Вы находитесь на странице: 1из 7

J.-Y. Pyun et al.

: Robust Error Concealment for Visual Communications in Burst-Packet-Loss Networks 1013

Robust Error Concealment for Visual Communications


in Burst-Packet-Loss Networks
Jae-Young Pyun, Jun-Suk Lee, Jin-Woo Jeong, Jae-Hwan Jeong, and Sung-Jea Ko, Senior Member, IEEE

Abstract — Compressed video sequences are very spatial redundancy in the time domain or frequency domain to
vulnerable to channel disturbances when they are transmitted estimate the lost macroblock (MB). They interpolate the lost
through the burst-packet-loss networks such as a wireless MBs by averaging available pixels in the neighboring MBs [3].
channel. An effective solution for packet loss is to perform These algorithms are easy to implement, but the reconstructed
concealment at the receiver. Most of the spatial and/or MBs have the blurring artifacts. Also, the SEC methods try to
temporal concealment methods estimate the true motion minimize a measure of spatial variation between the lost MB
vector (MV) associated with a missing block by using the MVs and the neighboring MBs in the frequency domain [4], [5].
of adjacent blocks. However, these methods can not conceal Although these algorithms in [4] and [5] produce the better
effectively the missing blocks, since the neighboring blocks reconstructed image quality than the time domain SEC
can be easily lost by the burst packet losses destroying a large
methods, they introduce heavy computational costs which are
area of coded video. In this paper, we present a robust error
unfavorable properties in the mobile environment. The TEC
concealment method of estimating the true MV of the missing
techniques use the temporal redundancy found in common
block by using the bidirectional motion tracking technique
within neighboring frames of a damaged frame. This motion video sequences to conceal the missing MBs. They estimate
tracking technique uses the linear trajectory feature of the the missing MVs and exploit the pixels from the reference
image block in the forward and backward prediction frames to fill the damaged areas with the motion compensation
directions. Experimental results show that the proposed MBs. Motion compensated temporal prediction has been
algorithm has superior subjective and objective video quality shown to be effective at concealing lost MBs [2]. The most
than conventional error concealment methods in the burst- precise way to find the lost MV is the motion estimation which
packet-loss environments . searches the best MV with the spatial coherence criterion.
Zhang et al. [1] estimates the MV by searching the most
Index Terms — Error concealment, boundary matching
similar neighbors of the lost MB in the previous frame. In [6],
algorithm, motion tracking, motion vector.
a block-matching algorithm is presented to match the upper
and lower adjacent MBs of the lost one to conceal the damage.
I. INTRODUCTION Motion estimation, however, is not proper to mobile terminal
With the generalization of mobile communication services, because of its heavy computational costs. Therefore, the error
recent interest has been focused on the transmission of real concealment method with the low computation complexity is
time video over wireless and other error-prone communication preferable to the mobile environments.
networks. A major drawback in these networks is that they As the well known error concealment method requiring low
introduce a significant number of transmission errors into the computation load, boundary matching algorithm (BMA)
digital bitstream. For compressed video sequences, selects the lost MV among a set of candidates based on a
transmission errors not only corrupt the current decoded frame, spatial smoothness constraint [7]. The lost MV can be selected
but cause the error propagation to succeeding frames which from the predefined MVs such as zero MV, median, and mean
degrades the quality of services. Therefore, it is considerably of MVs. The BMA, however, does not work properly when the
important to develop the error resilience techniques to edges are diagonal or anti-diagonal. To solve this problem, the
minimize the visual degradation caused by transmission error. modified boundary matching algorithm (MBMA) defines
Error resilience can be divided into four parts: error several boundary difference signals for different edge
detection, resynchronization, data recovery, and concealment orientations [8]. This TEC method selecting MV from
[1]. While all of these parts are important, error concealment candidates can effectively conceal the lost MBs when there are
by post-processing is an effective way to combat transmission a few corrupted MBs in the frame. However, in burst error
errors and has been widely investigated [2]. There are two conditions, these algorithms may fail due to the decreased
groups in the error concealment techniques. The one is spatial number of candidates. That is, they can not use neighboring
error concealment (SEC) method and the other is temporal MVs of the lost MB when the burst error occurred in the frame.
error concealment (TEC) method. The SEC algorithms use the In wireless communication channel, the narrow bandwidth
restricts the frame size as QCIF (176 х 144 pixels) so that a
Jae-Young Pyun, Jun-Suk Lee, Jin-Woo Jeong, Jae-Hwan Jeong, and frame is divided into just several packets. Therefore, several
Sung-Jea Ko are with the Department of Electronics Engineering, Korea rows of MBs of one frame are likely to be lost under the burst
University, Seoul, Korea (e-mail: {jypyun, oldy, zski, meeso2, error conditions. By this observation, we propose a robust
sjko}@dali.korea.ac.kr).
Contributed Paper
Manuscript received June 21, 2003 0098 3063/00 $10.00 © 2003 IEEE
1014 IEEE Transactions on Consumer Electronics, Vol. 49, No. 4, NOVEMBER 2003

