ul 2004 
lity are 
' source 
ntersec- 
 resolu- 
size (in 
ded as a 
ence, as 
partici- 
tions of 
eme has 
  
ntation 
ed to A. 
vas final- 
and also 
ita, using 
odel of a 
trance of 
Jyzantine 
columns, 
was used. 
1 60? ver- 
ut from a 
nm of the 
a distance 
leres were 
ks Survey 
1 Of scans. 
In total, 7 
hese were 
vhich con- 
f the soft- 
prove the 
ed (in Fig. 
alogue and 
ere, results 
gaPixel ca- 
taken with 
All images 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
The final, automatically generated, orthoprojection is presented 
in Fig. 6, left. The result is essentially satisfactory. A main im- 
perfection are small holes due to lack of texture (the camera has 
not been elevated). Most other defects — for example, a certain 
aliasing at edges of some parts which, however, may not be per- 
fectly perpendicular to the projection plane — are observed only 
with considerable zooming. On the right, a section 1s presented 
utilising all available model and texture information, including 
parts which are not visible on the left (e. g. a part of the arches). 
  
  
  
  
Figure 2. Intensity image of part of the central scan. 
  
A set of 18 signalised points served as geodetic ground control, 
most of which appeared on all images. Using our own software, 
a bundle adjustment with self-calibration (including coefficients 
k, and k; of radial symmetric lens distortion) was carried out in- 
volving 8 tie points. It was possible to achieve a high accuracy, 
as seen in Table 1. 
  
Table 1. Bundle Adjustment 
(7 images; 18 GCPs; 8 tie points) 
  
  
  
0, = 0.28 pixel 
  
c (pixel) 2573.94 + 1.36 
X, (pixel) 10.27 3: 1.05 
yo (pixel) 7.35+0.37 
ky (x10%) —2.98 + 0.04 
ko (x105) 4.90 + 0.19 
  
  
rms (GCP) = 0.3 mm 
o (TIE) - 1.3 mm 
  
  
  
  
Figure 3. The seven images used in the project. 
Using the data for camera calibration and exterior orientations, 
the images could be used to colour the projections of the model. Regarding speed, a PC (CPU Athlon XP 2.4 GHz, 512 MB RAM) 
required 5 min to project 3 million triangles onto the images, in 
a suitable order to facilitate the last step; 1 min to establish the 
Z-values for each orthoimage pixel; and 13 minutes for the pro- 
All images used here had been acquired successively, i .e. under duction of a final image 3720x2775 from 7 images 2592x1944. 
the same lighting conditions. Thus, as no radiometric variations 
among images were observed, texture averaging was considered 
as adequate (for pre-processing see e. g. Griin et al., 2001). The 
plane of the church fagade was selected as projection plane. The 
orthoimage pixel size was fixed to 2 mm. As already seen, the An algorithm has been implemented for the automatic synthesis 
3.3 Model texturing 
5. CONCLUDING REMARKS 
algorithm identifies image visibility of all model parts previous- of textured views, given a 3D triangulated mesh and precise ca- 
ly established as being visible in the direction of projection. The libration/orientation information for overlapping images. Model 
and image visibility are identified, to allow pixel colouring with 
the weighted average from all viewing images, whereby a basic 
results for an individual image are shown in Fig. 4. 
A next question to be noted, also noticed in Fig. 4, concerns the blunder detection tool allows to avoid artifacts. The product for 
matter of outlying colour values, particularly close to an occlu- the paticular object studied here is indeed very satisfactory. 
sion border (cf. section 2.3 and Fig. 1), a problem illustrated in 
Fig. 5. Above, an extract is seen of the orthoimage derived from Yet, further enhancements of the technique are both necessary 
all source images without blunder-filtering. The white artifact — and feasible. Thus, some hole-filling tools should be introduced. 
clearly originating from the occlusion border of the image in the Besides, more robust means for outlier avoidance need to be ex- 
middle — disappears if colour values deviating more than +o perimented with. For instance, closeness of a source pixel to an 
from the mean are automatically ignored (below). occlusion border can be taken into account, probably combined