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

TL;DR

This paper introduces SFGP, a probabilistic shape registration method based on Gaussian Processes, designed to effectively handle extensive missing data in 3D shape modeling, outperforming existing methods especially in detailed and deformed regions.

Contribution

It formulates shape fitting as multi-annotator Gaussian Process Regression, improving registration accuracy with missing data compared to prior approaches.

Findings

SFGP outperforms state-of-the-art registration methods with extensive missing data.

Demonstrated effectiveness on 2D and 3D datasets, including ears.

Better handling of detailed and deformed shape regions.

Abstract

We propose a shape fitting/registration method based on a Gaussian Processes formulation, suitable for shapes with extensive regions of missing data. Gaussian Processes are a proven powerful tool, as they provide a unified setting for shape modelling and fitting. While the existing methods in 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 SFGP shows better performance when dealing with extensive areas of missing data when compared to a state-of-the-art registration method and current approaches for registration with pre-existing shape models. Experiments are conducted both for a 2D small dataset with diverse transformations and a 3D dataset of ears.