inbul 2004 
nely Eqs.3 
(4) 
| points of 
d points of 
ion of DC 
model of 
aight lines 
ject space. 
idjustment 
onvergent 
EXTURE 
by several 
cquire the 
1ce facade 
only the 
tively are 
e question 
  
matched 
ited, there 
ise of the 
obtained. 
as mosaic 
stermined. 
sented. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
5.1 Selecting Mosaic Point Pair by Maximum Correlation 
Coefficient 
As we know, correlation coefficient is employed to determine 
whether the point pair of interest is corresponding? The larger 
the correlation coefficient is, the more likely it is that the point 
pair of interest is corresponding. Consequently it's taken for 
granted that the point pair with maximum correlation 
coefficient should be mosaic point pair. The displacements d,, 
d, are calculated by the mosaic point pair and facade textures 
are mosaic as Fig.7 illustrated. The result is not satisfying 
because even the point pair with maximum correlation 
coefficient is not really corresponding thanks to the similarity 
of texture. 
1 
1 
LE 
i € 
ü 
— 
HR 
i 
1 
"maa mc 
  
; 
  
Fig.7 The mosaic result employed the strategy of 
maximum correlation coefficient 
Hence to acquire a real corresponding point pair, geometric 
constraint needs to be introduced. 
5.2 Selecting Mosaic Point Pair Combined with Correlation 
Coefficient and Geometric Constraint 
Relative orientation process is implemented employed the 
corresponding point pairs matched above and the vertical 
parallax Q of every corresponding point pair is acquired 
meanwhile. Therefore the corresponding factor is calculated as 
below. 
(5) 
with: 
Te Gobi Le 
F;: the corresponding factor of point pair 7, the larger — 1s, 
i 
the more likely it is that point pair i is corresponding 
Q; : the vertical parallax of point pair i 
f; : the correlation coefficient of point pair i, the value 
of p; is 100 when point pair i is fully correlated 
  
Fig.8 The mosaic result employed the strategy 
combined with correlation coefficient and 
geometric constraint 
735 
According to the corresponding factor, the mosaic point pair is 
obtained and facade textures are mosaic as Fig.8 illustrated. 
The result is obviously satisfying. 
5.3 Tone Adjustment 
Because of the illumination, the color difference exists between 
each two images. In this paper, it is presented that an algorithm 
of tone adjustment by maximum intensity difference detected to 
eliminate the color difference. 
Firstly, the intersectional area is determined according to the 
corresponding points matched. The maximum intensity 
difference is then detected in this area and the tone of the right 
image is adjusted by linear mapping according to it. 
6. EXPERIMENTAL RESULTS 
Facade textures are acquired from image sequence extracted 
from video obtained by DV. The size of image is 720 pixels 
480 pixels. 
6.1 The Comparison between Two Kinds of Least Square 
Adjustment Models 
To compare the difference of convergent radius and stability 
between the model employed the constraint of straight lines 
bundle (Model A) and the model controlled by the constraint of 
known orientation of parallel lines in object space (Model B), 
   
     
  
(a) Model B 
Fig.9 Rectified Images (Histogram of Angle) 
  
(a) Model B (b) Model A 
Fig. 10 Rectified Images (Histogram of Angle with 
the Geometric Constrain of Normal Vector to 
Interpretation plane) 
  
    
(a) Model B (b) Model A 
Fig. 11 Rectified Images (Histogram of Angle with the 
Geometric Constrain of Normal Vector to 
Interpretation plane) 
the least square adjustments employed these two model are 
implemented based on the results of grouping lines by three 
methods, i.e. grouping lines by angle histogram, by angle