Performance Modeling and Optimization of Hierarchical B Pictures Based on Directed Tree

A general methodology for performance modeling and optimization of hierarchical B pictures used in video coding is proposed in this study. First, an approach of directed tree decomposition is used to denote hierarchical B pictures. In combination with a proposed linear model for compression efficiency, the performance of any hierarchical B prediction structure can be evaluated conveniently from the directed tree. Then with a dynamic programming method, the optimal tree for any group length of hierarchical B pictures can be set up elegantly from recursive subtrees. This method can be adapted to tradeoff between compression efficiency and random access ability. Besides, experimental results show that the optimal tree of compression efficiency achieves higher performance than the existing hierarchical B prediction structures.