A novel point pattern matching method is presented for medical image registration.It can jointly estimate the correspondence and non-rigid transformations between two point patterns that may be of different sizes.The algorithm utilizes the filter method for the correspondence probability

then the relaxation labeling is used for consistent result.And for final robust point pattern matching

it is embedded in an iterative matching frame

in which the Gaussian radial basis function is used for the non-rigid mapping.Experiments on both synthetic point-sets and on real world data show that matching errors of the algorithm were less than 0.13 while degree of deformation was 5%

outlier ratio was 50% and noise standard error was 5% respectively.