JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 6, NO.

2, JUNE 2010 49

Objective and Subjective Evaluation of Spatially

Transcoded Videos for Mobile Receivers
Carlos D. M. Regis, Raissa B. Rocha, Mylene C. Q. Farias, and Marcelo S. Alencar

Original scientific paper

Abstract Digital television produces video signals with differ- must be much lower in order to fit the display of a mobile
ent bit rates, encoding formats, and spatial resolutions. To deliver phone. One solution would be to transmit the video in full
video to users with different receivers, the content needs to be resolution and let the mobile receiver process the video in
dynamically adapted. Transcoding devices convert video from one
format into another. The reception of digital videos using mobile order to reduce its resolution. The problem with this solution is
receivers, implies that the spatial resolution of the video must be the limited processing capacity of a mobile device. Moreover,
adjusted to fit the small display. This paper presents subjective more processing implies an increase in energy consumption.
and objective quality analysis of spatially transcoded videos. A better solution is to process the video using a spatial
Transcoding algorithms that downsample the video frames using transcoder, before sending its signal to the mobile receiver.
the moving average, median, mode, weighted average and sigma
filters are considered. This idea is illustrated in the block diagram shown in Fig-
ure 1 [4], [5]. Transcoding before transmitting saves space
Index Terms Mobile TV, Performance evaluation, Video cod- and production time, because only the content with maximum
ing and processing, Transcoding.
resolution is stored. It also keeps the computational load of the
mobile receiver at a minimal, saving battery time and avoiding
I. I NTRODUCTION overheating.

I N a digital television scenario the video signal may have

different bit rates, encoding formats, and spatial resolutions
according to the type of transmission, the application, and
the receiver. Therefore, standards for digital television define
the reception of video signals in various formats for fixed or
mobile receivers. For example, the Brazilian digital television
standard, known as Integrated Services Digital Broadcasting Fig. 1. The cascaded pixel domain transcoder architecture to reduce the
Terrestrial Built-in (ISDB-Tb), allows the simultaneous trans- spacial resolution [4], [5].
mission of video using the compression standards MPEG-2
and H.264 [1]. This paper presents a comparison among different types of
spatial transcoding methods, which are intended for mobile re-
To flexibly deliver video to users with different available
ceivers. The quantitative performance analysis is presented for
resources, the content needs to be dynamically adapted. Video
two different video quality metrics: objective and subjective.
transcoding is the operation of converting a video from one
format into another [2], [3]. A format is defined by character- II. T HE T RANSCODING P ROCESS
istics such as bit rate, frame rate, spatial resolution, coding
The transcoding functions can be classified as homoge-
syntax, and content. For example, a TV program may be
neous, heterogeneous or as an additional type of function [6],
originally compressed at a high bit rate for studio applications,
[3]. The homogeneous transcoding changes the bit rate or the
but later it may need to be transmitted over a channel at a much
spatial and temporal resolutions. The heterogeneous transcod-
lower bit rate so that it can be displayed in a mobile phone.
ing performs standard (syntax) conversions, including conver-
In the particular case of the reception of digital videos using
sions between interlaced and progressive formats. The addi-
mobile receivers, there is a number of physical limitations
tional type of transcoding function provides different types of
when compared to using traditional television receivers. The
functionalities to the video being transmitted, as for example,
main restrictions are battery life, lower processing capacity,
resistance against errors (error resilience) and addition of
memory capacity, and small displays. Those restrictions im-
watermarks.
pose limitations on the type of video formats that can be
There are two major transcoder architectures: the cascaded
played on a mobile phone or any other device used for mobile
pixel domain transcoder (CPDT) and the DCT (Discrete Co-
reception. For example, the spatial resolution of the video
sine Transform) domain transcoder (DDT) [4]. This article,
Manuscript received September 2, 2009; revised June 23, 2010. adopts the first type of architecture, as shown in Figure 1.
Carlos D. M. Regis, Raissa B. Rocha and Marcelo S. Alencar are with The simplified encoder is different from a stand-alone video
the Federal University of Campina Grande (UFCG), Campina Grande, Brazil encoder since the motion estimation vectors, macroblock cod-
(E-mails: {danilo, raissa, malencar}@iecom.org.br)
Mylene C. Q. Farias is with the University of Braslia (UnB), Braslia, ing modes, and other coding information calculated for the
Brazil (E-mail: mylene@iecom.org.br) original video are reused to generate the new bitstream.
1845-6421/10/8184 2010 CCIS
50 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 6, NO. 2, JUNE 2010

by the filter. The filters have been chosen for their simplicity.
The 1 1 filter corresponds to a simple elimination.

