Performance Enhancement for Quality Inter-Layer Scalable Video Coding

Performance Enhancement for quality inter- layer Scalable Video
Coding
Abstract: One of advanced video coding
standard is scalable video coding (SVC)
extension of H.264/AVC, SVC provides
multimedia service within variable
transport environments. SVC is a large
computation complexity in encoding
processing, through using the exhaustive
search technique. The encoding processing
included the select macroblock mode and
motion vector layer. This paper introduces a
new proposed algorithm to reduce this
complexity with saving quality. Scalable
quality inter-layer performance enhancement
(SQILPE) proposed algorithm depends on
analysis amount of change in intensity value
of pixels in the MB and video statistics.
Experimental results show that the proposed
fast mode decision algorithm can achieve
computational savings up to 77.6% with
almost no loss in quality.
Keywords: moition vector, moition
estmation, Quality SVC, SNR SVC,
macroblocks.
1. INTRODUCTION
Many heterogenous demands required this
day for a video sequence. SVC is one of
recent video coding standards based on
H.264/AVC [1] [2], Which provided more
resilience in dealing with multiy different
requierments of video sequance ensursnce
efficient video coding Performance. Through
high compression ratio and saving the
quality[3].
The SVC/H.264 take considerable time in
coding processing for computation the best
mode of macroblock and motion estimation
through using the exhaustive search
techniques [4]. There are eight macroblock
models for inter prediction (SKIP,
MODE_16x16, MODE_16x8, MODE_8x16,
MODE_8x8, MODE_8x4, MODE_4x8, and
MODE_4x4) [5] [6] , and two intra modes
(INTRA_4x4 and INTRA_16x16 [7] [8].
For the monition vector estimation, there are
three-layers full pixel, subpixel (half pixel
and quarter pixel) [9] [10].
The reasons for using subpixel come from
difference reasons; Sampling: converting,
Lighting condition, Noise, Camera shaking:
camera shaking can bring motion effect to
stationary objects[11] [12].
Mayada Khairy
Computers and Systems Dept.
Electronics Research Institute, ERI
Giza, Egypt
Alaa Hamdy,
Electronics, Communications &
Computers Dept.
Helwan University
Cairo, Egypt
Amr Elsayed
Electronics, Communications & Computers
Dept.
Helwan University
Cairo, Egypt
Hesham Farouk
Computers and Systems Dept.
Electronics Research Institute, ERI
Cairo, Egypt
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 8, August 2017
240 https://sites.google.com/site/ijcsis/
ISSN 1947-5500

The Sub-pixel refinement can greatly
improve the performance of ME regarding
both compression ratio and decoded image
quality[13] [14].
Many fast modes decision schemes have
been proposed to solve the problem that
resulted from using exhaustive search
technique. Tae-Jung Kim and et al in [15]
introduced fast mode decision algorithm
based on coded block pattern (CBP) of 16 ×
16 mode in the current frame and a reference
block of best CBP. Seon-Tae Kim and et al.
in [16] suggested a method, depends on using
information of collocated macroblock and a
neighboring block in the base layer to encode
enhancement layer.He Li, Z. G. Li and et al.
in [17] developed an algorithm based on the
relation between a base layer (BL) and
enhancement layer (EL) number of
candidate mode.
This paper introduced proposed an algorithm
to enhancement the problem of encoding
complexity in two phases. First one for the
select the MB modes and the second phase
for select motion layer depends on many
characteristics as video statistics [18] [19],
change of pixel value intensity [20] [21],
object belong to background or foreground.
All details will be discussed in the following
sections.
The rest of the paper is organized as follows:
Section 2 gives the description of the
proposed algorithm. The proposed Algorithm
experiments and results; are introduced in
section 3. Finally, section 4 represents the
conclusion.
2. Scalable Quality Interlayer
Performance Enhancement(SQILPE)
Scalable Quality Inter-layer Performance
Enhancement (SQILPE) proposed algorithm
Outlined in fig.1. This algorithm depandes on
many charatrisctics; analyisis the change in
the pixels intistiy vlaue, stady the video
statstics, and extract back ground objects and
for ground objects, relation betwwen the size
of blocks and the quaity of the encoded MB.
Links these flages with others to achive our
targets from this proposed algorithm
Figure 1: Scalable Quality Inter-Layer Performance Enhancement proposed algorithm
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

