Transcript Document
Presentation of Research Paper Scalable Video Coding and Transport Over Broad-band wireless networks Authors: D. Wu, Y. Hou, and Y.-Q. Zhang Source: Proceedings of the IEEE , Volume: 89, Issue: 1 , Jan 2001, pp. 6 -20. Presented by: Yu Hen Hu Overview • Issues of wireless transmission of video • Scalable video coding • Network-aware end system • Adaptive Service © 2002-2005 by Yu Hen Hu 2 Wireless Video Communication System Preprocessing Source Coding Packetize, FEC Network queue Wireless transmitter error resilience Network layer Physical layer Time varying wireless channel © 2002-2005 by Yu Hen Hu Wireless receiver Network protocol Physical layer Network layer Packet recovery Source Decoding Postprocessing error concealment 3 Application: Video Communication • Encoded video – Consists of many segments of bit streams. – Bits highly dependent within each segment – Rate vary dramatically with type of frames, motion, etc. – Not all bits need to be transmitted – lossy compression – Deadline exists for each segment. © 2002-2005 by Yu Hen Hu • QoS – Subjective perceptual visual quality – Objective visual quality measures: PSNR, etc. • Control parameters – SNR scalability: • Quantization levels – Spatial, temporal scalability: • Frame rate, • frame size – Data partitioning – Entropy coding method – Type of frames, macroblocks, etc. 4 Characteristics of Wireless Video Links • High BER (bit error rate) – Due to fading channels (multi-path, shadowing) • Bandwidth variations – Movement of mobile unit – Hand-off between basestations – Noisy channel causes retransmission • Heterogeneity of end terminals – For multi-cast, and broadcast wireless system, one video stream serves multiple destinations with terminals of different capabilities. © 2002-2005 by Yu Hen Hu 5 Physical: Wireless Channels • Time varying channel characteristics (physical layer) – – – – – Fading Interferences Mobile clients Noise Channel estimation required. • Shared spectrum – Limited bandwidth – Sharing in both local spatial and local temporal domains © 2002-2005 by Yu Hen Hu • Resource constrained – Low power – Small form factor display • QoS measures – BER (bit error rate) • Control parameters – Transmission power – Modulation methods (softradio) • Impacts on network packet delivery – Delay – Transmission Error – Mis-match between bandwidth demand and available effective BW. 6 Uni-cast VS Multi-Cast • • • Uni-cast One stream serves one receiver Can not scale up • • • • © 2002-2005 by Yu Hen Hu Multi-cast One stream serves multiple receivers Packets need to be duplicated and transcoded scalable 7 Adaptive Service Framework © 2002-2005 by Yu Hen Hu 8 Scalable Video Coding • • • • © 2002-2005 by Yu Hen Hu Partition of video into layers SNR scalability: Different quantization levels Spatial scalability: Different resolutions Temporal scalability: Different frame rate 9 SNR Scalability 0.1234 0.12 0.12 0.0034 0.0034 0.12 0.0034 © 2002-2005 by Yu Hen Hu SNR Scalable Encoder Quantizer Q at enhancement layer has smaller quantization steps Example: DCT coefficient: 0.1234 Base layer quantized output: 0.12 Enhancement layer input: 0.0034 Enhancement layer output: 0.0034 0.1234 10 Spatial and Temporal Scalability • Lower layer bit streams are obtained from downsampled raw video. x = yn + upsample (yn-1 + upsample (yn-2 + upsample (yn-3 + … + upsample(y0) …)) • Down-sampling and upsampling are performed in both spatial and temporal domain. • Spatial domain: frame size • Temporal domain: frame rate © 2002-2005 by Yu Hen Hu 11 Applications of Scalable Video Coding to Wireless Channels • Adapt to multiple terminal characteristics – Each terminal (receiver) subscribes to different amount of video layers according to its own capability. • Adapt to variable band-width – Send appropriate amount of video layers for the currently available band-width • Network supports are needed to achieve above goals © 2002-2005 by Yu Hen Hu 12 Network aware rate scaling • If network condition (available rate) is known, – encoder can optimize the encoding decision to maximize the perceptual quality subject to rate constraint. – Rate control buffer size may be adjusted to avoid buffer overrun © 2002-2005 by Yu Hen Hu • Encoding decisions that affect rate include: – Quantization level – Coding mode (Intra, inter) selection. – Frame rate – Sending or dropping enhancement layer 13 Network Monitoring Criteria Method © 2002-2005 by Yu Hen Hu Type of monitoring Passive Active Frequency On-demand Continuous Replication Centralized distributed 14 Architecture of scalable video transportation from mobile to wired terminal Network monitoring and adaptation © 2002-2005 by Yu Hen Hu 15 Adaptive Services • Goal: – Rearrange network resources to meet the demand of wireless video transport • Strategy – Reserve minimum bandwidth for baselayer video stream – Adapt resources for enhance layer stream via traffic shaping • Method – Service contract: • specify traffic characteristics and QoS requirement – Call admission control and resource reservation • Ensure enough resources are available for individual services – Mobile multicast • Guarantee QoS during handoff – Sub-stream shaping © 2002-2005 by Yu Hen Hu 16 Transporting from Wired to Mobile Terminals © 2002-2005 by Yu Hen Hu 17 Sub-stream Traffic Shaping © 2002-2005 by Yu Hen Hu 18 Link Layer Error Control: ARQ and RCPC • ARQ (automatic repeat request) is a link layer error control method. • Resend only upon request from receiving end. • Advantage: • RCPC: ratecompatible punctured convolution: – If too late, don’t send – efficient usage of BW • Disadvantage: – Delay unbound © 2002-2005 by Yu Hen Hu 19 Service Comparison Guaranteed Adaptive-base Adaptive – enhanced Best-effort Path setup Y Y Y N Traffic characteristics Y Y N N End-to-end QoS guarantee Y If needed N N Network feedback N N If needed N Resource reservation Y Y N N Bounded delay, zero loss Small delay, low loss Better than best effort N Non-adaptive CBR/VBR Adaptive CBR/VBR Adaptive CBR/VBR None-real time Services QoS Target Applications © 2002-2005 by Yu Hen Hu 20