format of such photographs is associated with different kinds of 
degradations, so the use of such photographs in 3-D object 
reconstruction requires a kind of software that can be applied in 
processing wide range of photograph formats. The DLT 
procedure satisfies these requirements because the 
transformation coefficients are independent from the camera 
constants. 
The inner and exterior orientation of different photos means that 
we specified the DLT parameters not taking into consideration 
the different degradations occurring in X-ray photos but 
regarding them as modifications of transformation parameters. 
When determining the points, these differences are eliminated, 
on the other hand we think that this solution reasonably meets 
the accuracy requirements. For point determination we used the 
DLT program developed by the BUTE based on the Wong 
solution (Wong, 1975). With the help of a few adjustment 
parameters we can take some elements of X-ray degradation as 
distortion values into consideration as well. Testing the DLT 
program was done by the use of computerized simulation for 
one part, and by the processing of analogue and digital data of a 
test area for the other part. Figure 1, shows the test area of a 
monument structure. 
Second, the starting point of the new object coordinate system 
is the drafting point of the normalized stereo pair, the X-axis is 
the base itself and the Z-axis is perpendicular on the normal 
stereogram. 
Beside the above-mentioned conditions, the well known co- 
linear equations produce a relationship between the original and 
normalized stereo pairs, so through what is called basic 
exchange matrix a relationship can be formed between the two 
pictures and the new object coordinate system. This makes it 
possible to produce normalized stereo pairs that we solved with 
pixel transformation. 
During the transformation, it is practical to choose the camera 
constants of normalized stereo pairs higher than those of the 
original one, so that we don't lose any pixels in the process. The 
transformation itself was solved, similar to digital ortophotos, 
so that the location of each pixel of previously defined picture is 
searched in the original digital picture, the grey levels in the 
transformed picture are determined from the original picture 
depending on the nearest neighbourhood theorem. Figures 2 and 
3 show left and right stereo pair photographs in their original 
and normalized photographs that were used as a test sets. 
  
Figure 1. Test site 
3. DRAWING THE RESULTS 
The most important part of our task was depicting the results as 
they can be best used during surgical operations. At present 
stage of our research we have chosen two methods knowing that 
further development is, of course, needed to be carried out on 
them in case of using the methods during surgical intervention. 
Among the depicting methods of digital photogrammetry, the 
oldest and most simple one is the so-called split monitor 
solution. In this case, while looking at the normalized stereo 
pair with stereoscope, the 3D effect is formed. We used these 
relatively old and strict geometric conditions because this 
process bears the promise that the operating physician can see 
the stereoscopic structure of the vascular system during the 
surgical operation via pictures projected into an ocular device, 
which is one of our later goals of this research. The production 
of normal stereogram in the case of X-ray camera with the use 
of stereo radiograph is rarely possible, that is why in our 
solution we decided to apply the normalization of the general- 
arrangement stereo pair. If we know the orientation elements of 
photos to be evaluated we can solve this task. While doing so 
we take into consideration that the geometry of the normalized 
stereo pair correlates with the normal case of the terrestrial 
geometry when there is no cross parallax among the 
homologues points. 
When preparing the original and normalized digital photos, we 
have to fulfil two geometric conditions. First, the common axis 
of normalized stereo pairs must be parallel with the base. 
a) original left photograph 
  
b) normalized left photograph 
Figure 2. Left test photograph 
When checking the normal stereogram, the most important 
factor is the geometric accuracy. For this purpose we made 
object reconstruction from the normalized stereo pair photo 
coordinates, and then we examined the residual errors of known 
points of the normalized stereo pairs. The other geometrical 
control method is the examination of the cross parallax of 
normalized stereo pair since their theoretical value is zero. 
From among the non-geometric examinations, conditions 
known in the 3D vision were taken as a guiding principle. 
-24$8-