Kunii, Yoichi 
  
2.1.3. Similarity Function: Line tracking was performed using similarity 
function in this paper. The similarity is an index for representing how similar are I, 
the 2 lines, and the lines which have the highest similarity were discriminated as 
corresponding lines. Figure 4 shows 2 lines (L; and L;) and a middle separating 
line (L,,). The average distances of end points between L,, and 2 lines are D; and 
Dj, and the lengths of overlapping part of the 2 lines to the L, are P; and P,,. 
Consequently, the similarity of 2 lines is calculated by following equation. 
1 
k,D+k,P 
where, 
  
@) 
  
F, 
D=D,+D,, P= 
P 
c 
  
  
  
k, ,k,: coefficient parameters set by a user 
Figure 4. Similarity function 
2.1.4 Line Tracking for Stereo Matching: Sequential images for the house model was taken while a video camera was 
moving in horizontal direction, and the line tracking was performed using the sequential images and threshold value of 
similarity (0.7). On the other hand, affine coefficients are calculated using the corresponded points between the 2 
frames in the sequential images, and let make correspond for the points which were less than 0.7 similarity using the 
affine coefficient. 
From the test, 58 lines were extracted for the first frame, and 51 lines out of those lines were tracked correctly from the 
first frame to the last frame. 7 lines were not tracked and could be erased automatically. Consequently, 102 points by the 
both ends for first flame lines could be corresponded to the last frame. Thus, it can be said that the proposed stereo 
marching method is effective for automatic stereo matching. Figure 5 shows result of stereo matching. 
  
es were 
OW Was 
BA, N., 
Lucas- 
  
   
(a) Lines in the first frame (b) Lines in the last frame 
        
(c) Points in the first frame (d) Points in the last frame 
Figure 5. Result of Stereo Matching 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000. 461