. I LLL 
positions distances linearities 
  
Figure 6. Geometric constraints of control points used in SAPAGO 
3.4 Orientation of Images 
Orientation of images is performed by the bundle adjustment 
with geometric constraints such as length and linearity. 
Transformation equations of all images are simultaneously 
determined by SAPAGO (Simultaneous Adjustment of Photo- 
grammetric And Geometric Observations) developed by the 
authors. SAPAGO has a self-calibration capability and can 
correct image distortion due to principal distance, principal 
point location and lens distortion. 
3.5 Creating a Geometrically Corrected Image 
     
  
When the object can be treated as a plane, a DSM is not e 
necessary to create a geometrically corrected image. The only 
work of an operator at this step is to input a size of a pixel of a 
geometrically corrected image to be created. After input of the 
pixel size of an output image a computer executes geometric 
correction of each image following the determined transforma- (a) mosaicked image of 30 detailed images 
tion equation. Finally geometrically corrected images are 
connected each other to become a mosaicked image covering 
the whole of the object. 
When the object cannot be treated as a plane and stereo images 
of the target have been acquired, a DSM for generating ortho- 
images is extracted at first. Extraction of the DSM is executed 
automatically in principal. Next all of acquired images are 
transformed orthographically by using the extracted DSM. 
Finally each ortho-image is connected with the other, and a W 
mosaicked image covering the whole of the object is generated. ed 
me 
m 
4. APPLICATIONS OF THE METHOD ci 
Q 
Our method has been adopted by the following systems: = 
m 
(1) An image-based information management system for resto- is 
ration of cultural heritages a] 
This system adopts two-stages image acquisition way, nc 
because precise and accurate records of an old cultural Si 
heritage are required. Figure 7 shows two examples of av 
products of the system. The objects are mural paintings of a 
famous temple in Japan. Figure 7 (a) shows a mosaicked Ty 
image of 30 detailed images (6 columns by 5 rows), and Dt is 
: Figure 7(b) shows a mosaicked image of 3 detailed images (b) A mosaicked image of 3 detailed images ra 
(1 column by 3 rows). re 
Figure 7. Examples of the proposed method (1) Cc 
—190—