A robust and accurate facial 3D reconstruction method from images acquired by mobile device at home for facial growth monitoring University of Dundee

Overview

Context of  A robust and accurate facial 3D reconstruction method from images acquired by mobile device at home for facial growth monitoring project at University of Dundee

Diseases affecting facial growth require highly accurate facial 3D scan to diagnose, monitor and plan treatment. Currently, the gold standard for capturing such data is to use digital stereo-photogrammetry systems in a clinical setting with healthcare professionals allowing for sub-millimetre accuracy. Recently, 3D reconstruction from facial images captured by smartphone has shown potential as an alternative in the same settings. The project in which this PhD takes place aims to instead use patients’ smartphones at home to acquire by themselves images suitable for highly accurate facial 3D reconstruction through a purpose-made mobile application. 

In Computer Vision, highly accurate facial 3D models can be obtained from multiple images with optimization-based methods relying on motion analysis and geometrical constraints . Recently, deep learning approaches combining geometry and light properties estimation have emerged. While they achieve impressive results from images captured by laypersons, these methods do not leverage all the progress made to integrate multi-view geometry constraints in deep learning-based Structure-from-Motion. They also often rely on 3DMM models which are not representative of the morphology of patients with facial dysmorphology.