III. P ERFORMANCE E VALUATION

The performance evaluation of any video processing al-
gorithm must take into account the resulting quality of the
generated videos that use the proposed scheme. The most
accurate way to determine the quality of a video is by
measuring it using psycho-physical experiments with human
subjects, called subjective video quality assessment. Subjective
measurements are expensive and time consuming [13], but
they are considered an essential step in the process of choosing
the best video processing techniques.
Fig. 2. Representing the neighborhood of the central pixel with value ps . The other option is to use algorithms that give a physical
measure (objective metric) or estimate of the video quality.
Although the use of such metrics is fairly standard in the litera-
In this article the spatial resolution of videos in CIF (Com- ture, the outputs of these metrics do not always correlate with
mon Intermediate Format) (352 288 pixels) is reduced to human judgments of quality. Customarily, objective quality
obtain videos in QCIF (Quarter Common Intermediate Format) measurements have been limited to a few objective measures,
(176 144 pixels). The reduction from CIF to QCIF was such as the peak signal-to-noise ratio (PSNR). But, in the past
chosen because most equipments use the video files in both few years better video quality metrics that corresponds to the
formats. human perception of quality have been developed.
Both objective and subjective quality assessment techniques
The spatial transcoding operation consists of a down-
are used to estimate the quality and the performance of
sampling process with a factor of 2 : 1 in the horizontal
the spatial transcoder are presented. These techniques are
and vertical directions [7]. A filtering operation is used to
described next.
downsample the signal and consists of substituting pixels in
a 2 2 window (hence the 2 : 1 factor) by the result of a
filtering operation. In the general case, the pixels inside an A. Subjective Test Methodology
M M block are substituted by a single pixel, which has the The subjective video quality assessment technique used is
value calculated by one of the mathematical operations (filter): called Pair Comparison (PC) method [14]. The technique is
Moving Average [8], Median [9], [10], Mode [11], Sigma [12], usual in multimedia applications and provides results with
and Weighted Average. The operations are computed over the good precision.
neighboring N N surrounding block (N > M ). In the PC method the test sequences are presented in pairs.
The weighted average calculates the average of a particular Each pair of sequences corresponds to the same original
set of pixels but, in this case, each pixel may have a different sequence, but each sequence is processed by one of the
weight. For a block of neighboring pixels such as the one systems under test. The source sequence is treated as an
shown in Figure 2, different weight distributions and, con- additional system under test. The systems under test (A, B,
sequently, weight averages can be calculated. Three of these C, etc.) are generally combined in all the possible n(n 1)
weighted averages are used, given by the following equations: combinations forming pairs of sequences, such as AB, BA,
1 1 CA, etc. All pairs of sequences are displayed in both possible
g(x, y) = xs + (x0t + xt + x0u + xu ) , (1) orders (e.g. AB and BA). After each pair is presented, the
2 8
subject is asked to make a judgment on which element of the
1 1 0 pair is preferred in the context of the test scenario.
g(x, y) = xs + (x + xt + x0u + xu ) + The PC method is precise for differentiating among different
2 10 t
1 methods, even when the differences among them are not
+ (x0v + xv + x0z + xz ) , (2)
40 visible. This method was chosen because the size of the
displays makes it hard for the subject to differentiate between
1 1 two test sequences. Presenting them in pairs makes this task
g(x, y) = xs + (x0t + xt + x0u + xu ) +
4 8 easier. A total of 20 subjects were used in the psycho-physical
1 experiments. The subjects were students from the Federal
+ (x0v + xv + x0z + xz ). (3)
16 University of Campina Grande. Each subject watched four
times each of the combinations of the six test sequences. In
This article presents the analysis of those filters, with total, the subjects watched 120 video clips.
neighboring windows of size 1 1, 2 2, 3 3, and 4 4. The subjects used an answer sheet to record the judgment
For the windows of sizes 3 3 and 4 4 (N = 3 and N = 4), scores (Mean Observer Scores MOS) for each of the test
this corresponds to regions around the 2 2 block (M = 2), sequences. They used a scale of integer numbers, ranging
which is the area to be substituted by a single value calculated from 0 up to 10. The videos were displayed on the cell phone
REGIS et al.: OBJECTIVE AND SUBJECTIVE EVALUATION OF SPATIALLY TRANSCODED VIDEOS 51

