Fast CU size decision for HEVC intra coding by using local characteristics and RD costs

Abstract

The newest video coding compression standard is High–Efficiency Video Coding (HEVC) which supports high-resolution video sequences. Also, HEVC has about 50% bitrate saving compared with the former standard H.264/AVC. The quad-tree-based Coding Unit (CU) partitioning process is one of the most efficient technologies in HEVC. Its computational complexity, however, becomes main issue when attempting implementation with an encoder. A fast CU size decision algorithm for the intra-predicted picture of HEVC is proposed in this paper. We utilize the image complexity and RD cost values for early split decision. Also, the Bayesian decision rule are used for early termination of the CU partitioning process. Our experimental results show that the proposed algorithm significantly reduces the encoding time by about 55.7% with small BD-BR loss (1.56%) compared to the HEVC reference software HM 16.0.