error concealment technique to overcome the burst error. The the corrupted area do not have any available MVs since all
proposed algorithm estimates the lost MV by using the surrounding blocks are damaged by the burst packet losses. In
forward motion tracking (FMT) and backward motion this case, only zero MV and collocated MV (MVP) are
tracking (BMT). The FMT extrapolates every MV of available which do not help to reach a high quality of temporal
previously decoded frame and estimates the lost MV by using concealment [9].
the overlapped area between the motion extrapolation MBs
and the damaged MB. With the BMT, the proposed algorithm
projects every MV of the next frame into the current frame and Reference Frame
estimates the moving trajectory of the MBs by using the Candidate MVs
MVT[B]
Current Frame
overlapped area between backward motion compensated MBs MVL[R]
MVP
and the damaged MB. Then, two MVs estimated by BMT and Avg(MVT[B], MVL[R])
Median( MVT[B] , MVL[R] )
FMT are averaged and the averaged MV is handled as one of Zero
MVP
the candidate MVs of the damaged MB. MVT Corrupted block

This paper is organized as follows. The conventional error (Collocated MB) MVL
concealment methods are described in Section II. Section III Corrupted area
presents our proposed robust error concealment method using
BMT and FMT. Experimental results and conclusions are MVR
given in Section IV and V, respectively MVB

Fig. 1. MV candidates for the corrupted MBs used in MBA and MBMA
methods.
II. OVERVIEW OF ERROR CONCEALMENTS
In this section, we review a conventional well-known error
concealment technique MBMA with the low computation load, III. PROPOSED ROBUST ERROR CONCEALMENT
and introduce the candidate MVs of the corrupted MB. ALGORITHM USING BIDIRECTIONAL MOTION VECTOR
TRACKING

A. Modified Boundary Matching Algorithm (MBMA)


The existing methods such as the MBMA method can not
use all neighboring MVs since some of them may not be
The MBMA method was proposed to work properly when
received correctly by the burst packet losses. In the proposed
the edges in the boundary are diagonal or anti-diagonal. It
error concealment algorithm, the corrupted MVs among
utilizes the pixel elements adjacent to the boundary to
neighboring MVs are filled with new estimated MVs.
determine the edge orientation at the boundary of the lost
Therefore, the proposed algorithm can use all of the
block and then decides which boundary difference signal is to
neighboring MVs of the lost MB even when they are
be used in estimating the lost block. In particular, MBMA
unavailable.
includes two more boundary difference signals in the diagonal
Our proposed error concealment method consists of two
and anti-diagonal directions for each boundary side matching.
stages. In the first stage, the candidate MVs associated with
The boundary side matching is implemented through a side
missing blocks are estimated by using bidirectional motion
match distortion (SMD) technique finding the best-match
vector tracking (BMVT) technique consisting of FMT and
motion compensated MB [7], [8]. Indeed, the SMD method
BMT. In the second stage, a best-match motion compensated
finds one of MVs of the surrounding MBs with minimum gray-
MV is selected among candidate MVs including the candidate
level transition across the four boundaries (up, left, right, and
MV obtained by the BMVT and the damaged MB is concealed
down).
by using the motion compensated temporal replacement with
the selected MV.
B. Candidate MVs of the Missing Block
A. Candidate MVs of the BMVT
In general, conventional error concealment methods use
only available MVs that survived at the error-prone networks.
The candidate MVs of the proposed algorithm are the MVT,
Fig. 1 shows the MV candidates for the corrupted MBs used in
MVB, MVL, MVR, MVC, zero MV as shown in Fig. 2, where
MBMA. When packet losses are bursty, several consecutive
MVC is the MV of center MB. If MBs are corrupted by
GOBs (or slices) of a frame are corrupted. Therefore, the first
transmission error, the proposed algorithm estimates the MVs
corrupted MB can use the MV of top MB (MVT) and/or the
of MBs. For example, in Fig. 2, MVL and MVT are known
MV of left MB (MVL) to measure the boundary edge
and three MVs (MVR, MVB, and MVC) are candidate MVs
connectivity by using SMD. On the other hand, the last
estimated by the BMVT algorithm.
corrupted MB can use the MV of bottom MB (MVB) and/or
the MV of right MB (MTR). However, some damaged MBs in
J.-Y. Pyun et al.: Robust Error Concealment for Visual Communications in Burst-Packet-Loss Networks 1015

pixels in the areas ( A3 , A5 , and A7 in this example)


