M. Louw and F. Nicolls (South Africa)
Feature matching, MRF, LBP, sparse correspondence.
In this paper we develop a novel MRF formulation for
calculating sparse features correspondence in image pairs.
Our MRF terms can include cliques of variable sizes, and
solve these using Loopy Belief Propagation. To calculate
our MRF topology we develop a variant of the K-means
algorithm which we call the KN-means algorithm (where
each mean has a speciﬁed number of neighbours). The
method is compared to other state of the art sparse feature
correspondence algorithms and shown to compare well, es
pecially for less dense feature sets. Outliers are handled
naturally within this paradigm.