The proposed algorithm work on two phases
as follows :
First phase: select the best macrblocks
depandes on the the change in the intinisty
pixels value for MB and the vido statistics.
And tradeoff between blocks size and mode
where blocks are bigear as MB16*16 This
achive good and fast encoding processing at
the expense of quality on the other hand side
the smaller block size saving the quality with
incrasing unmber of bits and incrasing the
encoding time[22]. SQILPE proposed
algorithm illustrated in fig.2 start by divid
frame the block as MB 16*16 and analysis
intinisty value in the MB using varianc [23]
[24].using the following equation :
=
∑( )
……………… (1)
µ=
∑
….………………... (2)
Where an µ represent variance, and
mean respectively.
∗
<TH1; Which means the low
variance; in this case the MB 16*16 is the
best choice .now the proposed algorithm
select the mode (inter or intra )of the selected
size of MB depends on analysis the video
statistics using zero mean normalized Crosse
correlation [25] [26] according the following
equation:
Figure 2: select MB mode in SQILPE
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

ZNCC(F , F ) =
∑
∑ ×
∑ ×∑
………. (3)
F1 = x , y ,
F2 = (x , y )
• If ZNCC<THS: Which mean the
low video statistics; Get inter mode of
MB16*16 Whereas the propopilaity
to find the same MB in search
window is vary high as knowing the
inter mode using lower bit s than intra
mode
• Else if Tradeoff between inter
MB16*16 and intra MB16*16 using
(Sum absolute transformer
difference) SATD [10].
Else if TH1< ∗ <TH2: Which
means the medium variance and select MB
size and modes depends on video statistics as
following:
• If ZNCC<THS Tradeoff
between inter MB16*8 and , inter
MB 8*16
• Else if ZNCC>THS Tradeoff
between: inter MB 16*8, inter
MB 8*16 and intra MB16*16
Else Excludes the MB 16*16 size for inter
and intra mode and test MB 8*8 as following:
∗ < TH1 the MB
• ZNCC<THS Get inter MB
8*8
• ZNCC>THS Tradeoff
between: inter MB 8*8 and intra
MB 4*4
Is ∗ > TH1?
• ZNCC<THS Get intra MB
4*4
• ZNCC>THS Tradeoff
between: inter MB 4*4 and intra
MB 4*4
The second phase in SQFM: utilized to
select the best layer of MV with low
complexity. As mention before there are
many reasons for using the different layer in
selecting the MV -full pixel and sub motion
pixel (quarter and half). The way of selecting
the MV layer in this proposed algorithm
depends on the frame of video contains the
different objects with different statictics from
this point the proposed algorithm will start.
Classified the objectives on the frame to
background and foreground object[27] -here
the background object represents the object
with nearly fixed. As illustrated in fig.3 Star
by calculating the video statistics using the
ZNCC equation which already calculated
from phase one of proposed algorithm (1):
If ZCNN<TH1 Which mean the low video
statistics; here the applied of full MV and
disapply subpixel MV, and the mv calculate
as the following equation:
, , , = ∑ ∑ ( , )……..… (4)
B=SATD =HF ………………………….…… (5)
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

( ) , = (−1)∑
............................ (6)
Figure 3: select MV layer
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

If TH1<ZCNN< which mean the
medium video statistics, here sparated the
objects in frame to background and for
ground objects using Median approximation
median filter[28] as equations (7),(8) in case
of the target MB belong to the background
applied of full MV .
MB = ∑ ∑ |I(i, j, t) − Med(i, j, t −
1)| <Th................................……….…. (7)
Median = {(n + 1) ÷ 2}
the
value…….… (8)
If MB belongs to the foreground which
meaning:
MB = ∑ ∑ |I(i, j, t) − Med(i, j, t −
1)| > Th……………….………………. (9)
In this case applied full motion vector and
half pixel MV and bilinear interpolation to
interpolate macroblocks for sub motion
vector.
ZCNN>TH2 Which mean video statistics is
high.
If the MB belongs to the foreground object.
Test: MV1=full motion vector and sub
motion vector (half and quarter) with
utilising the bicubic interpolation to
interpolate macroblock. Else if MB belongs
to background objects using full MV.
3. Experiments and Results
The results of the proposed algorithm -
experimental results- where be compared to
JSVM 9.17 reference software, Seon-2
algorithm [29]All the results were be referred
to reference algorithms as a percentage value.
This work evaluates the proposed algorithms
on three standard test sequences. The test
sequences are "Forman" as low motion,
“Bus" as medium motion, and "Football" as
fast motion. Four parameters are used to
evaluating the performance. These
parameters are; saving in several sending bits
(ΔNSB), saving in ET (ΔET), saving in TT
across the network (ΔTT), and degradation in
peak signal to noise ratio (Y-PSNR). The
equations of the three parameters are shown
in equations [10-12].
The results show in table 1
ΔNSB=
( ) ( )
( )
∗ 100.... (10)
ΔET=
( ) ( )
( )
∗ 100……... (11)
Y-PSNR=PSNR(proposed) − PSNR(JSVC)
……………………………………….... (12)
First: “Foreman” test video:
By applying the SQILPE algorithm; in this
case, the highest significant reduction in
encoding time can be obtained up to 77.58%,
with negligible decreasing in quality 0.002
dB and increasing in Number of sending bits
up to -0.13% compared with JSVM 9.17.
Comparing “Foreman” test video with Seon-
2, the results become 12.29%, -0.126%,
1.025 dB for, Δ ET, Δ NSB and the increased
value in Y-PSNR respectively.
Second: “Bus” video sequence:
SQILPE algorithm; achieve 74.68%
reduction in encoding time, 0.54% increasing
in numbers of sending bits with negligible
decreasing in quality with 0.002dB compared
with JSVM 9.17.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

