Rate Control Methods Evaluation and Analysis for

H.263 and MPEG-4 Video Codec

Imran Ullah Khan[1], M.J.R.Khan [2], M.A.Ansari[3]
[1]
Research Scholar, Dept. Electronics & Comm. Eng., Mewar University, Chittorgarh, India
[2]
Faculty of Engineering, Mewar University, Chittorgarh,
[3]
India Faculty of Engineering,GBU Greater Noida,India
Email: [email protected] [1], [email protected][2] [email protected][3]

Abstract - Rate control plays an important role in video coding, control (SRC) is designed to meet both constraints (low-
although its not a normative tool for any video coding standard. latency and buffer constraints). A variable frame rate approach
In video communications, rate control must ensure the coded can be used to achieve the target bit rate. The similar
bitstream can be transmitted successfully and make full use of
technique can also extend at slice or macroblock layers [5-7].
the limited bandwidth. As a consequence, a proper rate control
The two rate control methods for MPEG-4 that are used in this
scheme is usually recommended by a standard during the
development, e.g. test model number 5 (TMN 5), TMN 8 and
work are as follows:
TMN 12 for H.263, and Verification Model 8 (VM8) for MPEG- OFFLINE: In this option, the bit rate control takes the form of
4, etc. Some work on rate control has been done for H.264/AVC changing the quantization levels over the frames encoded after
too. In the contribution, a rate control scheme based on VM8 is the first I- or P-frame. There will be no frame skipping in this
adopted by H.264/AVC test model. The performance evaluation mode.
suggested in this work will enable the user analyzing ONLINE: In this mode, the bit rate control is accomplished by
experimental results of different rate control methods and how adaptively changing the quantization values at the macroblock
optimum rate-distortion performance for H.263 and MPEG-4
level. Frame skipping is allowed in this mode.
video codec achieve.

Keywords Bit rate, coefficient of Correlation, Peak Signal to II. FUNCTION OF RATE CONTROL
Nose Ratio, Quantization Parameter and Rate control.
Rate control is that the encoder estimates the video bitrate
I. INTRODUCTION based on the network available bandwidth, ensures the coded
bitstream can be transmitted successfully and makes full use
The bit rate resulting from the DCT-based coding algorithm of the limited bandwidth. In other words, it is adjusting video
fluctuates according to the nature of the video sequence. output bits according to the channel is fixed or variable
Variations in the speed of moving objects, their size and transmission rate [8-11].
texture are the main cause for bit rate variation. The objective Now the core part of many video coding standards is the
of a rate controller is not only to achieve a constant bit rate for motion compensation and the DCT transform coding based on
transmission over a circuit switched network but also to block. The number of the encoder output bits of each frame is
maintain good quality of the reconstructed videos. A changing with the active input image. Therefore, the bitstream
transmission buffer is usually needed to smooth out the bit has the inherent characteristics of changing. If the coding
rate fluctuations, which are inherent in the interframe coding parameters remain unchanged in the compression process, the
scheme [1-4].The H.263 codec used in this work simulates bits of the consumption of different frame will be significantly
three rate control methods different. Due to the actual network bandwidth and storage
C1: Off line medium, if we have nothing to do with the bitstream, the
C2: Test model no.5 (TMN-5) video communication system is likely to go abnormally.
C3: Test model no.8 (TMN-8) Generally, using a buffer makes the output bitstream smooth.
Bit rate control and buffer regulation is an important issue for The buffer capacity has certain limitation (If buffer is too big,
constant bit rate (CBR) application. The scalable bit rate the propagation delay of real-time communication is longer

1
 Input Frame
which is difficult to be accepted). In order to prevent buffer
+ DCT Quantizer VLC
Output
overflow and underflow [5], rate control must be used in -
encoder.
III. OVERVIEW OF RATE CONTROL Inverse
Quantizer
Because of transmission bandwidth and storage space
limitation, video applications for higher compression ratio,
nondestructive coding can provide the compression ratio but IDCT
cannot satisfy the demand of actual video applications, but if
we can accept some degree of distortion, high compression + +
ME/MC
ratio is easy to get. Human visual system for high frequency
signals change not sensitive information loss, high frequency
Memory
part does not reduce subjective visual quality. Video coding
algorithm of mainstream DCT quantization method is adopted Fig.1: Block diagram of H.263 coder
to eliminate video signals, the visual physiology redundant Where, DCT - Discrete Cosine Transform VLC Variable Length Coding
than lossless higher compression ratio and will not bring the ME/MC- Motion Estimation/Motion Compensation

video quality decrease significantly [11, 12].

