Get documents about "MPEG Encoding
Document Sample
DIGITAL VI DEO SOLUTION WHITE PAPER
Advanced MPEG-2 Encoding
Introduction
This white paper introduces the advanced MPEG-2 encoding techniques featured on the
standard definition DiviCom™ MV50 encoder from Harmonic. This sixth-generation encoder
supersedes the market-leading MV45 encoder, adding extra processing power, new filters
and intelligent controls to consistently deliver the highest quality video in satellite, broadcast,
and xDSL networks, where bandwidth efficiency matters most.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Making Better Pictures with LookAhead Technology
First-rate encoding at low bit rates requires a combination of refined algorithms and powerful
noise reduction. Pioneered by DiviComTM, "LookAhead" processing is the core of the
Harmonic encoder. It is a technique that dedicates an MPEG-2 processor to fully analyze
incoming video content before the video meets the main processor. With this technique, the
encoder proactively applies the most efficient encoding and noise reduction strategies.
An MPEG-2 encoder is a complex mathematical model developed to match human eye
responses. MPEG-2 standards, developed from a decoding perspective, offers the freedom
to build more efficient performance on the encoding side. Harmonic uses MPEG-2
resources fully to minimize artifacts across a wide range of diverse content.
LookAhead Architecture
Cable headends, whether they are independent, centralized or remote, typically possess
common requirements allowing the reception of incoming services, the encoding of local
services, the modulation of the services for delivery to the home, and the overall control
of these processes. The Harmonic family of products is designed to address these require-
ments effectively and cost efficiently.
The LookAhead architecture harnesses the power of three processors in a combination that
enhances both the encoding and noise reduction, as shown in Figure 1.
Figure 1 - MV50 LookAhead Architecture
LookAhead Video Paths
Processor Feedback And Control
Paths
Delay
ENRGY Main
Processor MPEG-2
VIDEO Integrated
Buffer Noise
Reduction
MV50 Video Compression Core
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
As shown, the LookAhead architecture uses one MPEG-2 processor to extract statistics
from the incoming video almost 1 second ahead of the second encoding processor, while a
third processor provides the tools for motion-compensated noise reduction, making video
easier to encode. This proactive combination of video analysis and noise reduction helps the
encoder make better encoding decisions with available buffer and filtering resources to avoid
artifacts, thereby producing better pictures at lower bit rates.
Although LookAhead processing adds latency to the overall encoding process, the majority
of applications benefit from greater efficiency despite the delay.
How Does LookAhead Improve the Encoding Process?
Video has many characteristics: fast or slow action, film-originated, stills, scene cuts, dis-
solves, special effects, and so on. Algorithms, typically designed for general-purpose use, can
be developed to optimize the encoding to match the incoming content. LookAhead provides
the most effective techniques to eliminate artifacts as follows:
• Maximizes the use of standardized MPEG-2 resources: uses industry-leading DVxpert™
MPEG-2 silicon with Harmonic's custom proprietary microcode
• Improves encoder decisions: selects block-encoding and scan modes
• Customizes motion estimation
• Adapts for brightness: accounts for human visual response to overall scene brightness
that other encoders handle poorly
• Improves I-, P-, and B-frame rate control
• Eliminates short-term variations in quality ("I-frame beating")
• Optimizes the use of available bandwidth within the MPEG picture structure
• Applies appropriate noise reduction, adapting to each scene, thus removing noise while
retaining detail, edge definition, and sharpness.
The following sections describe in more detail how LookAhead improves the encoding
process.
Better Encoder Decisions
The encoding system uses LookAhead statistics to identify scene types so it can apply the
proper encoding strategy to produce maximum film, video, and text quality in the following
ways:
• When the encoder recognizes scene cuts and other complex effects, it overrides the
regular group of pictures (GOP) structure and places I-frames at the optimal moment.
• When the encoder detects still video scenes and knows they exist for the next N
frames, it encodes the still scene with proportionally larger I-frames to attain higher
resolution and greater clarity, while saving bits with smaller P- and B-frames.
For example, artifacts are sometimes visible if different techniques are used in adjacent
macro-blocks, which result in a noticeable discontinuity. The MV50 double-checks decisions
that may be inconsistent within adjacent blocks.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Brightness Adaptation
The human eye is more sensitive to artifacts in dark areas of content. The MV50 matches the
human eye response with LookAhead statistics used to identify darker regions, then adds
more bits to suppress visible artifacts in those areas.
