Non-rigid Medical Image Registration Based on the Thin-Plate Spline Algorithm

摘要

For the registration of large image distortion of some special places, the non-rigid registration is more accurate and necessary compared with the rigid registration. In this paper, a kind of non-rigid medical image registration algorithm based on the thin-plate spline is presented and discussed in detail. The hybrid energy formula is presented. In the process of registration, the thin-plate spline algorithm is proposed to deal with the whole image registration. In order to reduce and locate the registration region, the iterativeness of simulated annealing algorithm is proposed. In the local pending registration area, mutual information method is used and the local distortion is adjusted to resolve the incomplete match caused by the lack of characteristic points, which do help to get smoother image and better effect of non-rigid image registration. The dynamic outlier removal method is used to deal with outlier and make sure the accurate transformation function. In addition, some empirical data was modified for the purpose of getting better effect. It was proved by the experiments that the non-rigid medical image registration made a better registration effect for the large image distortion.