Temporal prediction improvement for parallel processing of HEVC

The novel coding standard high efficiency video coding (HEVC) is based on a block-based hybrid coding architecture. Parallel processing is important tool to accelerate both on the encoder and decoder side. Advanced motion vector prediction (AMVP) is an efficient inter prediction tool in HEVC. However, neighbor block dependence in AMVP coding is major barrier for parallel processing. In this work, focusing on the parallel processing ability in HEVC, a parallel temporal prediction algorithm is proposed that make the encoder decide the temporal candidate once in a larger coding unit for all the sub-units. Experimental results demonstrate that the proposed algorithm realize parallel temporal prediction candidate with only 0.11% BD-rate loss, compared with the reference software model.