Ensure Video Quality in Your Mobile Delivery

Oct 2, 2009 3:58 PM,
Marwan Jabri, Founder and CTO, Dilithium Networks

Though
mobile devices are becoming more sophisticated, with more processing
power and better usability, providing video across multiple platforms
remains a challenge.

Provided by 3G, HSDPA, and Long-Term Evolution (LTE), the boost in
bandwidth availability is fueling the rise in mobile video access from
handsets and netbooks. Larger screens, advances in compression
technologies, flat-rate data-access charges, and more efficient and
improved quality media adaptation technologies also are driving mobile
video access growth.

Variations across mobile screens, netbooks, and handsets dictate the
adaptation of entertainment content as terminals typically vary in
screen size, network access bandwidth, and media/codec processing and
protocol capabilities. The conversion from one format to another is
accomplished in real time or in an offline mode depending on the use
case.

For example, a content asset may be available in an MPEG2 transport
stream (MPEG2TS) file format and require delivery to a netbook user
connected over a 3.5G mobile network with an Adobe Flash player. The
video format in the original stream may be in MPEG2 video and Dolby AC3
audio. Delivering this content for best viewing on the consumer’s
netbook with the Adobe Flash player would necessitate the following
conversions:

• MPEG2TS to Flash video format (FLV)

• MPEG2 video to H.264 AVC

• AC3 audio to AAC

• Bandwidth from a high bit rate of 1 to 6 Mbits/s to few hundred kbits/s

• Screen size from HD (1080i/p, 720i/p) or SD/ED (480i/p) to VGA, HVGA, or QVGA

• And possibly the frame rate from 30 f/s to around 20 f/s

If the netbook consumer uses the Microsoft Windows media player, the
container format would have to be Microsoft’s Windows Media with
Microsoft’s Windows Media audio/video codecs.

Essential Media Adaptation

Depending on the use case, the media adaptation will need to be
performed on demand in real time or in an offline (batch) mode. The
most basic operation of media adaptation is video scaling, encoding or
transcoding, frame-rate, and bit-rate adaptation.

If video codecs are defined by standards, how can one encoder be
superior to another? Standards define valid bitstream structures and
decoders, but not encoders or encoding strategies. So as long as
compressed video bitstreams comply with the specification, the standard
has served its purpose.

As an example, most modern video standards use a motion prediction
tool. Motion prediction is usually about determining whether a portion
of a video frame (e.g., a 16- by 16-pixel macroblock) is present in a
previously encoded frame. If it is, then the amount of bits needed to
encode that portion can be significantly reduced by reusing the
portion.

 To do this, the encoders first need to determine the best matching
block in an area surrounding the macroblock in the previous frame
(motion estimation). On the basis of the match, the encoders next will
determine whether the macroblock is unchanged (encoder can skip) or if
some differences exist and can be encoded. (The encoding of the
differences will yield fewer bits than the encoding of the macroblock
itself.)

Each of these two steps can be implemented in various ways. Some
are more efficient, like compute cycles and memory utilization. Others
provide much greater quality. The second step can be invoked in a
variety of ways, according to an encoding strategy and the context of
the application—constant versus variable bit rate, offline-mode versus
real-time transcoding, etc. Hence, when it comes to video, adherence to
a standard is not a measure of quality.

Real-time or on-demand transcoding is challenging because the
transcoder has to convert the media stream from one codec to another
while maintaining appropriately short input to output latency. The
transcoder does not have the benefit or luxury of multiple passes over
the incoming bitstream to transcode while maintaining bandwidth
efficiency using information from a time later in a clip. The real-time
transcoder has to employ coding strategies that predict the changing
bandwidth needs of video frames and encode them accordingly. Failure to
do so leads to highly variable video quality and inefficient bandwidth
utilization. 

Bandwidth variations are common in networks in general, though they are
more severe in mobile and wireless networks. In today’s 2G to 3.5G
networks, quality of service is rarely in place or enabled, and
best-effort approaches are used. Available bandwidth to a 3/3.5G user
typically varies from 64 to 300 kbits/s. Fluctuations in available
bandwidth result in severe degradation in the customer experience.

Dynamic bit-rate adaptation (DBRA) is a general term used to describe
the ability of a delivery system to adjust its transmitted bandwidth to
match network fluctuations. DBRA approaches are different for real-time
streaming protocol (RTSP) and hypertext transfer protocol (HTTP)
progressive download (PD).

In RTSP, infrastructure providers and handset vendors are increasingly
adopting a standardized approach. For the RTSP DBRA scheme to be of
value, handset support and the support of RTSP streaming and relay
servers are essential.

For HTTP PD, there is no standardized way today to implement DBRA.
Different terminals and media-player technologies implement different
strategies. Some, like Apple’s iPhone (OS 3.0), require a multiplicity
of bit rates to be listed in the content asset, and the decision of
which one to select is left to the iPhone media player. Other streaming
and media-player technologies use different approaches.

Moving Forward with the Essentials

As
mobile video delivery technologies have emerged, designers are facing
new challenges in delivering video to any screen/terminal. Smart
encoding, real-time transcoding, and dynamic bit-rate adaptation
strategies are available today and are essential ingredients for
scalable software-based deployments delivering a high-quality customer
experience at a reasonable cost.