A novel method for global disparity vector estimation in multiview video coding

One of the latest extensions for H.264 video coding standard is the multiview video coding (MVC), which can achieve excellent coding efficiency by not only inheriting from H.264 video standard but also incorporating new tools for interview prediction on the multi-view video sequences. A new prediction mode, motion skip mode, was considered for interview prediction as part of MVC. Motion skip mode requires estimating the global disparity vector (GDV). In MVC reference software JMVM, GDV was obtained by frame matching on neighboring views, in which the high complexity is involved. Therefore, in this paper, we present a novel method of fast GDV estimation for the motion skip mode, which is based on edge projection map between adjacent views in MVC. Experimental results show that our proposed method can reduce the complexity while keeping R-D performances on MVC anchor pictures coding.