Adaptive motion vector resolution based on the rate-distortion cost and coding unit depth

Abstract

In current research on high-efficiency video coding (HEVC), motion vector resolution is always fixed to 1/4 pixel for the entire video sequence. Inter-coding with a fixed motion vector resolution can decrease the coding efficiency because the statistical properties of the local image are not considered. In this paper, we propose an adaptive decision scheme for motion vector resolution to improve the coding efficiency. The proposed scheme capitalizes on the tendency for a high-pel-precision level to be beneficial in terms of coding efficiency as the coding unit (CU) depth decreases. Also, we determined the strength with a rate-distortion (RD) cost and selected a predefined threshold set per slice level. Simulation results with respect to HM7.0 show that the proposed scheme provides a coding gain of 2.4% for a low-delay structure. Moreover, it was found that the average encoding time is reduced by 5%. The proposed scheme can also improve the coding efficiency at a slightly increased encoding time compared to conventional methods.