Iterative Tensor-Based Registration
Haacke E, Rajaram R, Ciulla C
The MRI Institute for Biomedical Research
This work presents an algorithm for registering 3D-MRI brain images. Processing steps consist of the application of: (i) a low pass Hanning filter to the k-space data, (ii) the Fourier Transform Theorem to correct for misplacement, and (iii) iterative enforcement of tensor-based registration. Results show improvement of registration both qualitatively as substantiated by a clear reduction of Gibbs ringing artifacts, and quantitatively by the reduction of the shift error to less than 0.001 of the voxel size for translation and 0.01 for rotation.