My image registration method uses the mean weighted ratio of two medical images and computes the minimum of this function using a novel Chebyshev polynomial approximation iteration loop Weighting is performed by setting the ratios between signal and background voxels to a standard high value and the mean value is computed over the union of the signal areas of the two images.
A large number of experiments were performed for rigid body registration during my master's work at the Department of Musculoskeletal Radiology at The Cleveland Clinic Foundation and showed that the accuracy of the method for 3D registration is sub-voxel using n=4 Chebyshev points .
The method is currently under development and I was able to achieve 2d non-rigid registration using n=4 Chebyshev points and a local geometric transformation model .
Another extention of the method is the 2D and 3D rigid registration using 1D and 2D binary projections respectively with n=5 Chebyshev points. This method was found to be more robust and accurate compared to two Mutual Information based methods.
In this Section there are links for :
HTML Hit Counter
|