Transcoding based optimum quality video streaming under limited bandwidth

advertisement
Transcoding based optimum quality video
streaming
under limited bandwidth
*Michael Medagama, **Dileeka Dias, ***Shantha Fernando
*Dialog-University of Moratuwa Mobile Communications Research
Laboratory
**Department of Electronics and Telecommunication Engineering
***Department of Computer Science and Engineering
University of Moratuwa
報告者:饒展榕
Outline
 Introduction
 Transcoding-based streaming
 Video quality measurement
 Test environment
 Results
I. INTRODUCTION
 Multimedia applications play a major role in data
communication.
 Applications such as YouTube® and metacafe® are some of
the most popular internet applications that are accessed not
only through internet-based PCs, but also through mobile
networks.
 Despite development of broadband technologies,limitations
exist still in rich video streaming.
 Especially the allowable data rate per user decreases as the
number of users that share the network increases.
 The data rate perceived by the user decreases as the user’s
distance to the base station increases in the mobile
environment. Therefore users will experience interruption of
the video.
 This paper addresses the study of adaptive video streaming
through transcoding with the data rate/data volume and
video quality measurement calculated though objective
measurement techniques.
 Linux based Video LAN Client(VLC)/ffmpeg environment
was set up to stream the video in a local area network.
 Netlimiter is used to observe the instantaneous data rate
perceived.
II. TRANSCODING-BASED STREAMING
 a.Transcoding
 b. Quantization factor
 c. Frame rate
 The video encoding process comprises of converting raw
video data into discrete cosine transformed data, quantization
and finally the process of run length coding to compress the
quantized DCT coefficients.
 Such data blocks are transmitted via the network so that the
receiver decodes the data which is the inverse process of
above to reproduce the data using a media player.
a. Transcoding
 Transcoding is the process of converting one form of video to
another in the compression domain.
 it is possible to stream the multimedia content with different
video parameters using transcoding where the video
information is scaled down.
 high resolution video requires a high data rate while lower
resolution video may require less.
 In that case resolution of the video might become a
transcoding parameter to be appropriately selected before
streaming to meet the available network bandwidth.
 Transcoding, or in general, video adaptation for scalability
could be in the form of temporal scaling, spatial scaling or
both.
 The frame rate is the key parameter that could be changed to
achieve temporal scaling.
 Spatial scaling is achieved by varying the quantization of the
encoding process.
 The quantization level defines how much information is
stored in a block of a frame in terms of bits.
 There is more information in each block when the level of
quantization is high, which leads to the requirement of high
data rate in the transmission and hence provides higher
quality.
b. Quantization factor
 (DCT) coefficients shall be quantized in many different levels
in the quantization process.
 The level of quantization or the resolution of the quantization
reflects the amount of information and the quality of a video
block of the frame being quantized.
 DCT coefficients in a block of the frame shall have certain
amount of information which is part of the total video
sequence.
 Information in the continuous domain gathered as DCT
coefficients are transformed into the discrete domain with
the DCT quantization.
 The“quantization factor” that comes with the ffmpeg video
library which used in this study maps to the level of
quantization that also directly relates to the quality.
 It is simply an integer index that map to the resolution of the
quantization when quantizing the DCT coefficients.
 Lower quantization factors reflect higher resolution leading
to higher data volume and better quality. Data volume and
quality drops as the quantization factor increases.
c. Frame rate
 The frame rate is the number of frames being processed for
streaming. The amount of the information in the video will
vary with varying the frame rate to achieve temporal
scalability.
 In a limited bandwidth situation, even a poorer quality video
at a lower frame rate may be perceived to be of a better
quality compared to the same video with interruptions.
 It was observed that videos having frame rates of 10 -15 give
poor quality of perception. Frame rate above 15 gives an
acceptable level of quality.
III. VIDEO QUALITY MEASUREMENT
 This is called the Mean opinion score (MOS) and has values
ranging from 5 for best to 1 for poor. The average of the
opinions of those polled will be the quality of the video.
IV. TEST ENVIRONMENT
V. RESULTS
VI. CONCLUSION
 Comparison of video quality and data rate/data volume with
the change of transcoding parameters is carried out in this
study. With low frame rate and low quantized resolution the
data volume of the video can be reduced to a great extent.
 Since the data volume is less, the required bandwidth also less.
 In low bandwidth situations, transcoding can be appliedto
meet the available resources.
Download