| This relatively known
work (*) is a strikingly simple algorithm for matching
points features across pairs of images.
Despite the well-known combinatorial complexity of the
problem, this work shows that an acceptably good solution can be
obtained directly by singular value decomposition of an
appropriate correspondence strength matrix.
The approach draws from the method proposed by Scott
and Longuet-Higgins but, besides suggesting its usefulness for
stereo matching, in this work a correlation-weighted proximity
function is used as correspondence strength to specifically
cater for real images.
(*) Hundreds of people requested the code and the
method has been used also in computer vision classes.
|
Publications:
M. Pilu "A
direct method for Stereo Correspondence based on Singular Value Decomposition",
IEEE
International Conference of Computer Vision and Pattern Recognition, Puerto Rico, June
1997 (Postscript)
M. Pilu, A. Lorusso
"Uncalibrated Stereo Correspondence based on Singular Value Decomposition",
British Machine Vision Conference, Sept 1997
Also available as HPL Technical Report:
M.
Pilu "Stereo Correspondence by Singular Value Decomposition" ,
HPLB Technical Report N. HPL-97-96, Sept 1997. (PDF)
Code:
Download the
code in Matlab
(shift+click to save to a location)
Download the
code in C including
the SVD correspondence method, simple fundamental matrix fitting with RANSAC, examples and
test and display routines in MATLAB. (Please read the
disclaimer here before downloading!) |
|
printable
version
