3D Skull Registration Based on Registration Points Automatic Correspondence

Abstract

Aiming at the precise 3D skull registration, ball volume integral invariants for multi-scale constraints is introduced to realize point correspondence between registration point sets of reference skull and that of target skull, a method of 3D skull registration based on ICP and TPS is proposed. Firstly, get reference skull and target skull unified to frankfurt coordinate system. Secondly, establish registration point sets for the two skull, the sets inlcude feature points calibrated by semi-automatic method and sampling points selected through octree. Thirdly, use the multi-scale constraints based on volume integral invariants to complete the point correspondence. Finally, carry out ICP to realize coarse registration and use TPS to achieve accurate global non-rigid registration. The experiments show our algorithm improves the precision of skull registration effectively.