When using a lossy coding method, it is related to the V. MPEG-4 AND ITS PARAMETERS
difference between the reconstruction images g (x, y) and the
original image f (x, y). Generally, the distortion factor D MPEG-4 have many advanced features, which were not
function can form according to need, such as selecting any available in its predecessors such as MPEG-1/2. MPEG-4
cost function, absolute square cost function etc. In the image combines some typical features of other MPEG standards, but
coding D is computed as[17]: aims to provide a set of technologies to satisfy the needs
authors, service providers and end users. It enables much
D E{[f(x, y) g(x, y)]2 .(1) functionality potential accessible on a single terminal and
higher levels of interaction with content. MPEG-4 achieves
these goals by providing standardized ways to support:
IV. H.263 AND RATE CONTROL METHODS coding, composition, multiplexing and interaction [19-22].
The key feature of MPEG-4 is decompose a video frames in
H.263 video coder has many advanced features, which were different layers of video object plane (VOP). The block
not available in its predecessor H.261. The block diagram of diagram of a MPEG-4 video coder is shown in Fig. 2.
H.263 video codec is shown in figure1. Image
Varying different parameters may control the features and Sequence Motion
+ Quantizer Video Output
functionalities of H.263, but for this work only those basic DCT Texture
Multiplex
parameters that provide trade-off between video quality, Coding
-
transmission bandwidth and computational complexity are
INVERSE
considered. QUANTIZE
R
Rate control method
The bit rate resulting from the DCT-based coding algorithm
IDCT
fluctuates according to the nature of the video sequence.
Variations in the speed of moving objects, their size and + +
texture are the main cause for bit rate variation. The objective S
of a rate controller is to achieve a constant bit rate for W Pred. 1
transmission over a circuit switched network. A transmission I Frame Store
T Pred. 2
buffer is usually needed to smooth out the bit rate fluctuations, C
which are inherent in the interframe coding scheme [13-18]. H Pred. 3
Fig.2 Block diagram of MPEG-4 coder
Where,
ME/MC DCT - Discrete Cosine Transform
IDCT - Inverse Discrete Cosine Transform
ME/MC- Motion Estimation/Motion Compensation
Shape Coding Pred. - Prediction
 VI. RATE CONTROL SCHEME IN MPEG-4 (i) The computation of target bit rate is based on the bits
available and the last encoded frame bits. However, there
The MPEG group officially initiated an MPEG-4 are fewer number of bits left for encoding thus, these bits
standardization phase with mandate to standardize algorithms can be assigned to this frame. A weighed average reflects
for audio-visual coding in multimedia applications, allowing a compromise of these two factors.
for interactivity, high compression, universal accessibility and (ii) A lower bound of target bit rate (F/30) is used so that the
portability of audio and video contents. Target bitrate for the minimal quality is guaranteed (where F denotes total
video standard is between 564 k bits/s for mobile target bits per second) [23-25].
applications and up to 4 M bits/s for TV/lm applications. (iii) The target bit rate is adjusted according to the buffer
MPEG-4 adopted VM8 (Verification Model 8) to realize rate status to prevent both overflow and underflow.
control. 3. Computation of the quantization parameter (Q) before
encoding
Initialize
(i) Q is solved based on the model parameters
(ii) Q is clipped between 1 and 31.
(iii) Q varies within 25% of the previous Q to maintain a
variable bit rate (VBR) quality [26-29].
Target Bit Calculation (T) for current frame
4. Encoding current frame
5. After encoding, model parameters are updated based on
the encoding results of the current frame.
(i) The data points are selected by using a window.
Estimation of Quantization Step
Q = R-1 (T, MAD)
(ii) Whose size depends on the change in complexity. If the
from T = R (Q, MAD) complexity changes significantly, a smaller window with
more recent data points is used.
(iii) The model is again calibrated by rejecting the outlier data
points. The rejection criterion is the data point and is
Encoding Current Frame discarded when the prediction error is more than one
standard deviation.
(iv) The next frame is skipped if the current buffer status is
above 80% [28, 29].

