Scene-Adaptive Disparity Estimation in Depth Sensor Networks Using Articulated Shape Models

Abstract

In this work a scene-adaptive approach for disparity estimation in depth sensor networks is presented. Our approach makes use of a priori scene knowledge to improve the low-level 3D data acquisition. We fit an articulated shape model to the given 3D data and leverage the resulting high-level scene information to make the estimation of disparities more robust to local ambiguities. We present early qualitative results to show the applicability of our method.