Files

Abstract

Multi-view video plus depth (MVD) is gathering huge attention, as witnessed by the recent standardization activity, since its rich information about the geometry of the scene allows high-quality synthesis of virtual viewpoints. Distributed video coding of such kind of content is a challenging problem whose solution could enable new services as interactive multi-view streaming. In this work we propose to exploit the geometrical information of the MVD format in order to estimate inter-view occlusions without communication among cameras. Experimental results show a bit rate reduction up to 77% for low bit rate w.r.t. state-of-the-art architectures.

Details

Actions

Preview