search

InsightSoftwareConsortium / itk_cucim

ITK Python filters accelerated with cuCIM
Apache License 2.0
3 stars 3 forks source link

Implement ants_registration #10

Open thewtex opened 2 years ago

grlee77 commented 2 years ago

The initial SyN code I implemented in cuDIPY only handled dense displacement grids and is missing support for control point grids.

I am not starting this now, but for reference, we can probably adapt some existing Python code for handling b-spline control point grids from nitransforms (MIT-licensed): https://github.com/nipy/nitransforms/blob/master/nitransforms/interp/bspline.py https://github.com/nipy/nitransforms/blob/22.0.0/nitransforms/nonlinear.py#L129

grlee77 commented 2 years ago

Also only a cross correlation metric is currently in cuDIPY.

A different MutualInformationMetric built on a Parzen joint histogram is the default in the rigid registration case, but has not yet been implemented on the GPU.