Fast motion and disparity estimation of multi-view video coding

To reduce the complexity of multi-view video coding, a fast joint motion and disparity estimation algorithm based on the view-temporal hierarchical B picture coding structure is proposed. By using the geometry correlation between motion and disparity vectors among neighboring view frames, a reliable starting point is found firstly. An iterative search strategy is then carried out around the starting point to get optimal motion vector and disparity vector in both temporal and view directions. The proposed algorithm takes use of the prediction vector of previous motion or disparity field for the next disparity or motion field. It can get both motion and disparity vectors simultaneously during one estimation process. Experimental results demonstrate that the proposed algorithm can successfully save up to 87.69% computational complexity as compared with JMVM full search algorithm with a negligible quality drop.