>y are 
dings 
eal of 
tures, 
ces of 
| task. 
rectly 
e best 
d and 
1, line 
ching. 
sts of 
been 
1996; 
). The 
; right 
'asure 
ion in 
ability 
  
  
Chunsun Zhang 
3.1 Line Extraction 
The input images are first filtered with Wallis filter for contrast enhancement and radiometric equalization (Baltsavias, 
1991). The technique developed in a previous project (AMOBE) is used to extract straight lines (Henricsson, 1996). 
The edge pixels are detected with the Canny operator. An edgel aggregation method is applied to generate contours with 
small gaps bridged based on the criteria of proximity and collinearity. All segments are checked using their direction 
along their length and split at points where the change in direction exceeds a given value. A test is conducted to see 
whether the consecutive segments can be merged into a single straight line. For each straight line segment we compute 
the position, length, orientation, and photometric and chromatic robust statistics in the left and right flanking regions. 
The photometric and chromatic properties are estimated from the “L”, "a" and "b" channels after an RGB to Lab color 
space conversion and include the median and the scatter matrix. 
3.2 Computation of the Similarity Score 
With known orientation parameters, the epipolar constraint can be employed to reduce the search space. The two end 
points of a line segment in one image generates two epipolar lines in the other image. With the approximated height 
information derived from DHM25 or DSM data, an epipolar band is defined (Baltsavias, 1991). Fig. 4 illustrates this 
idea. Therefore, a search region is determined in the right image for each segment in the left image. Any line included 
in this band (even partially) is a possible candidate, if it intersects the two epipolar lines (through the two line endpoints 
in the left image) within this band. For example in Fig. 4, lines i, j, k are accepted and will be compared with line pq in 
the left image for similarity measurement, while line r is rejected because it intersects eq outside q7, q2. The size/height 
of this search band is decreasing with edge length and orientation difference to the epipolar lines. The comparison with 
each candidate edge is then made only in the common overlap length, i.e. ignoring length differences and shifts between 
edge segments. 
  
PI 
Figure 4. The epipolar band pip2g2q: defines the search space for line pq 
For each pair of lines which satisfy the epipolar constraints above, their rich attributes are used to compute a similarity 
score. Therefore, the similarity score is a weighted combination of various criteria. A similarity measurement for length 
is defined as the ratio of the minimum length of the two lines, divided by the maximum one. Thus, the similarity score 
is defined as 
min( Lien , R 
= len len ) (1) 
len 
max( Lien » Rien ) 
We use the absolute difference between two line angles and an expected maximum difference in a ratio form similar to 
(1). The maximum value Tang is à predefined threshold value. Since in aerial image the Kappa angle has a large effect 
on the line angles, we rotate the lines with Kappa in left and right image respectively before computing the line 
similarity score by 
  
V zw T ang et | L ang di R ang / (2) 
ang. = T no = 
wis a weight related to line length, and given by w - wjw; , where W;, W, are computed for the lines in left and right 
image respectively. They are defined as a piecewise linear function as: 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 1011