure 5. It can be seen that 
features (for example, roads, rivers) in the derived mosaic fit 
each other (observe the smooth transition along the features 
within the resampled patches). A closer look at Figure 5 reveals 
the changes that took place in the northwest part of the city 
during the forty three years between the moments of capture. 
449 
  
  
  
memes C:iho-photo 1999 Lineer Features 
—— Matched Aenal 1956 Linear Features 
==. Non-Matched Aeral 1956 Linear Featutes 
  
  
  
  
  
  
Figure 4. Established correspondences between the 1956 aerial 
image and the 1999 ortho-photo line segments 
  
Having established the transformation function between the 
images, the input image can be resampled into the reference 
frame associated with the reference image. As explained in the 
previous section, the resampling is followed by applying Canny 
edge detection and majority filter to both images. Then, the 
resulting images are subtracted to produce a change image, 
which is enhanced by re-application of the majority filter. 
Figure 6-a shows the change image resulting from the 
registration of the 1956 aerial image with the 1999 ortho-photo. 
In this image, white areas indicate changes while black areas 
indicate parts with no change. Simple statistics show that there 
is roughly 50.6% change between the 1956 and 1999 imagery. 
Dividing the area into four quarters shows that the percentages 
of change, which occurred in the northwest, northeast, southeast 
and southwest parts of the image, are 74.8%, 66.4%, 34.4%, 
and 26.896, respectively. The sub-images (b, c, d, and e) in 
Figure 6 show different types of changes that took place. Sub 
image 6-b shows changes as a result of an urbanization activity 
(new residential community has been built). Sub image 6-c 
shows changes caused by trails in newly developed parks. 
Changes resulting from the construction of a new highway 
along the east side of the city is shown in sub image 6-d.