stanbul 2004 
) carve away 
will. First we 
be processed 
art of the true 
o voxels are 
th has been 
he following 
best view. 
mages. 
Set a fixed- 
t of the true 
nding image 
o images, in 
, we can first 
sure 3). From 
100Se two or 
sibility. This 
matching to 
image points 
'osscheck the 
ve reliability. 
rtion can be 
ting that this 
'oxels, which 
ved. So shell 
nber of fixed 
ls, until this 
a maximum 
ich evaluates 
fore deciding 
cond cube is 
le decisions. 
ceps track of 
e number of 
ade with any 
hes, with or 
ver, the two 
f successful 
ge will either 
alue inside a 
; scaled to fit 
5 is actually 
cube, we can 
1sidered. The 
e voxel, and 
llustrates the 
7. SUMMARY 
In this paper we presented shell carving as a refinement 
algorithm to the approximate model acquired by volume 
intersection method. An experimental image acquisition setup 
was explained on which the introduced algorithms were tested. 
Visibility information is recovered using line tracing. As an 
extension to line tracing, we introduced surface normal vector 
derived from a regional section of surface voxels. Two color 
image matching algorithms are introduced to search for image 
correspondences. Since RGB-triplets contain more information 
than a single gray value, these algorithms turned out to deliver 
more accurate results. We made use of the image orientation 
data for a knowledge based template grabbing. This takes 
perspective distortion into consideration and makes cross 
correlation insensitive to rotated images. The image matching 
was significantly improved by the knowledgc-based patch 
distortion. Fhe approximate model from volume intersection is 
improved successfully by combining all these tools and we 
presented some results. 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004 
REFERENCES 
Amanatides, J., Woo A., 1987. A Fast Voxel Traversal 
Algorithm for Ray Tracing. Proc. Eurographics '87, pp 1-10. 
Hartley R., Zisserman A., 2000. Multiple View Geometry in 
Computer Vision. Cambridge University Press. 
Kuzu, Y. Rodehorst, V., 2001. Volumetric Modelling using 
Shape from Silhouette. Fourth Turkish-German Joint Geodetic 
Days., pp. 469-476. 
Kuzu, Y. Sinram, O., 2002. Photorealistic Object 
Reconstruction using Voxel Coloring and Adjusted Image 
Orientations. ACSM/ASPRS Annual Conference, Washington 
DC, Proceedings CD-ROM, Proceed 00437.pdf. 
Matusik, W. Buehler, C. Raskar, R. Gortler, S. J. and 
McMillan, L., 2000. Image-Based Visual Hulls. SIGGRAPH 
2000 , Computer Graphics Proceedings, Annual Conference 
Series, pp. 369-374. 
Seitz, M. Dyer, R., 1997. Photorealistic Scene Reconstruction 
by Voxel Coloring. Proceeding of Computer Vision and Pattern 
Recognition Conference, pp. 1067-1073. 
  
  
Figure 10: Different viewpoints of the Nefertiti cube. The result of shape from silhouette (left), refinement by voting based 
carving (right)