International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
of the triangles that cited on the 3D model. Later, the colors of 
the inner pixels of the triangles should be mapped on to surface, 
(Watt and Watt, 1992). Since the scale differences between 3D 
model and pictures, it is obvious to see that there is no one to 
one correspondence between the pixels of the picture and the 
model. For this reason, to find the colors of triangle pixels, a 
color resampling is required. For the color resampling we used 
linear interpolation. 
While visualizing the texture mapped model, it is possible to 
integrate texture colors with illumination and shading 
equations. This integration process is done by OpenGL. Texture 
mapping is performed in two steps: (Watt and Watt, 1992). 
- Referencing the vertex coordinates of the polygonal model 
with (u,v) texture coordinates, 
- Filling the inner region of the polygons (triangles) by 
interpolating (u,v) coordinates (indirectly corresponding 
colors), on the pictures. 
In this study, we obtained (u,v) texture coordinates by using 
backward mapping of the vertex coordinates on to texture maps 
(images) by using 3D Affine transformation and 
photogrammetric bundle resection techniques separately. With 
both techniques, 3D model coordinates of the vertex points are 
projected on to 2D images. During the projection 
(transformation), z coordinates are being lost. For 3D Affine 
transformation, common points are selected from model and 
image by hand with our program. Then our program computes 
transformation unknowns with least squares adjustment. Z 
coordinates of the image points are all assumed zero. 
For the bundle resection, we use the control points. We know 
the model coordinates of the control points and also we measure 
the image coordinates of these points from image pairs by using 
ALSM matching. Then we compute exterior and interior 
orientation parameters with additional lens distortion 
parameters by using bundle adjustment. After the orientation 
parameters have been computed, model coordinates are mapped 
to image space by using spatial bundle rays resection equalities. 
Thus (u,v) texture coordinates are obtained. Then the mapping 
procedure begins. Before texture mapping we did not rectified 
the images according to the model. So there are many 
deformations on texture mapped model because of the height 
and perspective differences. We are still studying on rectifying 
images. 
6. SAMPLE IMAGES PRODUCED WITH MIPAS 
In Figure 1, digital images of the patient are seen with control 
points. These images were taken as soon as the scanning 
operation has finished. These images were used for texture 
mapping and photogrammetric evaluations. 
   
   
Figure 1. Control points and patient's face images 
In Figure 2, some volume models of the skin of the patient head 
which was created by our program have been shown. On the 
upper left in the figure, opaque skin volume model, on the 
upper right MIP volume with skin and bone structures and on 
the bottom, bone and skin volume models are seen. 
  
  
Figure 2. Volume model examples generated from CT slices 
With the volume models, by changing opacity, color and 
gradient values, it is possible to see the 3D medical data under 
desired illumination and shading conditions. When these values 
are changed, corresponding volume segmentation effect is seen 
synchronously on the screen. Thus it is possible to see the 
intended parts of the tissues. On the volume models, it is 
possible to learn the coordinates of picked points. 
In Figure 3, segmentation of tumor on MR image slices is seen. 
By using segmented images, individual tissues’ surface models 
are obtained. 
  
  
  
  
Figure 3. Segmentation of tumor on MR slices 
In Figure 4, on the left, tumor surface model and information 
about its coordinates, area, volume etc. are shown. On the right, 
tumor with brain cortex and on the right with skin and brain 
model is shown. 
  
Interna 
  
  
sac 
In the 
and th 
brain a 
from ( 
tumout 
mappit 
In Fig 
mappit 
  
Fi 
By us 
format 
it will 
many 
be dev 
plastic 
When 
would 
for m 
metho 
be ot 
evaluz 
Altan, 
and b