Joint Estimation of Bidirectional Prediction in H.264/AVC

摘要

An improved iterative joint optimization algorithm specific for the bidirectional prediction in H.264/AVC is presented in this paper. Beginning with the initial values resulted from the commonly-used independent search, the forward and backward motion vectors are alternatively and iteratively refined by searching over the quarter pixel accuracy neighborhood and targeting joint optimization objective, until the cost of the macroblock stops decreasing. At first, all seven macroblock partitioning modes are employed to find updated forward or backward motion vectors. Then best macroblock partitioning modes from the independent search are inherited to further simplify the algorithm. Computer simulation results demonstrate that this approach can greatly improve the performance of the H.264/AVC encoder at small cost of computational complexity increment.