Shape model fitting algorithm without point correspondence

In this paper, we present a Mean Shift algorithm that does not require point correspondence to fit shape models. The observed data and the shape model are represented as mixtures of Gaussians. Using a Bayesian framework, we propose to model the likelihood using the Euclidean distance between the two Gaussian mixture density functions while the latent variables are modelled with a Gaussian prior. We show the performance of our MS algorithm for fitting a 2D hand model and a 3D Morphable Model of faces to point clouds.