SVM-based Macroblock Mode Decision Algorithm for MVC

The exhaustive full mode decision algorithm is applied to both temporal and inter-view prediction by joint multi-view video coding (JMVC) to achieve the best performance for multiview video coding (MVC).Both the high compression ratio and the high complexity will seriously affect the practical use.A SVM-based macroblock mode decision algorithm for MVC using a JMVC platform is proposed in this paper.A three-stage SVM classifier is designed and the coded information between views is used as training data.In this paper, the applied three-stage classification strategy is discussed and the selected feature vectors are described in detail.Experiment results show that compared with the JMVC full mode algorithm, the encoding time can be remarkbaly saved by the proposed algorithm while maintaining nearly the same video quality.