dpshepherd
commented
2 months ago This looks promising: https://github.com/yongxb/OpticalFlow3d
Just need to validate order for mappings, if we use the forward or inverse based on library conventions, and then test with simpleITK warping.
Given @AlexCoul feedback and my experience with lack of robustness with DEEDS, look into a different deformable registration algorithm. It also is a pain to compile on certain systems and we are using Alexis' fork which is not ideal.