Multimodal image registration is a fundamental problem in medical image
analysis. In this paper, we propose a novel algorithm to compute the
local frequency representations of the multimodal data sets to be
registered. Local frequency representation can detect edge and ridge
information simultaneously. In this algorithm, we develop regularized quadrature
filters (RQFs) to compute local frequency maps, which are relatively
insensitivity to noise in comparison to standard QFs. The local
frequency maps thus obtained are used as an underlying representation
to which a statistically robust matching technique is applied, to
estimate a parameterized transformation between the volume data sets.
We present experimental results for registering several pairs
of CT-MR data sets along with comparisons to other matching
methods.
