An Optimized Probability
Estimation Model for Binary Arithmetic Coding
Jing Cui
1
, Shanshe Wang
2
, Nan Zhang
3
and Siwei Ma
4
1
School of Electrical & Computer Engineering
Seoul National University, Seoul, Korea
3
School of Biomedical Engineering, Capital Medical University, Beijing, China
2,4
School of Electronic Engineering & Computer Science
Peking University, Beijing, and
4
Peking University Shenzhen Graduate School, Shenzhen, China
1
Jingcui106@gmail.com
2
sswang@jdl.ac.cn
3
zhangnan@ccmu.edu.cn
4
swma@pku.edu.cn
Abstract— In this paper, we analyze the binary arithmetic
coding of High Efficiency Video Coding (HEVC) and the second
generation of audio and video coding standard (AVS2). Then an
optimized probability estimation scheme is proposed for
arithmetic coder. The proposed scheme is incorporated into the
HEVC reference software (HM 16.0) and AVS2 reference
software (RD 10.1). Experimental results demonstrate that the
proposed scheme can efficiently improve the coding efficiency of
entropy coding. The rate-distortion (R-D) performance gain can
be up to 0.21% for AVS2 and 0.30% for HEVC respectively.
Index Terms
— HEVC, AVS2, Context-based Binary
Arithmetic Coding (CBAC), Context-based Adaptive Binary
Arithmetic Coding (CABAC), probability estimation
I. INTRODUCTION
High efficiency video coding (HEVC) [1] is the latest video
coding standard developed by JCT-VC (Joint Collaborative
Team on Video Coding). Meanwhile, in China, AVS group
developed the second generation of audio and video coding
standard (AVS2) [2]. Compared to the previous video coding
standards, such as H.264/AVC [3] and AVS1 [4], coding
performance of HEVC and AVS2 has been significantly
improved.
To improve the coding performance, both HEVC and
AVS2 adopted binary arithmetic coding. What is different is
that HEVC adopted Context-based Adaptive Binary
Arithmetic Coding (CABAC) [5] and AVS2 adopted Context-
based Binary Arithmetic Coding (CBAC) [6]. These two
schemes are both based on information entropy theory [7] and
bring much coding performance improvement.
For the Binary Arithmetic Coding, the main procedure
mainly includes binarization, context modelling, binary
arithmetic coding and parameters updating. The binarization
process in current video standard employs the similar model-
probability distribution [8]. As for the context modelling, the
selection of probability model is referred as the context
modelling and the probability estimation is introduced to
estimate the probability of current symbol valued from 0 to
1[9]. In addition, the probability estimation model also affects
the probability update and range subdivision in the binary
arithmetic coding engine where the recursive interval
subdivision, parameters derivation, and relative updating are
performed based on the principle of arithmetic coder.
Therefore more compelling probability updating model is
needed especially for the motive signal source. Many research
works focus on how to optimize the binary arithmetic coding.
In [10], the proposed “virtual sliding window” method
provided a more outstanding compression rate compared with
look-up table index based entropy coder in [9]. Currently, the
virtual sliding window technique is widely explored in HEVC.
An integrated window sizes technique is introduced in [11] ~
[13], which gives a higher precision estimation model with
around 0.8% performance improvement in HEVC. In [14], a
counter-based window sizes scheme is proposed and brings
about 0.9% BD-rate saving. Therefore for probability
estimation, the smaller window size of each probability model
in the beginning of the sequence can improve the R-D
performance considerably and the changeable window size
tends to be more effective. The entropy coder CBAC used in
AVS2 made the similar affords to design an adaptive
probability estimation model to improve R-D performance,
although it causes computation complexity increase. In this
paper, we proposed a more effective scheme for probability
estimation in CBAC and we also incorporate our proposed
scheme as the entropy coder into HEVC to improve the
CABAC.
The rest of the paper is arranged as follows. In Section II,
probability estimation models in CABAC and CBAC are
shown briefly. Section III proposed our optimization scheme.
The experimental results both on AVS2 and HEVC are
presented in Section IV. Finally, Section V concludes paper.
II. P
ROBABILITY ESTIMATION MODELS IN CABAC AND
CBAC
Probability estimation plays a vital role in Binary
Arithmetic Coding. Probability estimation model is introduced
by adapting the internal state of the coder to the underlying
source statistics. Such feature enhances the compression
efficiency and it has been adopted into various entropy coding
schemes such as M coder, PIPE. However, HEVC and AVS2
adopted different probability estimation models to support
their respective arithmetic coding engine. In this section, we
will briefly introduce nature of probability estimation. Then
978-1-4673-7314-2/15/$31.00 ©2015 IEEE IEEE VCIP 2015