PDF version of this document

Main Loop

• For all registration iterations:
  • Set the level of precision for the optimiser to reach. At present it increases as registration proceeds, but ideally it should increase when the advancements made are small.
  • For all data instances:
    • If the current data instance is not a reference:
      • Set up the positions of knot-points. Currently random placements with a sensible distribution are made.
      • Given the knot-points positions, apply warps to the current data instance and seek the warp parameters which minimise the cost , where $x$ is the complexity of the model built from the entire set of data.
    • end if
  • end for
• end for
• Statistics and registration logging take place.