Hierarchical Inter Mode Decision for HEVC
Huanqang Zeng, Wenjie Xiang, Canhui Cai, Jing Chen
School of Information Science and Engineering
Huaqiao University
Xiamen, China, 361021
zeng0043@hqu.edu.cn, yyxiangwj@163.com,
chcai@hqu.edu.cn, jingzi@hqu.edu.cn
Kai-Kuang Ma
School of Electrical and Electronic Engineering
Nanyang Technological University
Singapore, 639798
ekkma@ntu.edu.sg
Abstract—The newest video coding standard--high efficiency
video coding (HEVC) can achieve higher coding efficiency, but at
the expense of tremendous computational complexity. To address
this problem, a fast inter mode decision algorithm, called the
hierarchical inter mode decision (HIMD), is proposed in this
paper. For each coding tree unit (CTU), the proposed HIMD
algorithm determines the optimal mode in time by checking the
early termination condition, which is developed based on the
observation that the relationship between the rate-distortion (RD)
cost and the size of prediction mode is a uni-modal function (only
one minimum value). Experimental results have shown that
compared with the exhaustive mode decision in HEVC, the
proposed HIMD algorithm is able to achieve a reduction of
computational complexity by 45.81% on average, while incurring
only 0.07dB loss in Bjontegaard delta peak signal-to-noise ratio
(BDPSNR) and 2.04% increment on the Bjontegaard delta bit rate
(BDBR).
Keywords—HEVC; coding unit; mode decision; early
termination
I. INTRODUCTION
High efficiency video coding (HEVC) is the newest video
coding standard [1-2], which is a successor of the H.264/AVC
[3] and developed by joint collaborative team on video coding
(JCT-VC) standard committee formed by video coding experts
group (VCEG) and motion picture experts group (MPEG). The
core objective of HEVC is to double the compression
efficiency while keeping almost the same subjective quality of
the reconstructed video, compared with H.264/AVC.
Similar to H.264/AVC, HEVC also adopts a block-based
hybrid coding framework. The improvement of coding
efficiency is achieved by exploiting many new techniques,
such as coding unit (CU) with variable block sizes varying
from 4×4 to 64×64, prediction unit (PU), transform unit (TU),
quad-tree structure, advanced motion vector prediction, and so
on [4]. However, the high coding efficiency of HEVC is
accomplished at the price of tremendous computational
complexity, which becomes the bottleneck of the real-time
applications of HEVC. In HEVC, the inter prediction occupies
more than 70% computational complexity. Therefore, how to
reduce the computational complexity of inter prediction in
HEVC while maintaining almost the same coding efficiency
has become an emerging research topic.
(a)
(b)
Fig. 1. (a) CU partitions and (b) PU partitions used in HEVC
In recent years, studies on this research topic can be found
in the literature [5-10]. Leng et al. [5] proposed a fast CU
decision algorithm at either frame level or CU level to
accelerate the encoding process. Cassa et al. [6] proposed a
method to reduce computational complexity by top skipping
and early termination. Kim et al. [7] suggested an early
detection of SKIP mode based on the evaluation of the
differential motion vector (MV) and the coded block flag
(CBF). Xiong et al. [8] presented a fast pyramid motion
divergence, which is then used to select those more likely CU
sizes. Shen et al. [9] presented a fast CU size decision method,
including the depth range selection and early termination by
making use of the neighbor and co-located CU information.
Lee et al. [10] proposed a fast encoding method to predict the
coding tree structure of current CU by using the same structure
of co-located CU in reference or temporally-previous frames.
In this paper, a more efficient inter mode decision
algorithm, called the hierarchical inter mode decision (HIMD),
is proposed. Based on the observation that there exists a
unimodal relationship (only one minimum value) between the
rate-distortion (RD) cost and the size of prediction mode, only
This work was supported in part
y the National Natural Science
Foundation of China under the Grants 61372107 and 61401167, in part by the
Xiamen Key Science and Technology Project Foundation under the Gran
3502Z20133024, in part by the Opening Project of State Key Laboratory o
Digital Publishing Technology under the grant FZDP2015-B-001, and in par
by the High-Level Talent Project Foundation of Huaqiao University under the
Grants 14BS201 and 14BS204.
2015 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) November 9-12, 2015
978-1-4673-6499-7/15/$31.00 ©2015 IEEE