International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004 
  
  
    
   
   
   
   
    
       
  
    
    
  
   
  
  
  
  
  
  
  
  
   
number of valid point pairs 
45 : 
40]. ————Ó EC E ; re anh fe : 
Terra qur qut Rr ERR / 
30 Mn ne a EE rer free fev vnm 
25 A V. une ases 
  
  
  
  
  
20 vem N ER RES CRT gérer aenenedre pet sensen ua demnesatne naan yeas Seva snr 
15 | 
10 earn ness Tamas Saeed : 
S N 
e 0 1 
correlation coefficient 
number of valid point pairs 
  
  
  
  
^ 
100 
1 | 
80} le 
f 
Agl.—— EE 
20] dtl eres teers nca ten N at 
m n 
E 1 
correlation coefficient 
Figure 6. Curves of correlation coefficients on all tracked point 
pairs by the LK method (top) and our IOFE method (bottom) 
  
   
  
   
  
    
  
  
   
   
   
  
   
  
  
LK [OFE NCC 
number of valid points 115 (17%) 312 (48%) | 366 (56%) 
registration accuracy 0.796 0.415 0.351 
(pixels) 
RMS 
X: 0.685 0.306 0.310 
Y: 0.384 0.277 . 0.162 
Computation time 0.55 0.59 5.47 
(second) 
  
  
  
  
  
    
   
  
   
     
     
   
  
   
  
   
    
Table 1. Statistic figures of tracking results by three 
methods (number of extracted feature points = 650) 
  
Figure 7. (left) the extracted feature points (yellow dots) overlaid 
with DV image; (right) the tracked points by the LK (red dots) and 
by the IOFE (yellow), where each point with the same tracking 
results is denoted by a single yellow point 
    
i 
: ‘A 
si 
uu 
Figure 8. A test image of 3D objects, where three sets of 
registration parameters are used for the areas A, B, and C, 
respectively 
  
  
  
  
  
  
  
  
IOFE NCC 
A | Number of valid points 2196 3996 
a, (pixels) 0.140 0.000 
B | Number of valid points 12% 12% | 
a, (pixels) 0.149 0.154 
C | Number of valid points 22% 24% 
  
  
  
  
  
Table 2. Statistic figures for DV images shown in Figure 8 
4. CONCLUSIONS 
In this paper, we present a new approach for automatic 
extraction, selection and transfer of corresponding image points 
in a series of sequential digital video (DV) images. It is called 
“iterated optical flow estimation (IOFE)". This approach 
utilizes and extends some well-known techniques and theories 
(e.g. the optical flow theory) as well as a proposed simple error- 
detection mechanism to achieve a much better efficiency than 
the well-known Lucas-Kanade optical flow estimation (LK) 
method. Compared with the traditional LK method, the IOFE 
approach significantly increases the maximum tracking distance 
and also improves the reliability of the tracking results. Our test 
results use the SONY DCR-PCII5 DV image sequences and 
show that the trackable range of 3-4 pixels in the LK method 
can apparently be enlarged to 30 pixels in this new approach. 
Moreover, the proposed error-detection mechanism simply 
utilizes the average gradient, normalized cross-correlation, and 
a simple image registration aided by least squares adjustment. 
Test results show that it can efficiently detect and delete wrong 
tracked points, and thus apparently improve the quality of 
automatic point transfer, and automatically provide accurate 
coordinates of a large number of corresponding image points in 
DV image sequences. 
This work aims at high precision automatic image triangulation 
for the automatic real-time mobile mapping vehicle system 
(MMVS). Some future works need to be done, e.g. high 
precision point measurement with a sub-pixel accuracy level, 
and rules for adding new tracking points. Thus, the IOFE might 
be improved so that it can be utilized in high precision photo 
triangulation for the real-time map-updating and other 
surveying purposes of a MMVS. 
    
    
D' 
Mi 
RS 
Fi: 
ns 
Pr 
atc 
ba 
Te 
Tr 
D 
PP 
De 
Ta 
cal