NOKIA N95. The distance between the subject and the device of one variable are related with the one of the other and given
was 18 cm. This distance was computed by multiplying the by [18]
height of the screen of the device by six (3 6 cm), as PA
recommended by ITU-T [14]. The tests lasted, on average, 30 j=1 [(j )(j )]
= qP , (9)
minutes. For presentation purposes all the videos were encoded
qP
A 2 A 2
j=1 ( j ) j=1 (j )
using the H.264 encoder with a bit rate of 243 kbit/s and
15 frames/s. in which A is the number of samples and e are the
variables to be related. When the correlation equals 1, it is
said to be strong.
B. Objective Metrics
Two metrics were used for objective evaluation of the video IV. R ESULTS
quality: PSNR and SSIM (Structural Similarity Metric). The
This section presents the results of the performance eval-
PSNR estimates the quality of a video frame by comparing a
uation (objective and subjective) of the spatial transcoders
reference to the corresponding processed version of it using
presented in Section II. The original videos used in the tests
the following expression [15]
were the Mobile, News and Foreman. Each video has 10
seconds and is publicly available for download [19]. Those
F
1 X M N 2552 videos were chosen because they contain a good mixture of
P SN R(x, y) = 10 log PM PN ,
F 2 texture, movement, and colors.
k=0 i=0 j=0 (x(i, j) y(i, j))

in which x and y are the original and processed pictures, i and

j are the spatial coordinates, M and N are the dimensions of A. Objective Evaluation
the frame and F the number of the frames. The transcoding techniques used in this subsection are the
The structural similarity metric (SSIM) attracted the at- Moving Average, Median, Mode and Sigma filters, with sizes
tention of the research community because it provides good 1 1, 2 2, 3 3, and 4 4 and the Weighted Average 1
results [16]. It is a full-reference objective metric that estimates (Equation 1), Weighted Average 2 (Equation 2) and Weighted
quality by measuring how the video structure of a processed or Average 3 (Equation 3) filters. The techniques are presented
distorted video differs from the structure of the corresponding in Table I.
reference (original) video. The structural information consists
TABLE I
of the attributes of the picture that reflect the structure of
T HE TRANSCODING TECHNIQUES .
objects in the scene, independent of the average luminance
and contrast. Number Filter
The SSIM metric defines the luminance, contrast and struc- 1 Simple Elimination
ture comparison measures, as given by the equations: 2 2 2 Moving Average
2x y 3 3 3 Moving Average
l(x, y) = , (4) 4 4 4 Moving Average
2x + 2y
5 2 2 Median
2x y
c(x, y) = , (5) 6 3 3 Median
x2 + y2 7 4 4 Median
xy
s(x, y) = , (6) 8 2 2 Mode
x y 9 3 3 Mode
The SSIM metric is given by the following equation 10 4 4 Mode
(2x y + C1 )(2xy + C2 ) 11 Weighted Average 1
SSIM(x, y) = . (7) 12 Weighted Average 2
(2x + 2y + C1 )(x2 + y2 + C2 )
13 Weighted Average 3
The constants, C1 and C2 , are given by 14 2 2 Sigma
C1 = (K1 L)2 and C2 = (K2 L)2 , (8) 15 3 3 Sigma
16 4 4 Sigma
in which L is the dynamic range of the pixel values, and
K1 and K2 are two constants whose values must be small,
such that C1 or C2 will cause effect only when (2x + 2y ) or Figure 3 shows the PSNR results obtained for the set
(x2 + y2 ) is small. For all experiments reported in this article, containing the spatially transcoded videos. For the Mobile
K1 = 0.01 and K2 = 0.03 [17], respectively, and L = 255, video the test showed that the best results were obtained with
for 8 bits/pixel in gray scale images. The values of the SSIM the 4 4 Sigma, 2 2 Sigma, and 4 4 Median filters. The
quality measure are between 0 and 1, with 1 as the best News video obtained the best results for the 22 Sigma, 22
value (better quality). Median and 4 4 Median filters. The Foreman video obtained
The correlation between objective and subjective measure the best results with the Weighted Average 3, 3 3 Moving
was calculated, which indicates the extent to which the values Average and 2 2 Sigma filters.
52 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 6, NO. 2, JUNE 2010

