Using Bayesian Classifiers For Low Complexity Multiview H.264/avc And Hevc Hybrid Architecture
Antonio Jesus Diaz-Honrubia, Johan De Praeter, Sebastiaan Van Leuven, Jan De Cock, Jose Luis Martinez, Pedro Cuenca

In order to enable a system which offers compatibility with currently existing H.264/AVC based systems, 3D functionality, and a low overall bitrate, a multiview H.264/HEVC hybrid architecture was proposed in the context of 3D applications and standardization. This paper presents an algorithm to reduce the complexity of this multiview hybrid architecture by reducing the encoding complexity of the HEVC side views. The proposed technique exploits the information gathered in the center view of the H.264/AVC encoder of this hybrid multiview architecture to make decisions on Coding Units splitting in HEVC side views using a Na´ve-Bayes probabilistic classifier. Thus, the proposal is quite novel since no similar works has been found in the literature and it exploits a new characteristic of the multiview HEVC streams which was not present in previous standards. Experimental results show that the proposed algorithm can achieve a good tradeoff between coding efficiency and complexity.