Enhanced Motion Vector Prediction for Video Coding

Abstract

Motion vector prediction (MVP) plays a crucial role in reducing the rate cost for coding motion vector (MV). However, the relative limited construction of MVP list restricts the potential efficiency. In this paper, an enhanced MVP scheme is proposed. In particular, we first modify the insertion order of MVP candidates according to the statistics. Then, the scaling process during MVP construction is modified to extend the application range. Moreover, an expanded-area motion vector prediction (EMVP) approach is adopted to further utilize the spatial correlation of local motion field. Simulation results shows that the proposed scheme can achieve better prediction for coding motion information.