Evaluating the feasibility and effectiveness of multi-perspective stereoscopy for 3D face reconstruction

This paper presents research on the feasibility and effectiveness of multi-perspective stereoscopy. A method is introduced to reconstruct a 3D face from a set of images taken by a camera array on a 2D plane. A physical setup is presented to dynamically emulate this array, along with a digital rendering setup. Quantitative measurements are shown on the performance of the physical setup, and the results of the digital setup are qualitatively compared. The influence of the camera baseline, subject surface angle and the number of cameras on the reconstruction quality is determined. Conclusions are drawn, but only for the current setup and methods.