Fig. 3. PSNR curves for the transcoded videos. Fig. 5. SSIM curves for the transcoded videos.

Figure 4 shows the result and the PSNR curves, respectively,

for the transcoded videos after coding using the H.264 codec.
The best transcoding results were obtained for the Mobile,
News and Foreman videos, using the 4 4 Sigma, 2 2
Sigma and Weighted Average 3 filters, respectively. The qual-
ity loss is noticeable when comparing those results with the
ones obtained after the H.264 coding. The best result was
obtained with the 4 4 Moving Average filter.

Fig. 6. SSIM curves for an encoded video after transcoding.

metric, were obtained for the 2 2 Median and 2 2 Sigma

filters. For the H.264 coded videos, after the transcoding
process, the best results were obtained with the 4 4 and
3 3 windows.

B. Subjective Evaluation
Fig. 4. PSNR curves for an encoded video after transcoding. The transcoding techniques used in this subsection are those
that provided the best results for the objective evaluation. The
techniques are presented in the Table II.
Figure 5 shows the SSIM results obtained for the set For the Foreman video the MOS scores are shown in Figure
containing the spatially transcoded videos. It can be observed 7. It can be noticed from the bar plots in Figure 7 that the best
that the best results for the Mobile video were obtained using results for that video were obtained using the 2 2 Sigma,
the 2 2 Sigma, 2 2 Median, and 4 4 Median filters. For 2 2 Median, Weighted Average 3, and 3 3 Median
the videos News and Foreman, the best results were obtained filters.
using the 2 2 Median and 3 3 Moving Average filters. For the Mobile video the MOS values from the experiment
Figure 6 shows the results and the SSIM curves, for the are shown in Figure 8. The best results for this video were
transcoded video after coding. obtained using the Weighted Average 3 and 3 3 Median
For the transcoded videos the best results, using the SSIM filters.
REGIS et al.: OBJECTIVE AND SUBJECTIVE EVALUATION OF SPATIALLY TRANSCODED VIDEOS 53

TABLE II
T HE TRANSCODING TECHNIQUES .
Number Filter
1 2 2 Sigma
2 2 2 Median
3 3 3 Moving Average
4 Weighted Average 3
5 3 3 Sigma
6 Weighted Average 2
7 Weighted Average 1
8 3 3 Median

Fig. 9. MOS bar plot obtained for the News video.

correlation for the videos Foreman ( = 0.3721) and Mobile

( = 0.3209). The News video provided a somewhat higher
correlation ( = 0.7745). As expected, the correlation between
the MOS and SSIM values was higher. For the Foreman
video the correlation was 0.5837, for the Mobile video it was
0.372, and for the video News the correlation was 0.8486. The
News video provided the highest correlation considering both
objective and subjective measures, which can be attributed to
the lack of movement in the video.
Overall, the filters that presented the best results were the
Fig. 7. MOS bar graph obtained for the Foreman video.
2 2 Sigma, 3 3 Median, Weighted Average 3, and 2 2
Median. Thus, an evaluation of the processing time for each
method has the potential to indicate which videos presented
the best results.

