A Fast Intra Coding Unit Size Decision Based on Statistical Learning for HEVC

High Efficiency Video Coding (HEVC) is an ongoing standard, and it employs the quad-tree block partitioning structure which includes coding unit, prediction unit, and transform unit. This content-adaptive coding tree structure can improve HEVC coding efficiency significantly, but it also consumes large computational complexity. This paper proposed a fast intra coding unit size decision algorithm to reduce the heavy complexity of HEVC encoding. First, the proposed algorithm reduced unit sizes search by using the classifier, which is based on the statistical learning. Second, an early largest unit size decision was designed to skip the checking of unnecessary unit sizes. As compared to the full search algorithm in HEVC reference software, experimental results show that the proposed algorithm achieves 50.4% computation saving on average with 1.83% bit rate increase and 0.070dB peak signal-to-noise ratio loss.