IPM Vision Group |
School of Mathematics Scientific Computing Center |
publication - software - links - people
Shape Matching |
In this work we presented a novel and fully automated method for measuring similarities between shapes and using it for shape matching. The computational complexity of reliable shape matching algorithms, to the extent that they exist, are generally formidable. Extensive tests on shapes with different orientations show that the proposed approach works just as well as the best available techniques (over 95% accuracy) and has the advantage that its computational complexity is better at least by a factor of fifty.
The CPU time requirement for the execution of this program (187 shapes) on a single 1.7 GHZ (Intel-Centrino) processor was 5.4 minutes. The Belongie-Malik (shape context for shape matching) algorithms on a dual 1.7 GHZ (Athlon) processor required over 180 minutes of CPU time.
The full description of the method and its MATLAB implementation will be available in April 2006.
To see results click here.
Authors:
Nima Razavi, Masoud Alipour.
© Copyright 2003-2005 - Institute for Studies in Theoretical Physics
and Mathematics (IPM).
All rights reserved. Please submit your comments or questions
here.
Last update:12/07/2005.