View
224
Download
1
Tags:
Embed Size (px)
Citation preview
H.264/AVC for Wireless Applications
Thomas Stockhammer, and Thomas Wiegand
Institute for Communications Engineering, Munich University of Technology, Germany
Image Processing Department, Fraunhofer Institute for Telecommunications Heinrich-Hertz-Institut
MoMuC 2003INTERNATIONAL WORKSHOP ON MOBILE MULTIMEDIA COMM
UNICATIONS
Outline
Introduction H.264/AVC
Compression Efficiency and Encoder Flexibility Features for Multimedia Messaging and
Wireless packet-based Streaming Features for Wireless Conversational Services-
Rate Control and Error Resilience Rate-Distortion Optimized Mode Selection
Selected Simulation Results Conclusion
Introduction
The primary goals for H264/AVC are improved coding efficiency and improved network adaptation.
H.264/AVC distinguishes between two different conceptual layers, the Video Coding Layer (VCL), and the Network Abstraction Layer (NAL).
Introduction
Due to the likely business models in emerging wireless systems, in which the end-user’s costs are proportional to the transmitted data volume, and also due to limited resources bandwidth and transmission power, compression efficiency is the main target for wireless video and multimedia applications.
This makes H.264/AVC coding an attractive candidate for all wireless applications including Multimedia Messaging Services (MMS), packet-switched streaming services (PSS) and conversational applications.
Introduction
For efficient transmission in different environments not only coding efficiency is relevant, but also the seamless and easy integration of the coded video into all current and possible future protocol and multiplex architectures.
The VCL specifies an efficient representation for the coded video signal.
The NAL defines the interface between the video codec itself and the outside world.
Structure of H.264/AVC video encoderC
on
trol D
ata Video Coding Layer
Data Partitioning
Network Abstraction Layer
H.320 MP4FF H.323/IP MPEG-2 Etc.
Coded Macroblock
Coded Slice/Partition
H.264/AVCCompression Efficiency and Encoder Flexibility
Figure 1 H.264/AVC Encoder realization with coding options.
H.264/AVCFeatures for Multimedia Messaging and Wireless packet-based Streaming
MMS (Multimedia Messaging Service) Compression efficiency : due to the strict
separation of encoding, transmission and decoding
IDR (Instantaneous Decoder Refresh): for random access and fast forward.
Rate control: for constant video quality Reliable transmission: for wireless link
layer
H.264/AVCFeatures for Multimedia Messaging and Wireless packet-based Streaming
Streaming: online transmission and decoding
Short-term variances in the bit-rate With an appropriate setting of the initial delay
and receiver buffer a certain quality of service can be guaranteed.
Wireless channels commonly provide a constant bit-rate and reliable transmission by using an acknowledged mode within a window of a few seconds.
H.264/AVCFeatures for Multimedia Messaging and Wireless packet-based Streaming
Long-term variances in the bit-rate: due to distance, shadowing, re-newed resour
ce allocation Channel adaptive streaming technologies
Adaptive media playout Rate-Distortion optimized packet scheduling Frame dropping if the channel rate fluctuates
In small range: non-reference frames => temporal scalability
In large scale: I frames, SP (Switching Predictive) frames
H.264/AVCFeatures for Wireless Conversational Services-Rate Control and Error Resilience
The low delay constraint has two main impacts on the video transmitted over wireless bearer services with constant bit-rate. Fast quantization parameter adaptation Temporally backward references in MC
An error-resilient video coding standard suitable for conversational wireless services has to provide to combat two problems: it is necessary to minimize the visual effect of errors within
one frame, and as errors cannot be avoided, the well-known problem of s
patio-temporal error propagation in hybrid video coding has to be limited.
H.264/AVCFeatures for Wireless Conversational Services-Rate Control and Error Resilience
Error-resilience features included in the H.264/AVC standard Slice-Structured Coding
A slice is a sequence of MBs and provides spatially distinct resynchronization points within the video data for a single frame.
Advantages: Packet loss probability can be reduced if slices and transmis
sion packets are relatively small. re-synchronization possibility within one frame
Disadvantages: Increase packet overhead Loss of intra-frame prediction
Group-of-Block (GOB) and Slice Interleaving Reduce no coding overhead in the VCL, but the costly R
TP overhead of up to 40 bytes per packet.
H.264/AVCFeatures for Wireless Conversational Services-Rate Control and Error Resilience
Flexible MB Ordering (FMO) FMO permits the specification of different patter
ns for the mapping of MBs. Data partitioning
reduce visual artifacts resulting from packet losses, especially if prioritization or unequal error protection is provided by the network.
Encoding of single MBs for regions Multiple reference frames
limit error propagation
H.264/AVCRate-Distortion Optimized Mode Selection
The concept of selecting appropriate coding options in optimized encoder designs for many video coding standards is based on rate-distortion optimization algorithms
The Lagrange parameter for appropriate weighting of rate and distortion has to be selected appropriately.
In the H.264/AVC test model, the Lagrangian mode selection is used for motion vector search as well as MB mode and reference frame selection.
Selected Simulation ResultsCompression Efficiency
Figure 2 Coding performance of H.264/AVC codec compared to state-of-the-art video coding standards for QCIF test sequence foreman at frame rate 10 Hz.
Selected Simulation ResultsSlices and Channel-Adaptive Intra Updates
Figure 3 Cumulative distribution of decoded PSNR for differentNAL unit erasure rates for the estimation of the expected distortion in the encoder.
8%
30%
Channel statistics are taken into account into the selection of the coding option in the encoder.
Packet loss rate=4%Pure R-D
Selected Simulation ResultsSlices and Channel-Adaptive Intra Updates
Figure 4 Cumulative distribution of decoded PSNR for differenterror-resilience strategies: channel-optimized intra updates with and without slice structuring for different assumed loss probabilities p.
Slice structuring + Channel adaptive
Slice structuring
Channel adaptive
Selected Simulation ResultsExploiting Feedback in Video Encoding
Figure 5 Cumulative distribution of decoded PSNR for differenterror-resilience strategies: RD-optimized intra updates with slice structuring, and feedback mode with and without slice structuring for delay d=2 and d=4.
RD-optimized + slice structuring
Feedback (2) + slice structuring
Feedback (2)
Feedback (4) + slice structuring
Feedback (4)
Conclusion
In addition to excellent coding efficiency, the design of H.264/AVC also takes into account network adaptation providing large flexibility for its use in wireless applications.
In experimental results based on common test conditions it has been shown that in case without any feedback, several slices in combination with channel-adaptive rate-distortion optimized mode selection is a promising approach.
In case of available feedback, the application of multiple reference frames to exclude error propagation without slice structuring provides excellent results.