Update Rate Distortion Model VII. SIMULATION , IMPLEMENTATION

T = R (Q, MAD) DETAILS AND RESULTS OBTAINED
First Estimator T = R (Q, MAD)
Remove Outlier
Second Estimator T = R (Q, MAD)
The software for video codec used in this work was developed
by university of British Columbia Canada; Version 3.0 is used
for this work. This coder can accept input video of various
formats and includes almost all options including that for
advanced mode defined for H.263 standard.
Fig.3: Procedure of the MPEG-4 VM8 rate control algorithm
MAD- Mean Absolute Difference For example some of the options which are used in this work
Q Quantizer Parameter T Target Bit are as follows (Numbers in bracket are default values).
-i <filename> original sequence (required parameter)
There are five steps in the MPEG-4 VM8 rate control -a <n> image to start at [0]
algorithm: -b <n> image to stop at [0]
1. Initialization -x <n> (<pels> <lines>) coding format
2. Computation of the target bit rate before encoding n=1 SQCIF n=2; QCIF n=3: CIF n=4: 4CIF
n=5 16CIF n=6; custom [ 12 : 11 PAR]
SQCIF (128 X 96) QCIF (176 X 144) CIF (352 X 288) e (x, y) = Intensity of output pixel (for each Y, U, V)
4 CIF (704 X 576) 16 CIF (1408 X 1152) The PSNR for entire video sequence is defined in terms of
-k <n> framer to skip between each encoded frame average PSNR
-c <n> rate control method i t
-r <n> target bit rate in bits/sec default is variable bit rate. Average PSNR PSNR(i).....(for each Y, U, and V)....(4)
t i 1
-m write repeated reconstructed frame to disk [off] Where t is total number of frames in video ,
In this Paper the performed of H.263 video codec is evaluated PSNR (i) is the PSNR value for ith frame.
for following Rate control options
C1 : OFF line Sequence title : Angiography
C2 : TMN5 (test model no.-5) Resolution : 176x144
C3 : TMN8 (test model no.-8) Number of frames : 300
The performance are evaluated in terms of subjective quality Color space : YUV 4:2:0
of reconstructed video, measured in terms of COC and peak Frames per Second : 30
signal to noise ratio defined in eq. 1 and 2 for different target Source : Uncompressed progressive
bit rates.
COC (Coefficient of Correlation) ,Mean square error (MSE)
and SNR are important parameters. The PSNR for each frame
is defined as:
X 1 Y 1

i( x, y) e( x, y)
x 0 y 0
COC
X 1 Y 1 X 1 Y 1 (1)
i( x, y) 2
x 0 y 0

x 0 y 0
e( x, y ) 2
Fig.4: Snapshot of Angiography video sequence, frame 30

The PSNR for each frame is defined as : TABLE-I: DETAIL OF THE VIDEO SEQUENCE MOBILE
255 2 255
PSNR 10 log 10 ( ) 20 log 10 ( ) ( for each Y ,U ,V ) S.No. Detail of Sequence
MSE MSE 1 Name Mobile
Where MSE is Mean square error
2 Size CIF (352x288)
Table 5 gives the summary of results obtained in this work
3 Total size[byte] 23612683
regarding the performance of MPEG-4 codec for various rate
4 Frames 300
control methods at 14.4 and 100 kbps target bit rates. It is
5 Playing time[s] 10.03Min
observed that online & offline rate control methods have
6 frame size[byte] 79
almost same performance as far PSNR is concerned. An 7 Max frame size[byte] 118449
Offline control method has less compression ratio and high bit 8 Mean frame size[byte] 78447.45
rate in comparison to online method. Online control (with 9 Mean bit rate[bit/s] 18827388.44
frame skipping) provides lowest bit rate & compression ratio. 10 Peak bit rate[bit/s] 31586400
Video coding using less number of frames can be used in 11 Mean I-frames[byte] 110923.58
applications such as video conferencing & video telephony. 12 Mean P-frames[byte] 81525.07
When rate control is applied and target bit rate is reduced, it is 13 Mean B-frames[byte] 73071.45
observed that the PSNR, bit rate and storage requirement also
reduces [4]. No control method corresponds to fixed
quantization parameter (Q=13) and independent of the target
bit-rate. It may not be suited for CBR applications.
X Y
1
MSE
XY
[i( x, y) e( x, y)]
x 1 y 1
2
(3)

Where, i (x, y) = Intensity of input pixel (for each Y, U, V)

