Figure 4.

(A) 3D meshes indicating the distance to the closest vertex point from the reconstructed model to the ground truth model obtained from FaceSCAN^3D. Difference values are plotted on the ground truth model. Vertex point distances are color coded, ranging from 0 mm to 2.95 mm (the 95^th percentile). (B) Descriptive curves for downsampling image count (Dimg) and image resolution (Dres). The blue and green lines indicate the quadratic fitting curves for the respective data. Dotted and dash-dotted lines indicate the respective 95% confidence interval of the fitting curves. X-axis values represent the relative downsampling rates corresponding to Table 1. Top left plots the MPC density relative to the full-resolution model, top right likewise for DPC density, bottom left for overall time consumption relative to the full-resolution model and bottom right for the average deviation to the full-resolution model in mm. (C) An example showing the area that needed to be removed in order to make the meshes covering a comparable area for each downsampling step for image count. (D) Additionally loss of vertex points in order to make the meshes covering a comparable area for each downsampling step (Dimg). The dashed area indicates the not interpretable result of having a negative loss of information as a function of downsampling. At 50% downsampling, 7.8% of spatial information was lost.