A Motion-Based Partitioning Algorithm for HEVC Using a Pre-Analysis Stage

Abstract

The High Efficiency Video Coding (HEVC) standard has shown large improvements in coding efficiency compared with previous standards. In particular, HEVC outperforms H.264/MPEG-4 Advanced Video Coding (AVC) by up to 50% in terms of bitrate reduction for similar perceptual quality. This improvement is the result of the introduction of new coding tools which enable the representation of data using fewer bits, but at the cost of long computation times. One of the most significant tools introduced by HEVC is the novel quadtree-based structure called the coding tree unit (CTU), which can be subsequently split into coding units (CUs), prediction units (PUs) and transform units (TUs), providing huge flexibility in the encoding. However, selecting the optimal tree partitioning requires the evaluation of a huge number of possibilities, which constitutes the most complex operation for the encoder. In order to tackle this task, this paper proposes a CTU/CU partitioning algorithm based on a pre-analysis stage. This stage performs a fast motion estimation that provides preliminary information to the encoder, including estimate distortion costs, which enables the building of the leastcost quadtree. On the basis of this tree, the evaluation performed by the encoder is restricted to a subset of PUs. After a thorough statistical analysis of numerous CU/PU combinations, we propose two different configurations that prioritize each of the target variables: coding efficiency and time reduction. In the former case, 58.09% of the encoding time can be saved at the cost of a 2.51% increase in BD-rate, while a larger reduction of 63.08% is achieved with a 3.43% increase in BD-rate in the latter case.

A Motion Estimation based Algorithm for Encoding Time Reduction in HEVC

“…This approach helps to decrease the encoding time. Cebrián-Márquez 9 , et al uses the pre-analysis stage, which performs block-based motion estimation to estimate Rate-Distortion cost. The estimated cost is used to build the optimal quad-tree by omitting a large number of unnecessary partitions, which reduces the encoding time.…”