Candidate MVs
Current Frame MVT
MVR
overlapping with MBe in Fn . The estimated MV of the lost
MVL
MVT MVB
MVC
MB in Fn is given by
MVC
Zero

MVL MVR M

n
∑ [α ⋅ MVi i
n −1
]
MVB MV = E
i =1
M
, (1)
Corrupted area
∑α
i =1
i

where α i is the weighting factor of MVi n −1 of MB i in frame


Fig. 2. MV candidates for the corrupted MB used in the proposed BMT
method. Fn −1 , MVi n −1 is the MV of the MB i in frame Fn −1 , and
B. Forward Motion Tracking (FMT) M depends on the size of the search range for motion
estimation used in the video encoder ( M = 9 for the 16 х 16
It is important to estimate the MV of the missing block such search range). If the MBe is not overlapped with any MB
that image continuity across its boundaries is preserved. In
projected from Fn −1 or MBs in Fn −1 are intra-coded, we set
[10], error concealment methods using motion extrapolation
has been introduced. If one frame has been heavily damaged MVEn = 0. The MVEn obtained by the FMT becomes the
and its previous frame has not been damaged, it makes senses
center candidate MV (MVC) of MBe . In the second step, the
to investigate how the MVs can be extrapolated to obtain a
reasonable prediction from a past reference frame. In other best-match motion compensated MV is selected among the
words, if the decoded MV from the current frame to the next candidate MVs and the selected MV is used for the motion
compensated temporal replacement to conceal the damaged
frame for a particular block is (vx , v y ) , the MV from the
MB.
previous frame to the next frame should be approximately
(2vx , 2v y ) . However, this method can not conceal the C. Backward Motion Tracking (BMT)
corrupted current frame but changes the reference frame for
the next frame. The BMT method uses the linear motion trajectory feature
To conceal the corrupted MB, we present a motion vector in the backward prediction direction while FMT uses the
extrapolation method of estimating its MV by using the feature in the forward prediction direction. With the BMT, the
overlapped area between the motion extrapolation MBs and proposed algorithm projects every MV of the next frame Fn +1
the damaged MB. Fig. 3 (a) shows the estimation procedure of
the FMT method. Assume that Fn is a frame with MBs into the Fn and estimates the MV of the MBe by using the
overlapped area between backward motion compensated MBs
corrupted by the packet loss and the previous frame Fn −1 has
and MBe . The method to estimate the corrupted MVs is
no error. In the first step, we extrapolate the MVs of Fn −1 into
shown in Fig. 3 (b). Assume that Fn +1 is not damaged by
the current frame Fn . For example, in Fig. 3 (a), three MBs
packet losses. First, three MBs (MBs 3, 5, and 7) in Fn +1 are
(MBs 3, 5, and 7) are projected into Fn using their MVs. The
projected into the frame Fn . They are expressed as the dashed
dashed squares in frame Fn represents the projected MBs
squares in frame Fn . Note that these three MBs have the
from Fn −1 . Since these MBs are overlapped with the corrupted
temporal correlation with MBe in the corrupted area of Fn .
MB MBe in Fn , they should have temporal correlation with
Thus, the MV of MBe can be estimated by the same weighted
the MBe . Thus, the MV of MBe can be estimated by taking average method which FMT used. It is defined as follows:
the weighted average of the MVs of MBs in frame Fn −1 . We
use the weighting factor α i that is equal to the number of
1016 IEEE Transactions on Consumer Electronics, Vol. 49, No. 4, NOVEMBER 2003

Correctly received MB
Overlapped
1 2 3 1 2 3 area with MBe
Corrupted
MV3
3
area
A3 AA3 3 2. MV estimation
MV3
4 5 6 4 5 6 of MBe A3 A3
A5 A5 MBe
MV5 A
MV5 5 Extrapolated
A6 A7 A5
6
A7 MB
MV7 7
7 8 9 Motion 7 A7 8 9
compenstation MV7 1. MV extrapolation
Corrupted
A7
MB
Fn-2 Fn-1 Fn
(a)

Correctly received MB
Corrupted
1 2 3 area 1 2 3 Overlapped
3 MV3 area with
A3 MV3 AA
MBe
A3 A3 33
4 5 6 A5 MV5
MBe 4 5 6
5 A5
A7 A5 1. Motion
A6 MV7 A6
A7
7 tracking MV5
7 8 9 2. MV estimation 7 A7 8 9
of MBe MV7
Corrupted MB

Fn-1 Fn Fn+1
(b)

Fig. 3. MV estimation of missing MB; (a) Motion vector extrapolation (FMT) and (b) Backward motion tracking (BMT).

