Accurate depth-map estimation for 3D face modeling

In the past few years several systems for tridimensional face reconstruction based on the analysis of 2D images have been proposed. The main goal of these systems is to provide fast and reliable 3D information for face recognition systems. Nonetheless, affordable image-based systems that are able to guarantee a high level of details and fast processing using commercial devices are far from being available. In this paper we propose a novel approach for a fast generation of detailed depth-maps of human faces based on a set of three calibrated cameras. The proposed algorithm is based on an fast implementation of the graph-cuts [1, 3] approach, which guarantees high-quality results in just a few seconds of processing time.