Registration of Arbitrary 3D Point Cloud Using Statistical Head Model

Abstract

The standardization of arbitrary 3D head mesh is necessary for many applications. This project explores one possible standardization method which made used of traditional 3D registration techniques (i.e. the Iterative Closest Point algorithm) and a given statistical head mesh model. During the process, missing information in the original head mesh can also be recovered based on the information contained in the statistical model. The project also seeks to automate the process by using Harr-like features to detect feature points required for the initial alignment.