Fitting a morphable model to pose and shape of a point cloud

The paper addresses the task of fitting a morphable head model to a dense, unstructured and untextured point cloud. The problem is typically approached in a multi-step process, comprising of generic non-rigid registration, conversion to the model's topology and fitting of the model. Here, a direct approach is proposed where the morphable model is fitted to the point cloud itself in an optimization process following the Iterative Closest Points framework. Along with shape parameters, rigid pose and scale arc estimated explicitly which leads to better fitting results than shape estimation alone. A compact formulation of a cost function is proposed, applicable also in the case of a model comprising of multiple, independent sub-models each of which describes a region of the face. The use of the algorithm for other tasks such as adding depth to mugshot-style face photos is discussed.