Study on the Optimization Method of Dynamic Reconstruction of 3D Non- Rigid Image

Background: The global registration algorithm is adopted to optimize the ICP algorithm, each match point is matched simultaneously for the entire depth image registered with ICP, in order to obtain the optimal solution of translation and rotation matrix for images of all perspectives. Methods: In the process of dynamic reconstruction of a three-dimensional (3D) non-rigid image, some image regions overlap, in the traditional 3D non-rigid dynamic image reconstruction method based on FDK, all the overlapping regions are calculated with ICP, which greatly increases computational complexity. A dynamic reconstruction optimization method for the 3D non-rigid image is raised based on the global registration, the disadvantages of ICP algorithm are analyzed, and the global registration algorithm is adopted to optimize the ICP algorithm, each match point is matched simultaneously for the entire depth image registered with ICP, in order to obtain the optimal solution of translation and rotation matrix for images of all perspectives. Result: The feature vectors in the process of 3D image reconstruction are classified. The empirical knowledge between the feature vectors and information are utilized to train 3D reconstruction of learning samples. The simulation results show that the method shortens the time of 3D image reconstruction and improves the accuracy of 3D image reconstruction.