Probabilistic Registration for Gaussian Process 3D shape modelling in the presence of extensive missing data

Gaussian Processes are a powerful tool for shape modelling. While the existing methods on this area prove to work well for the general case of the human head, when looking at more detailed and deformed data, with a high prevalence of missing data, such as the ears, the results are not satisfactory. In order to overcome this, we formulate the shape fitting problem as a multi-annotator Gaussian Process Regression and establish a parallel with the standard probabilistic registration. The achieved method GPReg shows better performance when dealing with extensive areas of missing data when compared to a state-of-the-art registration method and the current approach for registration with GP.