Ahmed Elaksher
2.1 Edge Feature
Edge segments are commonly used as a primitive feature in stereo matching. This feature appears in many images. In a
rural area image, it will not be possible to find strong geometric features as straight lines, circles, corners, etc. These
geometric features are usually parts of man-made structures such as buildings and roads. However, it is still possible to
extract edges from rivers, trails, ridges, shadows, and geomorphic features in the images. There are several well
developed edge detection methods. Strong edges with low noise would be ideal. Any detectors based on the first
derivative can fulfill the requirements.
2.2 Line Feature
A straight line is one of the most prevalent features used for analyzing urban area images. Because of occlusions and the
weakness of area based matching when abruptly changing elevations are present, features become an important factor
for handling this kind of image. Some researchers have commingled straight lines and edge segments. Actually it would
‘be more appropriate to separate straight lines from non-straight edges. Much prior work has been done on line
extraction. The existing line extraction tool, Burn's line extraction from the Khoros software, has been used in this
research.
5 5 $ PUR = ES : S & & ;
Left fier
Figure 1. Intersections of Epipolar Line and Extracted Lines
The final output of the extraction is given in term of the starting and ending coordinates of lines. Their endpoints can be
used to compute some useful attributes for matching such as line orientation, and length of a line. Figure 1 illustrates the
intersection between an epipolar line (labeled) and extracted lines overlaid on the original images.
2.3 Plateau Feature
In manual DEM collection, one of the features that human operators focus on are the boundary points of homogeneous
regions. A plateau feature is defined here as the end points of a 1D region with near constant grayscale. It will be
detected from an intensity profile of the points along an epipolar line. This means that it is a one-dimensional feature.
The plateau is particularly suited for urban imagery. Usually it represents an entity such as a rooftop, roof panel, field,
etc. If a profile is plotted, this feature is usually recognized easily by the human eye (Figure 2) i.e., it looks like a
plateau.
AINSI nr
200
rv
a ;
28 so 80 uw 106 125
Figure 2. Ideal Plateau Features on a Synthetic Intensity Profile
Attributes of the Plateau as well as its position are retained to be compared during the matching process.
268 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
