The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
644 
method does not work properly in this situation. Figure 3 shows accurately estimates camera pose and building dimensions 
the recovered camera pose and wire frame of the Burnside Hall provided that accurate image measurements are available, 
using MatLab. 
Unit: mm 
Dimensions 
measured 
using a ruler 
Dimensions computed 
from the image 
Left 
Box 
Width 
V 
71.1 
73.1 
M 
70.9 
Height 
V 
123.1 
97.9 
M 
122.6 
Right 
Box 
Length 
V 
72.1 
68.8 
M 
71.7 
Width 
V 
50.6 
47.6 
M 
49.6 
Height 
V 
15.3 
6.3 
M 
14.8 
Table 5. Comparison of the vanishing points based method with 
the model based method using real image Fig2.a 
Figure 3. Visualization of the recovered camera pose and wire 
frame of the Burnside Hall 
4. CONCLUSIONS 
This paper presented a method to recover 3D rectilinear 
building models from single monocular images. The method 
uses the correspondences between predefined 3D models and 
their corresponding 2D images to obtain camera pose as well as 
parameters of 3D building models. The camera orientation is 
first recovered followed by solving translation and the first 
building model dimensions. The direct computation of the 
initial estimate for camera rotation effectively solved problems 
in the previous approaches (e.g., Taylor and Kriegman, 1995), 
and the determination of camera pose as well as the first 
building model dimensions are much simpler than the previous 
methods (e.g., Debevec et al., 1996). Under the assumption of 
flat terrain, more 3D building models can be reconstructed 
based on recovered camera pose through model-to-image 
correspondence. 
Simulation experiments were carried out in order to investigate 
how the accuracy of the algorithm would be affected as 
different parameters were varied. The comparison using 
identical synthetic and real data shows that our method is 
significantly superior over the vanishing points based method. 
The experiments also show that our method robustly and 
REFERENCES 
Camillo J. Taylor and David J. Kriegman, 1995. Structure and 
motion from line segments in multiple images, IEEE 
Transaction on Pattern Analysis Machine Intelligence, 17(11). 
Caprile B. and Torre, V., 1990. Using vanishing points for 
camera calibration, International Journal of Computer Vision 
(4), No. 2, pp. 127 -140 
E. Guillou, D. Meneveaux, E. Maisel, and K. Bouatouch, 2000. 
Using vanishing points for camera calibration and coarse 3-D 
reconstruction from a single image, Visual Computer, vol. 16, 
no. 7, pp. 396^110. 
Debevec P., C. Taylor, and J. Malik, 1996. Modeling and 
Rendering Architecture from Photographs: A Hybrid Geometry 
and Image-based Approach, SIGGRAPH, pp. 11 -20. 
F. A. van den Heuvel, 1998. Vanishing point detection for 
architectural photogrammetry. In International Archives of 
Photogrammetry and Remote Sensing Vol. XXXII part 5, pages 
652-659. 
Horn, B.K.P, 1990. Relative Orientation, International Journal 
of Computer Vision Vol. 4, No. 1, pp. 59—78. 
Hoiem D., A. Efros, and M. Hebert, 2005. Automatic photo 
pop-up. In ACM SIGGRAPH. 
Kumar, Rakesh, and Hanson, Allen R, 1994. Robust methods 
for estimating pose and a sensitivity analysis. CVGIP: Image 
Understanding, v 60, n 3, p 313-342. 
Leibowitz D, Criminisi A, Zisserman A, 1999. Creating 
architectural models from images. In: Brunet R, Scopigno R 
(eds) Eurographics, Vol 18, pp 39-50, Blackwell Publishers. 
Liu, Y.; Huang, T.S.; Faugeras, O.D., 1990. Determination of 
Camera Location from 2-D to 3-D Line and Point 
Correspondences, IEEE Transactions on Pattern Analysis and 
Machine Intelligence, 12(1), Page(s):28 - 37 
S. Birchfield, 1998. An introduction to projective geometry. 
Unpublished, http://vision.stanford.edu/~birch/projective. 
T. Tuytelaars, M. Proesmans, and L. Van Gool, 1997. The 
cascaded hough transform. The proceedings of International 
Conference on Image Processing. 
Wang, R. and F. Ferrie, 2008. Camera Localization and 
Building Reconstruction from Single Monocular Images. The 
IEEE Computer Vision and Pattern Recognition (CVPR) 
Workshop on Visual Localization for Mobile Platforms, 
Anchorage, Alaska. 
Zhang Z., Zhang J., Zhang S., 2001. 3D Building 
Reconstruction from Single Images. Workshop on Automatic 
Engineering Surveying of Channel.