Integrated Noise Reduction
The MV50 architecture has a powerful new integrated noise reduction feature, known as
ENRGY, which couples noise reduction with the compression process. The new preprocess-
ing capabilities of the MV50 are tightly integrated with the LookAhead architecture.
ENRGY Suite of Filters
Noise is the spurious and unwanted component in a video signal. Preprocessing video to
remove noise at low bit rates is more efficient than encoding the noise, which wastes bits and
reduces picture quality. Removing noise without degrading video content requires both high-
quality filters and sophisticated control to remove noise automatically. The LookAhead archi-
tecture inserts a video preprocessor between encoding processors (the compression core).
This coupling of preprocessing and compression enables LookAhead statistics to assist the
intelligent application of pre-processing to provide the highest picture quality at the lowest bit
rates.
Noise reduction and video processing equipment has historically required highly skilled opera-
tors to manually tune the system to match the right parameters to the content. Network man-
agement systems offer access to the filter controls that have been designed around a menu
of templates that make it easy to select universal settings for operational convenience. Once
the basic control parameters have been set, through a user-friendly UI, the system automati-
cally aims to apply the correct filters to achieve optimum results.
The MV50 integrated noise reduction architecture, or ENRGY, consists of the following
filters:
• Motion-compensated temporal low pass filter
• Non-linear spatial filter
• Edge adaptive texture filter
• Temporal low pass filter
• Horizontal resolution filter
• Enhanced adaptive filter control with noise level estimation
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Figure 2 - MV50 integrated noise reduction architecture
Video Motion
Input Impulse Video Compensated Filters
Filter Temporal Horizontal
Filter Edge Adaptive Texture
Non Linear Spatial
Motion Adaptive Temporal
Noise Level Estimation and Border
adaptive filter Processor
control
MPEG-2
Video LookAhead Control Main TS out
MPEG-2 Paths MPEG-2
Processor Processor
MV50 ENRGY Integrated Noise Reduction and Video Compression Core
Other advantages to performing noise reduction within the encoder include more a compact
footprint, integrated control, and lower power requirements. The most compelling advantage
of the MV50 ENRGY solution, however, is leveraging the powerful MPEG-2 control parame-
ters from the LookAhead processor and applying them to the video pre-processing.
See the "ENRGY On The MV50 MPEG-2 Encoder" white paper for more details.
Optimal Motion Estimation
Motion estimation enables the system to extract the temporal redundancy from a sequence.
The MPEG-2 processors use a multi-layer, hierarchical search methodology that yields pre-
cise matches at half-Pel resolution. This provides greater search range than alternative meth-
ods producing exceptionally accurate motion vectors.
At low bit rates, the number of bits required to convey motion vectors becomes more signifi-
cant. The MV50 adds an extra layer of processing to minimize the overhead required to carry
the motion vectors. This technique will also replace inconsistent vectors when a more effi-
cient alternate method is available.
Further, the MV50 analyzes the incoming video signal and modifies its motion estimation to
match the content and thus further improve the encoded results. For each video frame, the
encoder varies where it spends its motion estimation processing power based on the amount
of motion in the frame, and on the presence or absence of special effects.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Maximum Use of the MPEG-2 Buffer
The MPEG-2 encoder must manage the buffer of the MPEG-2 Standard Target Decoder
("STD"), which provides a buffer model that forces the encoder to tightly control output bit
rates. This buffer model anchors the MPEG-2 specification around a standard decoding
foundation ensuring interoperability. With a predefined buffer size, the encoder must ensure
that the buffer never overflows or runs short of data ("underflow").
The LookAhead preprocessor extracts statistics to provide early analysis of video scenes so
that the main encoder can reduce safety margins and use the buffer more efficiently. Figure
3 illustrates how using more of the standardized buffer improves video quality. In effect, the
system can anticipate complex video sections, take precautions, and then prepare to sup-
press artifacts when the video sections arrive at the main processor.
Figure 3 - LookAhead architecture allows extended MPEG-2 "VBV" decoding buffer use to improve the ability of the encoder to
absorb short-term overload
MPEG Buffer level
std.
Safety margin
Single chip CBR encoding
MPEG Buffer level
std.
Safety margin
LookAhead CBR encoding - 13% more usable buffer
Border Processing
While watching a standard television set, viewers cannot typically see the picture border.
