Sciweavers

Share
PR
2006

Correspondence matching using kernel principal components analysis and label consistency constraints

8 years 10 months ago
Correspondence matching using kernel principal components analysis and label consistency constraints
This paper investigates spectral approaches to the problem of point pattern matching. We make two contributions. First, we consider rigid point-set alignment. Here we show how kernel principal components analysis (kernel PCA) can be effectively used for solving the rigid point correspondence matching problem when the point-sets are subject to outliers and random position jitter. Specifically, we show how the point- proximity matrix can be kernelised, and spectral correspondence matching transformed into one of kernel PCA. Second, we turn our attention to the matching of articulated point-sets. Here we show label consistency constraints can be incorporated into definition of the point proximity matrix. The new methods are compared to those of Shapiro and Brady and Scott and Longuet-Higgins, together with multidimensional scaling. We provide experiments on both synthetic data and real world data. 2005 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.
Hongfang Wang, Edwin R. Hancock
Added 14 Dec 2010
Updated 14 Dec 2010
Type Journal
Year 2006
Where PR
Authors Hongfang Wang, Edwin R. Hancock
Comments (0)
books