The International Archives oj the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
661 
Reliable points (green), edges with end points (red) Figure 4. 3D view on estimated 3D Edges with reliable 
frame 0 (upper), frame 133 (lower) points, side view (upper), top view (lower) 
5. CONCLUSIONS 
Video as an easy obtainable and low cost image data source has 
been of interest for many researchers in recent years. However 
recovering accurate structure of objects in realistic environment 
from video image sequence is still a difficult problem. 
Especially reconstruction from images captured by hand held 
camera is a challenging topic, whose image resolution is low, 
and not all comer points can be detected. Edge information can 
provide more constraints about objects’ shape. But quickly 
searching corresponding edges from image sequence and 
reconstructing their accurate position correctly is difficult. In 
this method, only edges near the reliable matched points are 
considered as edge candidates, which can reduce the search 
space for corresponding 2D edges in frames and is much faster 
than applying epipolar beam from end points to find candidates. 
So, this method can significantly simplify and speed up the edge 
matching procedure. In order to avoid a poor ray intersection 
geometry caused by the short baseline between the images, all 
the edge candidates are used to estimate a 3D edge. Based on 
the estimated variance factor, only good 3D edge estimations 
are accepted, which ensures the accurate position of matched 
3D edges. But it should be mentioned here that the decided end 
points may not correspond to comer points. 
Although we can not get enough information from edge 
matching results to reconstruct the whole objects, these matched 
edges contain relation with points. Further on, surface patch and 
other feature aggregates are connected, which are useful 
topological constraints for recovering 3D structures. 
REFERENCES 
2d3, 2008. Homepage boujou software, http://www.2d3.com 
(accessed 27 March 2008) 
Armstrong, M. and Zisserman, A., 1995. Robust object tracking, 
Asian Conference on Computer Vision, pp. 58-61. 
Baillard, C., Schmid, C., Zisserman, A. and Fitzgibbon, A., 
1999. Automatic line matching and 3D reconstruction of 
buildings from multiple views, International Archives of 
Photogrammetry and Remote Sensing, pp. 69-80. 
Baltsavias, E.P., 2004. Object extraction and revision by image 
analysis using existing geodata and knowledge: current status