Manuscript received May 30, 2015; revised December 7, 2015.
This work was supported by the National Natural Science
Foundation of China under Grant No. 61201371, the promotive research
fund for excellent young and middle-aged scientists of Shandong
Province, China under Grant No. BS2013DX022, and the Natural
Science Foundation of Shandong Province, China under Grant No.
ZR2015PF004.
Corresponding author email: wangchengyou@sdu.edu.cn.
doi:10.12720/jcm.10.12.1004-1011
of Communications Vol. 10, No. 12, December 2015
©2015 Journal of Communications
1004
Video Coding Based on Shape-Adaptive All Phase
Biorthogonal Transform and MPEG-4
Xiaoyan Wang
1
, Chengyou Wang
1
, Xiao Zhou
1
, and Zhiqiang Yang
1,2
1
School of Mechanical, Electrical and Information Engineering, Shandong University, Weihai 264209, China
2
Integrated Electronic Systems Lab Co. Ltd., Jinan 250100, China
Email: swwxy00800313@163.com; {wangchengyou, zhouxiao}@sdu.edu.cn; yangzhiqiang@ieslab.cn
Abstract—This paper proposes an efficient video coding
algorithm based on shape-adaptive all phase biorthogonal
transform (SA-APBT) and MPEG-4. Firstly, the input video
sequence is segmented into many arbitrarily shaped video
objects. Then the motion, shape and texture of all video objects
are encoded in accordance with priority. Shape coding uses
context-based arithmetic encoding (CAE). SA-APBT and
uniform quantization method are adopted in texture coding.
Experimental results show that reconstructed video sequence
using video coding algorithm based on SA-APBT and MPEG-4
obtains better effects including objective quality and subjective
quality compared with the one based on SA-DCT. And the
proposed algorithm uses uniform quantization instead of
complex quantization table in MPEG-4, which makes hardware
implementation easier.
Index Terms—Video coding, MPEG-4, shape-adaptive all
phase biorthogonal transform (SA-APBT), video object
I. INTRODUCTION
In video coding, the block-based coding has been
widely used in many video compression standards,
including MPEG-2 [1], H.263 [2] and H.264 [3]. The
disadvantage of the block-based coding is bad subjective
effects of reconstructed video sequence at low bit rates
[4]. Now in many applications such as video conference,
telemedicine and remote monitoring, observers tend to
focus on some objects in the video sequence, whereas
they are not interested in other objects, which promote
the development of the object-based video coding.
Because observed object can also get good reconstructed
quality using object-based coding even at low bit rates [5].
In MPEG-4 [6], video sequence is comprised of many
video objects with different physical significance. Each
video object can be accessed and manipulated separately
[7].
Shape-Adaptive Discrete Cosine Transform (SA-DCT)
[8] for arbitrarily shaped image segment coding is widely
adopted and included in MPEG-4 standard because of
low complexity and effective coefficients decorrelation.
But the shortcoming of it is the serious blocking effects at
low bit rates [9]. In order to solve this problem, this paper
proposes video coding algorithm based on Shape-
Adaptive All Phase Biorthogonal Transform (SA-APBT)
and MPEG-4. Firstly, the video sequence is segmented
into many arbitrarily shaped video objects. Then the
motion, shape and texture of all video objects are
encoded in accordance with priority. In texture coding,
SA-APBT and uniform quantization method are used.
Experimental results show that the objective quality and
subjective effects of the reconstructed video sequence
using the proposed algorithm based on SA-APBT are
better than that based on SA-DCT at same bit rates.
The rest of this paper is organized as follows. Section
II introduces two shape-adaptive transforms: SA-DCT
and SA-APBT. MPEG quantization and H.263
quantization methods used in MPEG-4 are explained in
Section III. And then in Section IV, MPEG-4 video
coding algorithm using SA-APBT is proposed. The
overall algorithm description of MPEG-4 video coding
using SA-APBT is firstly given. Thereafter the shape
coding and texture coding for all video objects are
described in detail. Experimental results and comparisons
between the proposed algorithm based on SA-APBT and
the one based on SA-DCT are presented in Section V.
Conclusions of the paper and discussion for future work
are given finally in Section VI.
II. SHAPE-ADAPTIVE TRANSFORM
A. SA-DCT
Conventional two-dimensional Discrete Cosine
Transform (DCT) [10] can only be used in rectangular
image coding and is not applicable to arbitrarily shaped
image segment coding. Let
X
and
C
represent an image
block and DCT matrix with size of
NN
respectively.
After two-dimensional DCT transform, transform
coefficients block
Y
can be denoted by
T
Y = CXC
(1)
1
, 0, 0,1, , 1,
( , )
2 (2 1)
cos , 1,2, , 1, 0,1, , 1,
2
i j N
N
ij
ij
i N j N
NN
C
(2)