Further, picture borders often contain considerable noise. With the MV50, operators can use
border processing, to apply unseen and stronger compression on the picture's boundaries.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Applying Generous Processing Power
The MV50's processing power assists various aspects of the encoding process by thoroughly
analyzing options to make the best decisions. Two examples are:
• While in the adaptive scan mode, the encoder applies zigzag and alternate scans,
compares results, and selects the most efficient scan mode on a frame by frame basis.
• The MV50 applies more than two billion comparisons-per-second in order to detect
pixels that represent spurious impulses.
Statistical Multiplexing: DiviTrackXE
Statistical multiplexing is a well-proven technique that enables a number of encoders to
"pool" their bandwidth. Excellent statistical multiplexing requires accurate complexity analysis,
timely bit rate allocations, efficient encoding, and quality noise reduction. The MV50 was
specifically designed to support the DiviTrackXE statistical multiplexing application. The
MN20 mulitplexer orchestrates this closed loop system, with each encoder producing a VBR
output that is aggregated in the MN20 to produce a Multiple Program Transport Stream at
the desired bandwidth with minimal null packet overhead.
The additional processing power on the MV50 has enabled a new feature, Intelligent Priority
Control (IPC) to enhance the DiviTrackXE statistical multiplexing application. It gives the oper-
ator a mechanism to prioritize the system so that it allocates more bits to a premium program.
IPC is more than a simple weighting scheme. It increasingly protects the priority channels as
the stress on the system increases. Thus, you can assign a priority attribute to each channel
in the DiviTrackXE pool. When the complexity increases beyond a certain norm (when quality
starts to suffer), the system activates the priority mechanism and progressively applies more
bits to the most favored channels.
The LookAhead architecture and ENRGY noise reduction provide all the attributes for an
exceptionally efficient system.
For more detailed information, see the “DiviTrackXE - Advanced Statistical Multiplexing” white
paper.
Ultra Low Bit Rate Applications
The definition of what constitutes “broadcast quality” has changed over time. The Internet has
played a role in this change. It is clear, however, that there is growing demand to support ultra
low bit rate applications based on MPEG-2. To help meet the needs of new “broadcast”
opportunities, the MV50 encoder has been designed to help the service provider deliver
acceptable quality for ultra low bit rate applications.
All the advanced techniques described in this paper have been developed to provide better
video quality at the lowest bit rates. There has been added focus to develop operating modes
that are able to offer useful quality at the very lowest bit rates.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
Operators are typically looking to push MPEG-2 further, to deploy solutions that are “good
enough”. What does this really mean? Typically, operators want a system that is able to deliver
“near broadcast quality” pictures most of the time and can also gracefully deal with overload
scenarios. To get the lowest rates requires efficient encoding and pre-processing strategies
designed to respond effectively to the complexity overload scenarios.
Minimizing compression artifacts is a challenge when coding video sequences containing high
motion and spatial frequency (high detail) content at low bit rates. Crude techniques such as
heavy spatial and temporal filtering make video easier to encode at low bit rates but the end
result is a soft and blurred image that is usually unacceptable to viewers. More sophisticated
techniques are required. The MV50 encoder is equipped with new modes and filters that may
be deployed to deal with the overload scenarios. Two key requirements at low bit rates are (1)
to deal with scene cuts or flashes and (2) to deal with tough sequences such as sports in an
“acceptable” way.
The Edge Adaptive Texture (EAT) filter on the MV50 encoder has some aggressive modes
that high high value for low bit rate applications. This filter identifies areas of texture and
applies a degree of filtering to reduce the waste in allocating bits to image structure that
cannot be coded well (i.e., it is removed or attenuated prior to encoding, rather than by the
encoder). The filter offers a trade-off between compression artifacts and motion rendition that
is similar to content that has been originated on film. To ensure that valuable edge detail is
preserved, complex mechanisms have been developed to protect edge detail. Results show
that the filter is good for sports scenes. As an example, insteaf of the pan-across the crowd
scene looking as though it has been badly encoded the filter produces an effect on the back-
ground crowd scene similar to analog camera blur. The edge preserving mechanisms ensure
that the foreground detail such as players and the ball remain sharp.
Development effort to exploit the potential of the MV50 encoder is expected to produce fur-
ther gains that will clearly demonstrate this platform’s ability to push MPEG-2 a significant bit
further.
