An Improved 3D Human Face Reconstruction Approach Based on Cubic Splines Models

In this paper, we develop a new hybrid active vision/geometric modeling approach dedicated to 3D human face recovery. Initially, a 3D coarse reconstruction is obtained via a structured-light assisted stereo sensor. Here, the stereo matching problem is resolved through two main stages: (a) sub-pixel stripe edge localization of projected patterns of light, and (b) correspondences establishing based on adaptive dynamic programming optimization technique. Next, the introduction of smooth interpolation models achieves the fine reconstruction. Here, cubic spline curves are employed in order to improve the quality of the reconstructed models. Indeed, they allow us to produce dense and details preserving reconstructions by following the control points from the coarse reconstruction stage. Furthermore, we present some reconstruction results and discuss both qualitative and quantitative evaluations of the proposed reconstruction scheme.