MVBn , i.e.,
M

n
∑ [α ⋅ MVi i
n +1
]
MVBIn = Average[ MVEn , MVBn ] . (3)
MV = B
i =1
M
, (2)
∑α
i =1
i
If Fn −1 and/or Fn +1 are/is damaged, we use the selection
rule given in the Table I. For example, when Fn −1 is heavily
n +1
where α i is the weighting factor of MVi of MB i in frame
damaged and Fn +1 is correctly received, the BMT method is
n
Fn +1 . After the estimated MV MV is calculated, it becomes
B chosen to estimate the candidate MV of MBe .
the MVC of MBe . In the same way as the FMT method, the
best-match motion compensated MV is selected among the TABLE I
SELECTION OF MV ESTIMATION METHODS
candidate MVs and the selected MV is used for the motion
compensated temporal replacement to conceal the damaged Fn −1 Fn +1 MV estimation method
MB. G G BMVT
G B FMT
D. Bidirectional Motion Vector Tracking (BMVT) B G BMT
Simple temporal replacement or
B B spatial error concealment
The MVEn and MVBn estimated by the FMT and BMT method [9]
methods might be different, since image objects in a sequence G : Correctly received
do not move orderly in the real video stream. Assume that two B : Heavily damaged
neighboring frames Fn −1 and Fn +1 are received correctly. We After estimating the candidate MVs associated with the
n
first perform BMVT by taking the average of MV and E
missing block by using the BMVT, we use the SMD method to
determine the true MV of the missing block. Fig. 4 shows the
J.-Y. Pyun et al.: Robust Error Concealment for Visual Communications in Burst-Packet-Loss Networks 1017

IN
OUT

(a) (b)

Fig. 4. Matching area for the side match distortion; (a) Matching area
for MVT and MVL, (b) Matching area for MVB and MVR

(a) (b) (c)

(d) (e) (f)

Fig. 5. Results of different concealment methods at frame 4 of the Foreman sequence; (a) Original frame, (b) Corrupted frame, (c) MBMA,
(d) FMT, (e) BMT, (f) BMVT.

TABLE Ⅱ
PSNR AND CPU TIME OF THE ERROR CONCEALMENT METHODS IN THE SEQUENCE FOREMAN
Selected frame number
Methods Avg.
4 5 10 11 23
PSNR 25.14 22.94 26.76 25.09 27.84 25.55
MBMA
CPU time 2.33 2.23 2.5 2.46 2.3 2.38
PSNR 25.71 22.91 28.39 26.33 27.82 26.23
FMT
CPU time 2.54 2.39 2.92 2.69 2.92 2.69
PSNR 26.23 22.61 28.06 26.79 28.60 26.46
BMT
CPU time 2.63 2.69 2.85 2.69 2.53 2.68
PSNR 26.20 23.60 28.42 27.21 28.00 26.69
BMVT
CPU time 2.75 2.56 2.79 2.77 2.58 2.69

matching area employed for the SMD. In particular, when


MVT and MVL are correctly received, we use the matching
area as shown in Fig. 4 (a). On the other hand, when MVB
and/or MVR are correctly received, the matching area as
shown in Fig. 4 (b) is utilized. For the other cases that all
candidate MVs are lost or received correctly, either matching
area can be used.
1018 IEEE Transactions on Consumer Electronics, Vol. 49, No. 4, NOVEMBER 2003

TABLE Ⅲ
VIDEO QUALITY AND CPU TIME OF THE ERROR CONCEALMENT METHODS IN THE SEQUENCE TENNIS
Selected frame number
Methods Avg.
2 3 4 16 21
PSNR 21.45 21.33 20.44 22.16 22.08 21.48
MBMA
CPU time 2.69 2.23 2.31 2.46 2.54 2.45
PSNR 22.80 22.11 22.46 24.22 20.08 22.33
FMT
CPU time 2.85 2.09 2.69 2.85 2.62 2.74
PSNR 22.15 21.30 20.63 22.36 24.30 22.14
BMT
CPU time 2.77 2.38 2.89 2.38 2.62 2.60
PSNR 22.44 22.18 21.43 23.38 23.07 22.50
BMVT
CPU time 2.85 2.54 2.23 2.77 2.62 2.60

