Hypercube based interview prediction for multi-view video coding

Multi-view video coding (MVC) is being standardized by the MPEG committee. One of the important factors is improving the prediction efficiency by exploiting the spatio-temporal redundancies among the views. The prediction structure has to be selected to balance prediction efficiency as well as coding complexity. Increasing the number of reference views will improve the prediction but also increases the encoding and decoding complexity. A secondary goal is to compactly represent the prediction structure and the dependencies. In this paper we present a hypercube inspired prediction structure for inter-view predciton. The proposed structure balances the efficiency of prediction and coding complexity well. We present the results of motion estimation efficiency and compare it with a ‘grid’ prediction model. We show that using a Hypercube based model will improve the overall efficiency of the system.