Dowman, Ian 
  
  
Fig. 8: Views of the completed model. 
As can be seen, the texture mapped building is almost identical to that shown in Figure 7 from a visual point of view. It can also be 
used for making crude 3D measurements and volume-oriented computations, provided that its NFR is structured properly. 
The following section examines the ATPT based on the geometric quality of the textures. 
5.1 Geometric quality 
To evaluate the geometric quality of the textures, two parameters are defined: white pixels and discontinuities. As already mentioned, 
the textures are created by passing the 3D grid points to the ATPT and extracting the corresponding pixel values. Consequently, the 
system works properly only if the image coordinates of the grid points are computed correctly. For this, the image selection and the 
coordinate computation parts of ATPT need to function correctly and accurately. Otherwise, the grid points are projected to wrong 
places. If the projection of a grid point falls outside the frame of a CCD image, its pixel value is set to 255 which corresponds to white 
colour. Alternatively, if the projection of a grid point is computed to be within the frame of a CCD image but not on the right place, a 
wrong value for the pixel of the point is defined. The result of such malfunctioning, is the appearance of discontinuities on the textures. 
Figure 9 (a) shows the individual images used to process the texture of one of the building facades. The dimensions of this facade of the 
building are 17.5 x 7.5m , and the dimensions of the corresponding texture are 350 x 150 pixels. The texture of this face was processed 
with the spacing interval of 5cm. The texture processing was carried out with no supervision or manual interactions required. The only 
interaction was to define the location of a few files and the value of some parameters, like the spacing interval, to initialise the program. 
  
(a) (b) 
Fig. 9: Individual images and the corresponding texture. 
Figure 9 (b) shows the rectified mosaic of the CCD image portions forming the texture of the face. As can be seen, neither white 
pixels nor discontinuities are present on the texture. This indicates that from a geometric point of view ATPT has functioned properly. 
However, there are some contrast discontinuities over the right area of the texture, which are the result of the Automatic Gain Control 
(AGC). AGC is a property of most CCD cameras used to average the image brightness to a consistent level. As a result, darker 
areas become brighter while lighter ones become darker. It has the desirable effect of compensating for the ambient lighting, but the 
undesirable effect of changing the signal level of an area of the image, if the lighting conditions change elsewhere in the image. A 
method to reduce this effect has been developed by Varshosaz (1999). 
Figure 10 shows another example. The dimensions of the corresponding facade are 22.7 x 13.6m. The texture of this face has 227 x 136 
pixels and was created with a spacing interval of 10cm. Similar to the previous example, no white pixels or discontinuities are present 
on the texture. The textures of the rest of the building faces were processed using images from single stations and in all of them no 
geometric discontinuity or white pixels were observed. These textures were used to form the final VRM presented in Figure 8. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 185