Fig.5: Snapshot of Mobile video sequence

 TABLE II: ANGIOGRAPHY VIDEO SEQUENCE SIMULATION RESULTS FOR C1, C2 & C3

Angiog Angiography Video Sequence

C1 C2 C3
raphy 42
OBR PSNR PSNR PSNR
COC COC COC 40
(Kbps) (dB) (dB) (dB)
38

Y-PSNR(dB)
120.7 0.99111 29.87 0.99107 29.59 0.99109 29.72
36
C1
203.47 0.99848 32.58 0.9985 32.62 0.99857 32.75
34
C2
304.26 0.99971 34.31 0.9997 34.29 0.99978 34.56
32 C3
404.32 1.00004 35.63 1.00001 35.49 1.00006 35.78
30
507.65 1.00029 36.71 1.00023 36.22 1.00033 36.87
28
607.9 1.00066 37.51 1.00052 37.1 1.0007 37.7 120.7 203.47 304.26 404.32 507.65 607.9 710.06 810.02 913.65 1018.13

710.06 1.00108 38.19 1.00067 37.61 1.00119 38.45 Bit Rate(KBPS)

810.02 1.00117 38.88 1.00101 38.1 1.00123 39.03
913.65 1.00132 39.55 1.00111 38.59 1.00137 39.63 Fig.6: Comparative Performance of H.263 Video Coder for Angiography
video sequence for three Rate control methods C1, C2, C3
1018.1
1.00139 39.98 1.00113 38.99 1.00145 40.2
3

