Automatic landmark annotation and dense correspondence registration for 3D human facial images

TL;DR

This paper introduces a fully automatic, robust, and efficient non-rigid registration method for 3D facial images, enabling dense correspondence mapping useful for diverse applications like genetics and forensics.

Contribution

The novel approach combines automatic landmark annotation with a thin-plate spline protocol for dense registration, improving accuracy and efficiency across different ethnicities.

Findings

Method is robust and accurate across ethnic groups.

Enables high-throughput analysis of facial features.

Provides dense correspondence mapping for 3D facial images.

Abstract

Dense surface registration of three-dimensional (3D) human facial images holds great potential for studies of human trait diversity, disease genetics, and forensics. Non-rigid registration is particularly useful for establishing dense anatomical correspondences between faces. Here we describe a novel non-rigid registration method for fully automatic 3D facial image mapping. This method comprises two steps: first, seventeen facial landmarks are automatically annotated, mainly via PCA-based feature recognition following 3D-to-2D data transformation. Second, an efficient thin-plate spline (TPS) protocol is used to establish the dense anatomical correspondence between facial images, under the guidance of the predefined landmarks. We demonstrate that this method is robust and highly accurate, even for different ethnicities. The average face is calculated for individuals of Han Chinese and Uyghur origins. While fully automatic and computationally efficient, this method enables high-throughput analysis of human facial feature variation.