Table 1: Results of SQILPE proposed algorithm
Comparing “Bus” test video with Seon-2, the
results become 11.75%, -0.486%, 1.008 dB
for, Δ ET, ΔNSB and increase in Y-PSNR
respectively.
Third: “Football” video sequence:
SQILPE algorithm; achieve 70.54%
reduction in encoding time, a negligible
number of sending bits 0.98 with decreasing
in quality 0.006dB compared with JSVM
9.17.
Comparing “Football” test video with Seon-
2, the results become 21.99%, -0.92%, -1.006
dB for, ΔET, ΔNSB and Y-PSNR
respectively.
4. CONCLUSION
This new proposed algorithm paper
introduced for quality Inter-layer residual
SVC. SQILPE proposed algorithm achieve
significant saving in encoding time with
small decreasing in quality and small
increasing in number of sending bits
comparing with JSVM 9.17. moreover,
comparing with other proposed algorithms as
Seon-2 the SQFM algorithm improving the
quality and the encoding time with negligible
increasing with NSB.
Acknowledgment
The paper in the core technical gives
opportunity to express my profound gratitude
and deep regard for Eng. Gamal Saleh
Chairman of Systems Design Company, for
his great support and honest consultation
specialisation for valuable feedback and
constant encouragement.
References
[1] S.-T. Kim, K. R. Konda, and C.-S. Cho,
“Fast mode decision algorithm for
spatial and SNR scalable video coding,”
in 2009 IEEE International Symposium
Video
Sequence
Test
Measurement
Parameter
SQILPE
SQILPE Vs.
JSVM 9.17
SQILPE Vs.
Seon-2
Foreman
ΔNSB % 0.13- 0.126-
ΔET% 77.58 12.29
Y-PSNR dB -0.002 1.025
Bus
Δ NSB % -0.54 0.486-
ΔET% 74. 68 11.75
Y-PSNR dB -0.004 1.008
Football
ΔNSB % 0.98 0.92-
ΔET% 70.54 21.31
Y-PSNR dB -0.006 1.013
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

on Circuits and Systems, 2009, pp. 872–
875.
[2] T. Wiegand, L. Noblet, and F. Rovati,
“Scalable video coding for IPTV
services,” IEEE Trans. Broadcast., vol.
55, no. 2, pp. 527–538, 2009.
[3] H. Kibeya, F. Belghith, M. A. B. Ayed,
and N. Masmoudi, “Fast coding unit
selection and motion estimation
algorithm based on early detection of
zero block quantified transform
coefficients for high-efficiency video
coding standard,” IET Image Process.,
vol. 10, no. 5, pp. 371–380, 2016.
[4] J. Bankoski, P. Wilkins, and Y. Xu,
“Technical overview of VP8, an open
source video codec for the web.,” in
ICME, 2011, pp. 1–6.
[5] N. Bahri, I. Werda, A. Samet, M. A. B.
Ayed, and N. Masmoudi, “Fast intra
mode decision algorithm for H264/AVC
HD baseline profile encoder,” Int. J.
Comput. Appl., vol. 37, no. 6, pp. 8–13,
2012.
[6] M. Khairy, A. Elsayed, and A. Hamdy,
“Low Complexity for Scalable Video
Coding Extension of H. 264 based on
the Complexity of Video,” Int. J. Adv.
Comput. Sci. Appl., vol. 7, no. 12, pp.
220–225, 2016.
[7] T.-D. Chuang, Y.-H. Chen, C.-H. Tsai,
Y.-J. Chen, and L.-G. Chen, “Algorithm
and architecture design for intra
prediction in H. 264/AVC high profile,”
in Proc. Picture Coding Symp, 2007.
[8] M. Khairy, A. Hamdy, A. Elsayed, and
H. Farouk, “Efficient Enhancement for
Spatial Scalable Video Coding
Transmission,” Math. Probl. Eng., vol.
2017, 2017.
[9] B. Lee, S.-C. Cha, H.-J. Ha, and W.
Han, Method and apparatus for
supporting motion scalability. Google
Patents, 2005.
[10] Q.-F. Zhu, Method and system for
mode decision in a video encoder.
Google Patents, 2005.
[11] A. Abdelazim, “Fast Motion
Estimation Algorithms for Block-Based
Video Coding Encoders,” University of
Central Lancashire, 2011.
[12] N. Stupich, “Low Power Parallel
Rolling Shutter Artifact Removal,”
Carleton University Ottawa, 2014.
[13] W. Lin, K. Panusopone, D. M.
Baylon, M.-T. Sun, Z. Chen, and H. Li,
“A fast sub-pixel motion estimation
algorithm for H. 264/AVC video
coding,” IEEE Trans. Circuits Syst.
Video Technol., vol. 21, no. 2, pp. 237–
242, 2011.
[14] A. Abdelazim, M. Yang, and C.
Grecos, “Fast subpixel motion
estimation based on the interpolation
effect on different block sizes for
H264/AVC,” Opt. Eng., vol. 48, no. 3,
pp. 030504–030504, 2009.
[15] E. Maani and A. K. Katsaggelos,
“Optimized bit extraction using
distortion modeling in the scalable
extension of H. 264/AVC,” IEEE Trans.
Image Process., vol. 18, no. 9, pp.
2022–2029, 2009.
[16] S.-T. Kim, K. R. Konda, C.-S. Park,
C.-S. Cho, and S.-J. Ko, “Fast mode
decision algorithm for inter-layer coding
in scalable video coding,” IEEE Trans.
Consum. Electron., vol. 55, no. 3, pp.
1572–1580, 2009.
[17] H. Li, Z. G. Li, and C. Wen, “Fast
mode decision algorithm for inter-frame
coding in fully scalable video coding,”
IEEE Trans. Circuits Syst. Video
Technol., vol. 16, no. 7, pp. 889–895,
2006.
[18] J. Ascenso, C. Brites, and F. Pereira,
“Content adaptive Wyner-Ziv video
coding driven by motion activity,” in
Image Processing, 2006 IEEE
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

