Abstract
High efficiency video coding (HEVC) achieves significant higher coding efficiency than previous video coding standards. In the intra prediction of HEVC, prediction unit sizes from 4×4 to 64×64 are employed, respectively defined as levels 1 to 5, to achieve higher coding efficiency. Nevertheless, this is at the expense of high computational complexity. This paper proposes a fast algorithm for the intra prediction of HEVC, which can efficiently decide the level on the basis of the roberts-cross edge detector. The proposed algortihm utilizes the high correlation between regional texture and prediction unit partitioning. It is mainly composed of a bottom-up level decision method and an efficient decision flow based on an authentic image feature. Furthermore, chrominance information is also employed to decide the prediction unit partitioning. The experimental results for the proposed algorithm demonstrated that it achieved a task with greatly reduced computational complexity compared with the original HEVC. The average time-saving is approximately 37%, while the increase in bit rate and decrease in PSNR are negligible.
