International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
  
histogram with geometric constraint of normal vector to 
interpretation plane and by initial value acquired artificially. 
The results of rectification are illustrated as Fig.9~11. 
The distributing of vanishing point coordinates, angular 
parameters and focus is shown in Fig.12 based on the results of 
least square adjustment above (The hollow denotes model A 
while the solid 
  
      
7. CONCLUSIONS 
The algorithm of rendering of building facade texture from 
land-based image sequence has been presented. Some 
experiments has been conducted with land-based image 
sequence acquired by DV. The result is satisfying and several 
  
  
  
   
  
   
  
  
  
   
  
ai model A" œ 
4 The distributing of The distributing of The distributing of 
E Eh vanishing point X vanishing point Y angular parameters 
fom! 22 uie (Model B) (Model B) (Model A) 
distributing 
above, a 9 
conclusion is DT 0 eT > 
easily drawn a © P udi distributing = K 
that: model B 9 vanishing point Y TX e distributing of|Z ° 
Das larger (Model A) o angular parameters 
convergent (Model B) 
radius and 
higher stability e 
than those of 
   
   
Istributing of 
vanishing point X 
model A. (Model A) 
Namely, the 
values 
acquired by (a) The distributing of vanishing point 
model B are 
generally 
optimal while 
those acquired 
by model A are locally optimal. 
6.2 The Elimination of Lens Distortion 
There are 408 lines extracted in the image as Fig.13a illustrated. 
And then 68 lines which lengths are over 60 pixels are 
dimidiated and extracted again. The result is shown in Fig.13b. 
The lens distortion is estimated used those lines extracted after 
dimidiation and the result of correction of lens distortion is 
illustrated in Fig.13c. As we can see, The lines curving thanks 
to lens distortion obviously became straight after lens distortion 
correction. That is the lens distortion was eliminated. 
  
c. Result of lens 
distortion correction 
b. Lines extracted 
after dimidiatiaon 
a.Lines extracted 
Fig13. Lens distortion correction 
6.3 The Mosaic of Facade Texture 
Fig.14 is the comparison of mosaic results of facade texture 
employed two strategy presented above. The mosaic result used 
the strategy of selecting mosaic point pair combined with 
correlation coefficient and geometric constraint is obviously 
much better than that employed the strategy of selecting mosaic 
point pair by maximum correlation coefficient. 
After the processes of tone adjustment and texture analysis, 
more refined facade texture is acquired as Fig.15 illustrated. 
(b) The distributing of angular 
parameters 
4 ——00—099— 
9 (c) The distributing of focus £f 
Fig. 12 Convergence Distributing 
  
b. combined with correlation coefficient and 
geometric constraint 
Fig.14 the comparison of mosaic results of facade 
texture 
conclusions are drawn as below. 
» The curvature of straight lines caused by lens distortion 
can be detected used those lines extracted after the 
dimidiation of every long straight line. Therefore the lens 
distortion can be estimated according to the curvature 
detected. The method to estimate lens distortion presented 
in this article is proven to be straightforward and effective 
by the experiments carried out. 
« it is taken for granted that color information can be 
employed to remove the occlusion of trees because of the 
  
  
  
    
  
LE an NENNEN 
Inn | RR E 
Fig. 15 Refined Texture for Building Facade 
   
  
  
   
  
    
   
    
   
   
   
    
       
   
    
    
  
    
   
  
   
  
  
  
  
    
    
   
    
  
   
    
   
    
  
  
     
  
   
    
  
   
    
Intern 
REFE 
B Ca 
Camel 
Cipoll 
1999. 
image 
Comp 
Debev 
Archit 
Image 
ACK! 
We wi 
of Chi 
Video 
suppor