CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey 
close up top images are obtained. The camera was lifted with a 
tripod to provide top images. However, the recessed portion of 
the top floor and the roof could not be well documented since 
lift-support and aerial photogrammetry were necessary for these 
work. Apart from close ups, overviews were taken. Overview 
images are not for measuring, but they cover the façade in a few 
photographs and help one to orient himself within the image set. 
Apart from multi-image technique, single image rectification is 
selected to be used for the photogrammetric plotting of the 
details. For this process, the digital camera was used. Since 
these images were for texture mapping and orthophoto 
production, it was important to shoot them with minimum 
perspective possible. Within this frame, unit areas between 
every two pilasters (6,5 x 14 m) were documented in a single 
photo taken from the ground level. 
4. PRODUCTION OF PHOTOGRAMMETRIC MODEL 
In order to produce the photogrammetric model; first, the metric 
images were digitized on a desktop scanner. Then, these images 
were evaluated with an image evaluation software. 
4.1 Scanning 
Photogrammetric evaluation in desktop environment requires 
images in digital format. Films shooted with reseau metric 
cameras are best suitable for scanning because of the possibility 
of controling image deformation caused by the scaning process. 
A wide range of high quality image scanners with high optical 
resolution, good geometric quality and color depth with 14 bit 
for each color is available. In short, large film formats with a 
reseau in combination with desktop scanners with high 
resolution provide good results (Pomaska, 2001). 
imago 2 
good intersection angle 
Figure 3. Principle of taking metric images 
In this study, films shooted with Rolleiflex 6006 metric reseau 
camera were scanned on AGFA Duo Scan T 2500. With this 
scanner, a resolution of 2500 ppi can be reached. The object 
size in regular images was approximately 15 x 15m. This 
corresponded an image size of 60 x 60mm. Therefore, the image 
scale was calculated as 1/250 (15 000 / 60 = 250). According to 
the scanning theorem (Luhmann, 2000), 5mm should be taken, 
if 10mm is the aimed object resolution. Then, the result of the 
calculation 15 000mm (object size) divided by 5mm (object 
resolution) is 3000. This is the number of pixels. The end- 
product of the scanning process will be 3000 by 3000 pixels in 
size. This is around 1000 dpi. For getting a unique image 
resolution, the object resolution was selected with 0.005m. A 
surface of 15x 15m is then given with a resolution of 3000x3000 
pixel. On the other hand, the production of good colours in the 
scanned images requires the calculation of a profile for the 
scanner. Here, scanning was carried out in RGB colours. For the 
improvement of the images via the image processing program 
Photoshop, gamma correction of 1.5 and sharpening was made. 
Image files were named considering film numbers and image 
numbers, and saved as bitmaps. 
4.2 Evaluation of Images 
There are various software for evaluating metric images. In this 
research, CDW (Close Range Digital Workstation) was used for 
the evaluation of metric images. This software provides aid in 
image refinement. Correction of photo coordinates is known as 
image refinement. Image deformation results from different 
sources. These are lens distortion and geometric deformations 
such as lack of film flatness or orthogonality (Atkinson, 1996). 
CDW handles the correction of lens distortion and geometric 
deformations. At the same time, the software measures image 
coordinates (Pomaska, 2001). 
In this study, library of images for left, middle and right parts 
(see section 3.2.2) was developed. Scanned images were saved 
in their correct folders. Evaluation of each library (part) was 
carried out separately; then, the results were combined. The 
following process was repeated for each set in CDW 
environment: 
1. The system fixing table was filled in. From the 
control points, measured with a total station, seven 
values for defining the degrees of freedom for a co 
ordinate system were selected and entered in the 
system fixing table. 
2. Camera data was introduced: 6050 middle format, 
reseau 6006, focal length: 51.24, XH: -0.25, YH: 
0.16, rO: 20, angle unit: GON. 
3. Images were oriented. During this process, the reseau 
crosses were utilised. The crosses at the four corners 
of the central cross were checked in. This gave way to 
an automatic selection of the rest of the crosses. If 
more than 50% of the crosses are oriented 
automatically, the transformation is considered as 
sufficiently accurate. In turn, the image file is saved. 
4. Image coordinates were measured. Images 
documenting the same sub-section of the building 
surface were grouped together. At least two images 
viewing the same area with a different perspective is 
necessary for measurements (see section 3.2.2). Then, 
‘measure image coordinates’ was chosen on the 
toolbar. Six matching points that are observed on the 
images of the sub-section worked on are marked. For 
example, left top corner of window A is marked in the 
images viewing it. The six points should be 
distributed evenly on the selected elevation portion. In 
addition, these points should not belong to the 
shadowy and hidden areas. 
5. Object coordinates and camera positions are 
calculated. Theoretically, this process is composed of 
a series of resections followed by intersections