3D facial landmark localization using texture regression via conformal mapping

3D facial alignment typically requires the accurate estimate of facial landmarks. Most existing landmark detection methods use geometry characterization or resort to regression algorithms performed on point clouds or range images. A method combining both 3D geometry information and 2D texture has rarely been investigated. In this paper, we propose a novel 3D facial landmark localization algorithm, based on conformal geometric mapping, that can convert a 3D model to 2D using both geometry and texture information. Then a two-layers-regression method is used to improve the stability of landmark localization on the 2D geometry images. This method is impervious to pose changes and robust with respect to changes in expression. We evaluated the proposed approach on publicly available datasets and demonstrate how the use of 2D regression methods boosts the robustness and accuracy of 3D facial landmark localization.