VIII.
ANALYSIS OF EXPERIMENTAL
RESULTS
TABLE III: MOBILE VIDEO SEQUENCE SIMULATION RESULTS FOR MPEG-4 VIDEO CODEC
Table 1 describe the Details of the Video sequence Mobile
(a)No Table 2 gives the Simulation results for C1, C2 and C3 for
(c) Bit rate
Contr (b) Bit (d) Bit rate video sequence Angiography and Table 3 gives simulation
Target Control
Various ol Rate Control results for MPEG-4 video codec.
bit rate Online(with
Parameter over Control Online(without
(Kbps) frame The Angiography video sequence has 59 frames, with fixed
bit Offline frame skipping
skipping) background texture and different moving objects. Since
rate
Average performance of rate control methods C1, C2, & C3 depend
48.44 34.12 34.55 34.54
PSNR Y (dB) upon speed, size and texture.
Average C3 giving the best performance among all the three rate
49.78 38.15 38.83 39.12
PSNR U (dB) control methods, and C2 is giving worst quality. There is
Average improvement of 0.13-1.54dB in C3 over C1 and C2 and 0.02-
49.88 38.35 38.87 39.41
PSNR V (dB) 1.32dB in C1 over C2.
100
Bit rate
98.95 99.5 99.9 99.9
(kb/sec)
IX. CONCLUSION
Compression
7:01 95:01:00 95:01:00 95:01:00
ratio
The performance of H.263 codec is analyze and evaluated
Average
44.44 27.85 27.83 27.49 with rate control methods. Effect of different rate control
PSNR Y (dB)
method on different video sequence is shown in this Paper
Average The figure 6 compares the performance of H.263 codec with
45.78 33.7 34.75 33.68
PSNR U (dB) three rate control methods for video sequence Angiography.
Average
46.36 34.15 35.17 33.95 Since no rate control method gives best performance for all
PSNR V (dB) sequences, hence the use of particular method very much
14.4
Bit rate 1478. depend on the nature of video sequence.
44 15.7 27
(kb/sec) 3
Also In this paper we have investigated the effect of rate
Compression
7:01 205:01:00 556:01:00 304:01:00 control methods on the rate-distortion performance of MPEG-
Ratio
4 video codec. It is observed that for the constant bit-rate
(CBR) applications, both on-line and off-line rate control
methods gives the obtained bit-rate very close to the target bit-
rate for medium and high target bit rates. However, for low bit
rate applications only online rate control method results in the 21. http:\\www.ubvideo.com
actual bit-rate being close to the target bit rate. Further it is 22. D.Marpe, T.Wiegand and G.J.Sullivan, The H.264/MPEG Advance
video coding standard and its applications, IEEE Communications
observed that the use of the deblocking filter at the decoder Magazine, vol.44, no.8, pp.134-144, Aug.2006.
improves the subjective quality of the video codec at the 23. V. Lappalainen, A. Hallapuro, T. D. Hmlinen,"Complexity of
Optimized H.26L Video Decoder Implementation," IEEE Trans. Circ
expense of computational complexity. There are many other and Syst. for Video Technol., vol. 13, pp. 717-725, July 2003.
parameters and functionalities in MPEG-4 which are aimed to 24. Information Technology Coding of Audio-Visual Objects Part 10:
be investigated in future. Advanced Video Coding, ISO/IEC International Standard 14496-10,
June 2003.
25. Compressed video communications, by Abdul H. Sadka,published by
REFERENCES John Wiley and sons,2002.
26. Manjanaik.N, Manjunath.R Intra Frame Coding for Advanced Video
1. Video coding for low bit rate communicationITU-T, Recommendation Coding Standard to Control PSNR and Bitrate Using Gaussian Pulse
H.263, ver.1, 1995. International Journal of Emerging Science and Engineering (IJESE)
ISSN: 23196378, Volume-2, Issue-5, March 2014.
2. http://spmg.ece.ubc.co/h.263pleas/h.263 plus.html
27. A. M. Tekalp, Digital Video Processing, Prentice Hall, 1995.
3. ITU-T H.263 Encoder, version 2,Signal Processing and Multimedia
Group, University of British Columbia, Canada. 28. J. Ribas-Conklin, S. Lei. Rate control in DCT video coding for low-
delay communication.IEEE Trans. Circuit Syst. Video Technol. Feb
4. Frank H. P. Fitzek MPEG4 and H.263 Video Traces for Network 1999, 9(1):172-185 proposed draft of adaptive rate control. JVT H017.
Performance Evaluation, IEEE Transactions on Circuits and Systems , 8`s Meeting: Geneva, 20-26. May, 2003.
Vol.2 (2001).
29. Ronda J, Eckert M, et al. Rate control and bit allocation for MPEG-4.
5. Y. Wang, J. Ostermann, Y. Q. Zhang, Video processing and IEEE Trans. Circuit Syst. Video Technol. 1999, 9(8): 1243-1258.
communications, Prentice Hall, New Jersey, 2001.
6. ITU-T H.263 Encoder, version 2,Signal Processing and Multimedia
Group, University of British Columbia, Canada.
7. ISO/IEC JTC/SC29/WGl 1.MPEG 97/M1931. Joint rate control for
multiple video objects based on quadratic rate-distortion model[s].
8. Brown, M., D. Bushmitch, K. Kerpez, D. Waring & Y. Wang, (2009)
Low-bit rate video codec parameter evaluation and optimization, in
Military Communications Conference, MILCOM 2009, IEEE, pp. 1-20.
9. Sandeep Narain, Syed Hasan Saeed MIMO System Performance
Evaluation for High Data Rate Wireless using Space time codes with
orthogonal structureIJMER Vol-4, Issue-6,Mar-2014.
10. Syed Hasan Saeed Yogendra Singh, Shailendra Kumar Implementation
of Switching System for High Speed Data Transmission using
VHDLIJETT Vol-11, Issue-3 pp: 114-117 2014.
11. http://www.ispreview.co.uk, Accessed 5/12/2013.
12. Kidwai Naimur Rahman, Alam M., Khan Ekram, Beg Rizwan,(2012),
"A Efficient memory no list Set partitioned embedded block (NLSK)
wavelet image coding algorithm for low memory devices", International
Journal of Signal Processing, Image Processing and Pattern Recognition
(IJSIP), SERSC, Korea, Vol. 5,No. 4, 2012 ,pp 93-106.
13. www.mobile.ecs.soton.ac.uk/peter/h263/h263.html.
14. http://www.acm.org/sigs/sigmm/mm97/papers/yu/paper.html.
15. Video coding for low bit rate communication, Int.
Telecommun.Union-Telecommun. (ITU-T), Recommendation H.263,
version 1,1995; version 2, 1998; version 3, 2000.
16. A. Garzelli A. Andreadis, G. Benelli and S. Susini, \FEC coding for
H.263compatible video transmission," in International Conference on
Image Processing, Santa Barbara, CA, Oct. 1997.
17. M.Ghanbari,Video Coding: an introduction to standard codes;IEE
press, London 1999.
18. P.Topiwala, G.sullivan, A.Joch, and F.Kossentini, Performance
evaluation of H.26L, TML8 vs. H.263++ and MPEG-4,ITU-T Q.6/SG6
(VCEG) Tech. Rep. N8, sept.2001.
19. Hung-Ju Lee, Ya-Qin Zhang. Scalable Rate Control for MPEG-4 Video.
IEEE Trans. Circuits and Systems for Video Technology2000, 10(6):
878-894.
20. ITU-T Recommendation H.264 and ISO/IEC 14496-10 (MPEG-4)
AVC,Advanced Video Coding for Generic Audiovisual Services,
Version 3: 2005.