Now, we conclude that the proposed robust error


IV. EXPERIMENTAL RESULTS
concealment technique can be a useful alternative to existing
techniques due to its low computational complexity and
To evaluate the proposed error concealment technique, two acceptable PSNR performance.
video test sequences, Foreman and Tennis, are used. These
QCIF test sequences of temporal resolution 10 frame/s are
encoded by H.263+ (TMN8) [11], [12] at 96 Kbps. In our V. CONCLUSIONS
experiment, 5 % of the decoded frames are randomly selected
and the selected frame Fn is partially damaged with the When video streams are transmitted over error-prone
networks, the compressed video can suffer severe degradation.
corruption ratio of 50 %. To reduce the video quality degradation caused by the burst
Fig. 5 (a) and (b) show the original and damaged frame, packet losses, we have proposed the robust temporal error
respectively. Fig. 5 (c), (d), (e), and (f) show image frames concealment, named BMVT. The experimental results show
recovered by the MBMA, FMT, BMT, and BMVT methods, that BMVT method can increase decoded video quality even
respectively. It is seen that the MBMA method does not under the burst packet losses. Even though the motion is large,
produce satisfactory result, since the image details are not fully the proposed method approximately can track the moving
recovered due to burst errors. Note that the image frames of objects by using the linear trajectory feature of the motion.
the FMT and BMT methods are quite different owing to the Therefore, the proposed method can estimate the MVs of the
different moving speed of objects between the neighboring corrupted MBs even in the sequences including fast moving
frames. In particular, the FMT produces the lower video objects. The proposed temporal error concealment can be
quality than the BMT method in the selected frame. However, applicable to the video decoder requiring low computation
the recovered image using BMVT shows the improved complexity and the robust error concealment to high scene
subjective quality than other recovered images. This is because activity.
BMVT uses both temporal and spatial correlation while
MBMA uses the only spatial correlation based on the SMD
approach. Table II and III show the PSNR results and CPU REFERENCES
time of the error concealment techniques. To quantify the
[1] J. Zhang, J. F. Arnold, and M. R. Frater, “A cell-loss concealment
computational complexity of each method, we measured the technique for MPEG-2 coded video,” IEEE Trans. Circuits and Syst.
decoding time required for each technique and compared it to Video Technol., vol. 10, no. 4, pp. 659-665, Jun. 2000.
the decoding time for an undamaged sequence. Thus, a CPU [2] Y. Wang and Q. F. Zhu, “Error control and concealment for video
communication: a review,” in Proc. IEEE, vol. 86, pp. 974-997, May.
time of 2.0 indicates that the decoding time for the damaged 1998.
sequence was twice the time required to decode the [3] S. Aign and K. Fazel, “Temporal and spatial error concealment
undamaged sequence. In the MBMA method, the consumed techniques for hierarchical MPEG-2 video codec,” in Proc.
GLOBECOM, pp. 1078-1083, 2000.
CPU time is larger than 2 owing to the boundary matching [4] Y. Wang, Q. F. Zhu, and L. Shaw, “Maximally smooth image recovery
procedure. However, MBMA does not conceal properly the in transform coding,” IEEE Trans. Communications, vol. 41, no. 10, pp.
damaged MBs caused by the burst packet losses since the 1544-1551, Oct. 1993.
[5] H. Sun and W. Kwok, “Concealment of damaged block transform coded
neighboring MBs are also damaged. However, BMVT method
images using projections onto convex sets,” IEEE Trans. Image
produces the better video quality than MBMA at the cost of processing, vol. 4, pp. 470-477, Apr. 1995.
CPU time of 12.5 %, improving the PSNR by 1.1 dB on the [6] S. Tsekeridou and I. Pitas, “MPEG-2 error concealment based on block
average. matching principles,” IEEE Trans. Circuits and Syst. Video
Technology, vol. 10, no. 4, pp. 646-658, Jun. 2000.
J.-Y. Pyun et al.: Robust Error Concealment for Visual Communications in Burst-Packet-Loss Networks 1019