C. Processing Time
Another important factor that should be considered when
comparing different algorithms is the processing time, or
computational complexity of the algorithm. Table III shows the
processing time for each of the transcoding algorithms under
test, to indicate the time spent as the filter window increases.
Table III shows that the Sigma and Mode filters demand
longer processing times as compared to the Moving Average
and the Weighted Average filters. This is why those techniques
need to be compared. Also, the Median processing time is
slightly higher than for the Average filter. Considering only
the processing time, the best results were obtained for the
filters: Weighted Average, 2 2 and 3 3 Moving Average,
Fig. 8. MOS bar plot obtained for the Mobile video. and 2 2 Median.

V. C ONCLUSION
For the News video the MOS gathered from the experiment This article presented an analysis of the subjective and
are shown in Figure 9. The best results for this video were objective quality of spatially transcoded videos. The transcoder
obtained using the 2 2 Sigma and 2 2 Median filters. operation consisted of downsampling the video frames us-
The correlation between the MOS and PSNR results for ing Moving Average, Median, Mode, Weighted Average and
each transcoded video was calculated, resulting in a low Sigma filters with sizes 1 1, 2 2, 3 3, and 4 4.
54 JOURNAL OF COMMUNICATIONS SOFTWARE AND SYSTEMS, VOL. 6, NO. 2, JUNE 2010

