Freesurfer comes with a variety of different registration tools designed for specific purposes. This page is supposed to highlight a few of them and describe the context in which they should be used:
mri_robust_register: 2 inputs, rigid/affine, same or cross-modality, unbiased
A tool for pairwise image registration. Originally designed for highly accurate same modality and same subject registration (across time). It can compute rigid or affine transforms. It is robust with respect to outlier/anatomy change (removing their influence in the registration) and is inverse consistent (symmetric). More recent versions (dev) are capable of registering across modality in a symmetric fashion. robust_register is also used to register histology scans (part to whole hemisphere), etc.
mri_robust_template: n inputs, rigid/affine, same-modality, unbiased
A tool to register several inputs (same modality) to an unbiased average template, that it creates at the same time. Registration can be rigid or affine. Inputs are usually of the same subject, e.g. at different times. robust_template is based on robust_register and inherits the robustness to outlier. It is used in recon-all in the motion correction step (to co-register several within session scans to create one higher SNR input) and for longitudinal processing to create the unbiased average template.
bbregister: 2 inputs, rigid, cross-modality, needs surfaces
This program performs within-subject, cross-modal registration using a boundary-based cost function. The registration is can be 6, 9, or 12 DOF DOF (rigid-to-affine). It is required that you have an anatomical scan of the subject that has been processed in Freesurfer.
mri_cvs_register: 2 inputs, non-linear, same-modality, needs surfaces
This program performs subject-to-subject or subject-to-atlas non-linear volume registration using a combined volumetric and surface-based (CVS) registration algorithm. It is required that you have an anatomical scan of the subject that has been processed in Freesurfer!
mri_em_register: 2 inputs, linear, atlas registration
This generates a linear talairach transform (lta file) from T1 anatomical to gca file (probabilistic atals). It is used in recon-all to create transforms/talairach.lta .
mri_ca_register: 2 inputs, non-linear, atlas registration
This generates a multi-dimensional talairach transform from T1 anatomical to a gca file (probabilistic atals) using a linear talairach (registration) file as initialization. It is used in recon-all to create transforms/talairach.m3z .
other registration tools
tkregister - graphical tool to manually register two images
FreeView - graphical viewer that can also do simple rotations, cropping