[7] W. M. Lam, A. R. Reibman, and B. Liu, “Recovery of lost or Jin-Woo Jeong received the B.S. degree in Computer
erroneously received motion vectors,” in Proc. ICASSP’93, pp. V417- Engineering with Department of Computer Science and
V420, Apr. 1993. Engineering from In-ha University, in 2002. He is now a
[8] J. Feng, K. T. Lo, H. Mehrpour, and A. E. Karbowiak, “Cell loss M.S. candidate in electronics engineering with the
concealment method for MPEG video in ATM networks,” in Proc. Department of Electronics Engineering at Korea
IEEE Global Telecommunications Conference, vol. 3, pp. 1925-1929, University. His research interests are Mobile QoS, IP QoS,
Nov. 1995. and Error concealment.
[9] S. Valente, C. Dufour, F. Groliere, and D. Snook, “An efficient error
concealment implementation for MPEG-4 video streams,” IEEE Trans.
Consumer Electron., vol. 47, no. 3, pp. 568-578, Aug. 2001.
[10] A. Vetro, H. Sun, Y. Chen, and S. Kung, “True Motion Vectors for
Robust Video Transmission,” in Proc. of SPIE Visual Communications Jae-Hwan Jeong received the B.S. degrees in electronics
and Image Processing, pp. 230-240, Jan. 1999. engineering with the Department of Electronics
[11] ITU-T/SG-15 Video Coding Experts Group, Video codec test model, Engineering from Korea University, in 2002. He is
TMN8, 1997. currently working toward the M.S. degree in electronics
[12] J. Ribas-Corbera and S. Lei, “Rate control in DCT video coding for low- engineering with the Department of Electronics
delay communications,” IEEE Trans. Circuits Syst. Video Technol., vol. Engineering at Korea University. His research interests are
9, no. 1, pp. 172-185, Feb. 1999. in the areas of Mobile QoS, IP QoS, Handoff, video signal
[13] G. S. Yu, M. M. K. Liu, and M. W. Marcellin, “POCS-based error processing and multimedia communications.
concealment for packet video using multiframe overlap information,”
IEEE Trans. Circuits Syst. Video Technol., vol. 10, no. 4, Jun. 2000.

Sung-Jea Ko (M'88-SM'97) received the Ph.D. degree in


1988 and the M.S. degree in 1986, both in Electrical and
Computer Engineering, from State University of New
York at Buffalo, and the B.S. degree in Electronic
Engineering at Korea University in 1980. In 1992, he
joined the Department of Electronic Engineering at Korea
University where he is currently a Professor. From 1988 to
Jae-Young Pyun received the B.S. degree from Chosun 1992, he was an Assistant Professor of the Department of
University, and M.S. degree from Chonnam University, Electrical and Computer Engineering at the University of Michigan-Dearborn.
Kwang-ju, Korea, both in Electronics Engineering, in 1997 From 1986 to 1988, he was a Research Assistant at State University of New
and 1999, respectively. He is now a Ph.D. candidate in York at Buffalo. From 1981 to 1983, he was with Daewoo Telecom where he
electronics engineering with the Department of Electronics was involved in research and development on data communication systems.
Engineering at Korea University. In 1999, he joined the He has published more than 200 papers in journals and conference
Research Institute for Information and Communication proceedings. He also holds over 10 patents on data communication and video
Technology, where he is currently a research engineer. His research interests signal processing.
include Mobile QoS, IP QoS, and video compression and communication. He is currently a Senior Member in the IEEE, a Fellow in the IEE and a
chairman of the Consumer Electronics chapter of IEEE Seoul Section. He has
been the Special Sessions chair for the IEEE Asia Pacific Conference on
Circuits and Systems (1996). He has served as an Associate Editor for Journal
of the Institute of Electronics Engineers of Korea (IEEK) (1996), Journal of
Jun-Suk Lee received the B.S. degree in information Broadcast Engineering (1996 - 1999), the Journal of the Korean Institute of
engineering in 1997 and the M.S. degree in electronics Communication Sciences (KICS) (1997 - 2000). He has been an editor of
engineering with the Department of Electronics Journal of Communications and Networks (JCN) (1998 - 2000). He is the
Engineering from Korea University in 2000. He is 1999 Recipient of the LG Research Award given to the Outstanding
currently working toward the Ph.D. degree in signal Information and Communication Researcher. He received the Hae-Dong best
processing at Korea University. His research interests are paper award from the IEEK (1997) and the best paper award from the IEEE
video signal processing and multimedia communications. Asia Pacific Conference on Circuits and Systems (1996).

Вам также может понравиться