TABLE III
[3] J. Xin, C.-W. Lin, and M.-T. Sun, Digital video transcoding, Proceed-
P ROCESSING TIME FOR A VIDEO . ings of the IEEE, vol. 93, no. 1, pp. 8497, Jan. 2005.
Transcoding Method Time(seconds) [4] J. Xin, M.-T. Sun, B.-S. Choi, and K.-W. Chun, An HDTV-to-SDTV
spatial transcoder, Circuits and Systems for Video Technology, IEEE
Simple Elimination 0.47
Transactions on, vol. 12, no. 11, pp. 9981008, Nov 2002.
2 2 Moving Average 1.30 [5] Y.-R. Lee, C.-W. Lin, S.-H. Yeh, and Y.-C. Chen, Low-complexity
3 3 Moving Average 1.13 DCT-domain video transcoders for arbitrary-size downscaling, 2004
4 4 Moving Average 3.89 IEEE 6th Workshop on Multimedia Signal Processing, pp. 3134, Sept.-
2 2 Median 1.59 1 Oct. 2004.
3 3 Median 5.00 [6] I. Ahmad, X. Wei, Y. Sun, and Y.-Q. Zhang, Video transcoding: an
4 4 Median 13.69 overview of various techniques and research issues, Multimedia, IEEE
Transactions on, vol. 7, no. 5, pp. 793804, Oct. 2005.
2 2 Mode 7.78 [7] P. Yin, M. Wu, and B. Liu, Video transcoding by reducing spatial
3 3 Mode 8.22 resolution, International Conference on Image Processing, vol. 1, pp.
4 4 Mode 56.47 972975, 2000.
Weighted Average 1 0.75 [8] T. Acharya and A. K. Ray, Image Processing - Principles and Applica-
Weighted Average 2 1.19 tions. Hoboken, New Jersey, USA: John Wiley & Sons, Inc., 2005.
Weighted Average 3 3.42 [9] M. Ahmad and D. Sundararajan, A fast algorithm for two dimensional
median filtering, Circuits and Systems, IEEE Transactions on, vol. 34,
2 2 Sigma 5.76
no. 11, pp. 13641374, Nov 1987.
3 3 Sigma 12.06 [10] A. Bovik, T. Huang, and J. Munson, D., A generalization of median
4 4 Sigma 20.50 filtering using linear combinations of order statistics, IEEE Transactions
on Acoustics, Speech and Signal Processing, vol. 31, no. 6, pp. 1342
1350, Dec 1983.
[11] H. Wu and K. Rao, Digital Video Image Quality and Perceptual Coding.
Boca Raton, FL, USA: CRC Press Taylor & Francis Group, 2006.
The objective quality evaluation of the transcoded videos [12] R. Lukac, B. Smolka, K. Plataniotis, and A. Venetsanopoulos, Gen-
used the PSNR and SSIM metrics. The PSNR results were eralized adaptive vector sigma filters, International Conference on
considered satisfactory, and the filters 4 4 Median, 2 2 Multimedia and Expo. ICME 03., vol. 1, pp. I53740 vol.1, July 2003.
[13] T. H. Falk and W.-Y. Chan, Performance study of objective speech qual-
Sigma, and 22 Median produced the best results. The results ity measurement for modern wireless-voip communications, EURASIP
obtained with the SSIM metric were also satisfactory, and the Journal on Audio, Speech, and Music Processing, p. 11 pages, 2009.
filters 2 2 Sigma and 2 2 Median showed the best results. [14] ITU-T, ITU-T Recommendation P.910, subjective video quality assess-
ment methods for multimedia applications, September 1999.
A subjective experiment was performed to obtain a more [15] Q. Huynh-Thu and M. Ghanbari, Scope of validity of psnr in im-
reliable quality assessment of the transcoded videos. The age/video quality assessment, Electronics Letters, vol. 44, no. 13, pp.
experiment was performed using the Pair Comparison (PC) 800801, 19 2008.
[16] R. de Freitas Zampolo, D. de Azevedo Gomes, and R. Seara, Avaliacao
method described in the ITU-T P.910 Recommendation [14]. e comparacao de metricas de referencia completa na caracterizacao
The test sequences were displayed on the NOKIA N95 cell de limiares de deteccao em imagens, XXVI Simposio Brasileiro de
phone. Subjects used a scale of discrete numbers, ranging from Telecomunicacoes - SBrT 2008, Sept. 2008.
[17] Z. Wang, L. Lu, and A. C. Bovik, Video quality assessment using
0 to 10 to inform their quality judgments (MOS). structural distortion measurement, in in Proc. IEEE Int. Conf. Image
The data gathered from the subjective experiment showed Proc, 2002, pp. 6568.
that all transcoded videos presented MOS values above 7, [18] J. Hu and J. Gibson, New rate distortion bounds for natural videos
based on a texture dependent correlation model in the spatial-temporal
which is an acceptable subjective evaluation. The data also domain, in Communication, Control, and Computing, 2008 46th Annual
showed that the 2 2 Median and 2 2 Sigma filters provided Allerton Conference on, Sept. 2008, pp. 9961003.
the best results in terms of quality. [19] YUV video sequences, http://trace.eas.asu.edu/yuv/index.html,
November 2008.
The computational complexity of the proposed transcoding
algorithms was also tested. Considering only the processing Carlos Danilo Miranda Regis was born in Guara-
time, the best results were obtained for the Weighted Average, bira, Brazil. He received his Bachelor Degree in
2 2 and 3 3 Moving Average, and 2 2 Median filters. Electrical Engineering, in 2007, and his Masters De-
gree, in 2009, in Electrical Engineering, both from
The spatially transcoded algorithms using the 2 2 Median the Federal University of Campina Grande (UFCG),
and 2 2 Sigma filters produced the best results, both Brazil, where he is currently a doctoral student. He
objectively and subjectively, and these techniques are the is with the Iecom Executive Staff of the Journal
of Communication and Information Systems (JCIS),
most appropriate to perform spatial transcoding. In particular, since 2006. He is currently a substitute professor
the 2 2 Median filter has the advantage of requiring less at the Federal Institute of Education, Science and
processing time. Technology of Paraba (IFPB), Brazil. His current
interests include video quality metrics, video processing, multimedia, Digital
VI. ACKNOWLEDGMENTS TV, mobile TV and video transmission.
The authors would like to thank CAPES and CNPq for
funding this work and Iecom for providing the equipment and Raissa Bezerra Rocha was born in Campina
Grande. She is a masters student of Electrical Engi-
facilities. neering at the Federal University of Campina Grande
R EFERENCES (UFCG) and is currently taking part in research at
the Iecom.
[1] M. S. Alencar, Digital Television Systems. Cambridge Univerty Press,
2008.
[2] M. Bonuccelli, F. Lonetti, and F. Martelli, Temporal transcoding for
mobile video communication, in the second Annual International Con-
ference on Mobile and Ubiquitous Systems: Networking and Services.
Citeseer, 2005, pp. 1829.
REGIS et al.: OBJECTIVE AND SUBJECTIVE EVALUATION OF SPATIALLY TRANSCODED VIDEOS 55