Measuring Quality
Video quality relates directly to the efficiency of the algorithms and the bit rate. Evaluating
video quality is important, but unfortunately, there are no metrics that can consistently charac-
terize the measurement of video quality as it is perceived by a viewer. The exceptional ability
of the human eye to judge video quality is difficult to model. Although objective measurement
tools are available, the results are often inconsistent and can produce misleading conclusions.
The most astute operators rely heavily on human eye testing and side-by-side comparisons.
Judging quality when changing parameters results in a tradeoff between resolution and arti-
facts. Whether a change is an improvement or not depends on the video source, the bit rate,
and the viewer's preference. The most refined systems that perform at low bit rates provide
transition and background quality, ability to deal with noise, and fast action motion.
DIGITAL VI DEO SOLUTION WHITE PAPER ADVANCED M PEG-2 E NCODING
What Effects Should You Look For?
Harmonic recommends running a side-by-side comparison with a reference to categorize
and compare the frequency and severity of compression artifacts. Look for settings on the
MV50 that minimize the following effects:
• Edge effects (mosquitos) - noticeable around edge of objects
• Blocking or tiling in response to motion or fades and dissolves
• Blocking or tiling in flat or background areas
• Soft definition
• Motion blur or ghosting
• Pulsing effects, which are annoying and obvious even from a distance
• Non-graceful adaptation effects, which are unnatural or surprising in look
To learn the most about a system, work it aggressively with difficult content at low bit rates.
Consider a wide range of content types, including movies, sports, animation, and studio pro-
gramming.
For more information, see the "Guidelines on Video Quality Evaluation" white paper.
Conclusion
The MV50 encoder is a powerful combination of programmable hardware and software
designed to support challenging compression applications.
Testing video to determine the most appropriate compression rates and settings is a chal-
lenge. Thus, Harmonic is continuing to develop the MV50 platform's extra processing capa-
bilities that reinforce its position as a premium vendor of MPEG-2 compression solutions.
Contact Harmonic for more information about the MV50 encoder.
Harmonic Inc. Tel: +1.408.542.2500 Copyright Harmonic. 2001 All rights reserved. The Harmonic name and logo are registered tradzemarks
of Harmonic Inc. The names of other companies and products mentioned herein may be trademarks of their
549 Baltic Way Fax: +1.408.542.2511 respective owners. Specifications are subject to change without notice.
Sunnyvale, CA 94089 www.harmonicinc.com WP2001-2213 08/01
HARMONIC INC. CONTACT US
Am e r i c a s A s i a - Pa c i f i c
Americas Sales Headquarters Harmonic (Asia Pacific) Limited
549 Baltic Way Suite 703-704, CMG Asia Tower
Sunnyvale, CA 94089 The Gateway, 15 Canton Road
Phone: +1.800.788.1330 inside the U.S. Tsimshatsui, Kowloon
+1.408.542.2500 outside the U.S. Phone: +852.2116.1119
Fax: +1.408.490.6708 Fax: +852.2116.0083
Latin America Harmonic International Inc. B.R.O.
Phone: +1.760.751.3543 Room 510-511, Office Tower A, COFCO Plaza
Fax: +1.760.751.3508 8 Jianguomennei Ave.
Beijing, China 100 005
E M EA
Phone: +86.10.6522.4832
U.K., Middle East and South Africa
Fax: +86.10.6522.4875
21 Progress Business Centre
Whittle Parkway
Slough, Berkshire SL1 6DQ
Phone: +44.1.628.600.100
Fax: +44.1.628.666.736
Europe, CIS and Africa
Continental Square, 4 Place de Londres
Saturne Building, 2nd Floor
ROISSY CDG Cedex, 95727
Phone: +33.1.48.62.92.12
Fax: +33.1.48.62.92.36
Harmonic Inc. Tel: +1.408.542.2500 Harmonic. 2003 All rights reserved. The Harmonic name and logo are registered trademarks of Harmonic
Inc. The names of other companies and products mentioned herein may be trademarks of their respective
549 Baltic Way Fax: +1.408.542.2511 owners. Specifications are subject to change without notice.
Sunnyvale, CA 94089 www.harmonicinc.com 02/24/03
Related docs
Other docs by aakram23
Comment: Hitachi & Sony Supplying 4-Inch LCD Displays For An One iOS Instrumentation
Views: 33 | Downloads: 1
How To Fix iPhone, iPad, iPod Ingest Wi-Fi Connectivity Stock After Upgrading To iOS 5
Views: 49 | Downloads: 0
