Virtual view interpolation and prediction structure for full parallax multi-view video

Abstract

In this work, we propose virtual view interpolation algorithm and prediction structure for full parallax multi-view video coding. The proposed virtual view interpolation algorithm estimates the disparity information by bilateral criterion and disparity smoothness. The interpolated view is used as additional reference frame, and novel direct mode is applied to it. Simulation results demonstrate that our proposed virtual view interpolation algorithm shows the better quality than the conventional interpolation algorithm. Moreover, the proposed prediction structure, which uses the interpolated view as an additional reference with novel direct mode, shows the significant coding gain. © 2009 Springer-Verlag Berlin Heidelberg.