International Conference on, 2006, pp.
605–608.
[19] H. Gong and S.-C. Zhu,
“Intrackability: Characterizing video
statistics and pursuing video
representations,” Int. J. Comput. Vis.,
vol. 97, no. 3, pp. 255–275, 2012.
[20] R. J. Radke, S. Andra, O. Al-Kofahi,
and B. Roysam, “Image change
detection algorithms: a systematic
survey,” IEEE Trans. Image Process.,
vol. 14, no. 3, pp. 294–307, 2005.
[21] L. Li and M. K. Leung, “Integrating
intensity and texture differences for
robust change detection,” IEEE Trans.
Image Process., vol. 11, no. 2, pp. 105–
112, 2002.
[22] M. P. Sharabayko and N. G. Markov,
“H. 264/AVC Video Compression on
Smartphones,” in Journal of Physics:
Conference Series, 2017, vol. 803, p.
012141.
[23] G. K. Rohde, A. S. Barnett, P. J.
Basser, and C. Pierpaoli, “Estimating
intensity variance due to noise in
registered images: applications to
diffusion tensor MRI,” Neuroimage, vol.
26, no. 3, pp. 673–684, 2005.
[24] J. Sijbers, D. Poot, A. J. den Dekker,
and W. Pintjens, “Automatic estimation
of the noise variance from the histogram
of a magnetic resonance image,” Phys.
Med. Biol., vol. 52, no. 5, p. 1335, 2007.
[25] V. Thirumalai and P. Frossard, “Joint
reconstruction of multiview compressed
images,” IEEE Trans. Image Process.,
vol. 22, no. 5, pp. 1969–1981, 2013.
[26] K. Briechle and U. D. Hanebeck,
“Template matching using fast
normalized cross correlation,” in
Aerospace/Defense Sensing, Simulation,
and Controls, 2001, pp. 95–102.
[27] M. Heikkila and M. Pietikainen, “A
texture-based method for modeling the
background and detecting moving
objects,” IEEE Trans. Pattern Anal.
Mach. Intell., vol. 28, no. 4, pp. 657–
662, 2006.
[28] B. Weiss, “Fast median and bilateral
filtering,” Acm Trans. Graph. TOG, vol.
25, no. 3, pp. 519–526, 2006.
[29] S.-T. Kim, K. R. Konda, C.-S. Park,
C.-S. Cho, and S.-J. Ko, “Fast mode
decision algorithm for inter-layer coding
in scalable video coding,” IEEE Trans.
Consum. Electron., vol. 55, no. 3, pp.
1572–1580, 2009.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

Performance Enhancement for Quality Inter-Layer Scalable Video Coding

More Related Content

What's hot (18)

Similar to Performance Enhancement for Quality Inter-Layer Scalable Video Coding (20)

Recently uploaded (20)

Performance Enhancement for Quality Inter-Layer Scalable Video Coding