Mylene C.Q. Farias received her B.Sc. degree in

electrical engineering from the Federal University
of Pernambuco (UFPE), Brazil, in 1995, and her
M.Sc. degree in electrical engineering from the State
University of Campinas (Unicamp), Brazil, in 1998.
She received her Ph.D. in electrical and computer
engineering from the University of California Santa
Barbara, USA, in 2004, for work in no-reference
video quality metrics. Dr. Farias has worked as a
research engineer at CPqD (Brazil) in video quality
assessment and validation of video quality metrics.
She has also worked as an intern for Philips Research Laboratories (The
Netherlands) in video quality assessment of sharpness algorithms and as
a consultant for Intel Corporation (Phoenix, USA) developing no-reference
video quality metrics. She is currently an assistant professor at the University
of Braslia (UnB), Brazil. Her current interests include video quality metrics,
video processing, multimedia, watermarking, and information theory.

Marcelo Sampaio de Alencar was born in Serrita,

Brazil in 1957. He received his Bachelor Degree
in Electrical Engineering, from Federal University
of Pernambuco (UFPE), Brazil, in 1980, his Mas-
ters Degree in Electrical Engineering, from Federal
University of Paraiba (UFPB), Brazil, in 1988 and
his Ph.D. from University of Waterloo, Canada,
in 1994. Marcelo S. Alencar has more than 30
years of engineering experience, and 21 years as an
IEEE Member, currently as Senior Member. For 18
years he worked for the Department of Electrical
Engineering, Federal University of Paraiba, where he was Full Professor
and supervised 30 graduate and several undergraduate students. Since 2003,
he is Chair Professor at the Department of Electrical Engineering, Federal
University of Campina Grande, Brazil. Between 1982 and 1984, he worked
for the State University of Santa Catarina (UDESC). He spent some time
working for MCI-Embratel and University of Toronto, as Visiting Profes-
sor. He is founder and President of the Institute for Advanced Studies in
Communications (Iecom). He has been awarded several scholarships and
grants, including three scholarships and several research grants from the
Brazilian National Council for Scientific and Technological Research (CNPq),
two grants from the IEEE Foundation, a scholarship from the University of
Waterloo, a scholarship from the Federal University of Paraiba, an achieve-
ment award for contributions to the Brazilian Telecommunications Society
(SBrT), an award from the Medicine College of the Federal University of
Campina Grande (UFCG) and an achievement award from the College of
Engineering of the Federal University of Pernambuco, during its 110th year
celebration. He published over 220 engineering and scientific papers and
twelve books, including Digital Television Systems, by Cambridge University
Press, Communication Systems, by Springer, Principles of Communications,
by Editora Universitaria da UFPB, Electromagnetic Waves and Antenna
Theory, Probability and Stochastic Processes, Digital Cellular Telephony,
Digital Telephony, Digital Television and Communication Systems, by Editora
Erica Ltda. He also wrote several chapters for different books. Marcelo S.
Alencar has contributed in different capacities to the following scientific
journals: Editor of the Journal of the Brazilian Telecommunication Society;
Member of the International Editorial Board of the Journal of Communications
Software and Systems (JCOMSS), published by the Croatian Communication
and Information Society (CCIS); Member of the Editorial Board of the Journal
of Networks (JNW), published by Academy Publisher; Editor-in-Chief of
the Journal of Communication and Information Systems (JCIS), special joint
edition of the IEEE Communications Society (ComSoc) and SBrT. He is
member of the SBrT-Brasport Editorial Board. He served as member of the
IEEE Communications Society Sister Society Board and as Liaison to Latin
America Societies. He also served on the Board of Directors of IEEEs Sister
Society SBrT. He is a Registered Professional Engineer. He is a columnist for
the traditional Brazilian newspaper Jornal do Commercio, since April, 2000,
and is currently Vice-President External Relations of SBrT. He is member of
the IEICE, in Japan, SBMO, SBPC and SBEB, in Brazil.