Transmitting Scalable Video over a DiffServ network

1
Transmitting Scalable Video over a DiffServ
network

• EE368C
• Project Proposal
• Sangeun Han, Athina Markopoulou
• 1/30/01

2
References - Motivation

• Scalable Video Coding Transmission
• U.Horn B. Girod, Scalable video transmission
  for the Internet, Computer Networks and ISDN
  Systems, 1997.
• M. van der Schaar H.Radha, A hybrid
  temporal-SNR FGS for the Internet video, IEEE
  Trans. on Circuits and Systems for Video
  Technology.
• J.Kimura F.Tobagi, Perceived quality and
  bandwidth characterization of layered MPEG-2
  video encoding, SPIE 1999.
• S.McCanne, N.Vetterli V.Jacobson, Low
  complexity video coding for receiver -driven
  layered multicast, JSAC 1997.
• Differentiated Services
• http//www.ietf.org/html-charters/diffserv-charter
  .html
• RFC 2475, RFC 2597
• Software
• NS simulator http//www.isi.edu/nsnam/ns/
• ITU-T Recommendation H.263 (Annex O)

3
Overview

• Scalable video coding
• MPEG-2, H.263 SNR, Spatial, Temporal,
• MPEG-4, H.26L FGS
• Transmission over the network
• In general
• DiffServ
• Our scenario
• Simulation setup
• Issues

4
Scalable Video Coding
5
Fine-Granularity Scalability

• Notes
• Problems limited scalability, error propagation
• Standards MPEG-4, H.26L
• FGS advantages transmission over networks w/ BW
  variation, error resilience

6
Transmission Loss

• Small loss translates into drastic quality
  degradation (loss of important data temporal
  dependence)
• Transmission over the Internet is lossy

• Need to use the available bandwidth to send the
  most important data

7
Solutions

• Feedback adapt transmission rate to variations
• Disadvantages complexity, granularity of BW
  adjustments, delay in feedback, overhead,
  inappropriate for multicast or high variability
• Receiver Driven Layered Multicast
• Problems overhead, delay, granularity of BW
• Smoothing Admission control
• Idea limit stream and load variability.
• Disadvantages complexity, overhead, delay, model
• Loss happens control its effect by dealing with
  it intelligently
• Unequal error protection
• Priority dropping

Use the available bandwidth for the most
important data
8
Priority Dropping

• Loss is inevitable. Limit its effect when it
  happens. Prioritize information according to
  importance (contribution to quality)
• Drop packets according to their priority

Advantages simple sender, handles heterogeneous
receivers short term congestion
9
QoS architectures for the Internet

• Best effort no guarantees
• Integrated Services (IntServ) Per-flow
  guarantees
• Differentiated Services (DiffServ) Per aggregate
  guarantees

10
Example of a DiffServ node

• Packet are marked (DSCP)

• Each packet is treated according to this marking

AF11
11
AF class

• IETF DiffServ WG in RFC 2597 defines 4 Assured
  Forwarding (AF) classes
• Each AFx class marking with AFx1, AFx2, AFx3
• Minimum BW guaranteed for the AFx aggregate.
• 3 dropping priorities (1,2,3).

12
Scalable video in DiffServ

• Use a particular AF Class for Video
• Mark different layers with different AF dropping
  priorities
• Define mechanisms to be used for the Video AF
  Class
• Give rationale on how to create layers to work
  together with AF

13
Approach

• Simulation trying scenarios

Playback buffer
H.263 Encoder Layering
RTP Packet.
Decoding Error Conceal.
Depackt.
DiffServ Network
Marker
(Packet Loss)
Layered video
Layered video
RTP Transport
14
Issues

• Purpose show benefit of combining layering and
  priority dropping
• Network point of view Provide recommendations
  for DiffServ standardization
• How many priorities are really needed?
• How to configure AF class?
• How to choose the layering parameters?
• Coding point of view Explore benefit of
  layering/FGS
• under Internet loss scenarios
• E.g., Tradeoff between motion smoothness and
  quality of pictures under buffer loss, or mix
  with bursty data
• Significance of fine granularity for real
  scenarios

15
Suggestions..?