International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
  
distortion) were not significant. Errors in check points were +5 
mm and +12 mm, expressed as standard and maximum residual 
errors in coordinates, respectively. 
Data reduction was carried out in a DPW (Socet Set” v. 4.4.2). 
A new camera file was created with the selfcalibration results 
and the computed outer orientation parameters were also 
imported (as initial approximations). Next, an automatic point 
measurement process was launched in order to increase the 
triangulation network density. Then new orientation parameters 
were computed with all data. 
A digital surface model (DSM) of the vault was created using 
matching techniques and manual edition. The DSM post 
spacing was 0.025 m with a total of 76500 points. 
This DSM allowed a contour map of the vault with an interval 
of 5 cm and subsequent profiles of main sections of the vault. 
These sections and contour map have illustrated clearly the 
large deformation of the vault, which at present is near to 
collapse. Finally, an orthophotograph (3 mm GSD) was created 
(Figure 8). 
3.3.3 Image rectification. The belfry, all inner walls and some 
outer walls in the north side were photographed with the semi- 
metric (reseau plate) and non metric, analogue and digital, 
cameras. Both stereoscopic pairs and convergent photographs 
were taken and the two networks were used in order to reach a 
favourable geometry for bundle adjustment and control point 
densification. In these areas image rectification was found to be 
suitable, since most walls and arches could be projected onto a 
surface without accuracy loss (the required accuracy should be 
better than +4 cm) and the final project cost had to be moderate. 
Analytical rectification has been made by means of the well 
known two-dimensional projective transformation (Novak, 
1992): 
_ A,X + A3Y + à; 
extreytl (1) 
y. bix+ bay+ b, 
extrevyrl 
where u, v = image coordinates 
X, y = planimetric object coordinates 
2;, Ge 83, Dj Ds Ds c, €; 7 projeciive 
transformation coefficients 
Once transformations were computed, each single image was 
warped using ENVI? program (Environment for Visualizing 
Images, from Research Systems Inc.) The resampling was made 
by bilinear transformation. The computed radial distortion 
coefficients (by selfcalibration) were taken into account to 
avoid curved edges due to the large distortion values of the non 
metric lenses used. 
In most cases, a single image was not enough to cover the 
whole wall or arch surface. So rectified image mosaics were 
necessary. Seam lines between rectified images were smoothed 
by selecting an edge feathering distance of 20-30 pixels. 
Finally, a radiometric adjustment was made to homogenize the 
mosaics. The adjustment was made with usual image 
processing software, Adobe® Photoshop®. 
Figure 9 shows the mosaic formation results. In this case the 
mosaic corresponds to the south side of the belfry and an inner 
wall of the sacristy. A total of 6 photographs and 16 control 
points were used. In some photographs the number of control 
points was 4 or less, so in order to apply the projective 
transformation with redundancy, additional photographs were 
necessary. A bundle adjustment was carried out for the control 
point network densification. Most part of photographs was 
taken with the Canon D30 and 20 mm lens (Figure 9B/F). 
However, the higher section of the tower (Figure 9A) was 
photographed from out of the church with the Pentax camera 
and 200 mm lens. This camera was used since object-camera 
distance outside the church was excessively considering the 
Canon D30 resolution. Besides, it was impossible a shot from 
the sacristy ground because lack of visibility and the photo had 
been highly oblique. 
  
Figure 9. Inner wall in the sacristy and south side of the belfry. 
Mosaic of geometric (rectified) and radiometric corrected 
images (G). 
Also the UMK images have been used for image rectification 
mosaics. So in outer walls (east and north sides) all available 
cameras were used. Figure 10A shows a mosaic in the outer 
cast walls with UMK (walls, taken from the mobile hoist, 
Figure 4) and Hasselblad (tower) photographs. Rectified single 
images from the different walls (with several orientations) were 
mosaicking. Planes were projected onto a surface parallel to the 
belfry west side. Figure 10B shows the north side walls (Canon 
D30) and belfry (Pentax camera). 
In other areas of interest, some complications appeared. So, in 
the arches crossing the nave (Figure 1C), space limitations and 
height of the arches made necessary large number of photos 
(Figures 11, 12 and 13). Besides, in some parts of the arches the 
photos had to be highly oblique. These condition added 
problems to the mosaic formation. On another hand, these 
inconvenient made more complicate the set up of the 
control/check point network. 
  
Interna 
etr 
Fig 
Figure 
employ 
of the 
compli 
excessi 
points 
densifi 
is illus 
  
   
Fi gure 
Figure 
stereoy 
the st 
limitat