Surface measures for accuracy evaluation in 3d face reconstruction

In this paper, we propose surface based metrics to evaluate the accuracy of the 3D face reconstruction algorithms. The continuous 3D surfaces obtained due to reconstruction algorithms for a 3D face is considered to obtain a numerical estimate. The surface based measures namely the volume of difference, average edge distance and intercept edge distance metrics are proposed. These measures are computed by considering continuous 3D surfaces to ascertain the quality of the reconstruction techniques. Experimental validation of the proposed approach is performed using the data acquired by 2D cameras and 3D scanning technologies.

► A new frame work for accuracy evaluation of reconstructed 3D face model. ► In this numerical approach, the numerical estimation is obtained by computing the newly introduced surface based metrics. ► The surface based metrics namely volume of difference, average edge distance and intercept edge distance metrics are proposed to obtain the numerical estimate. ► The proposed approach is independent of algorithms designed for 3D face reconstruction and of equipment.