rekIMSA

Information about rekIMSA

Published on February 27, 2008

Author: Callia

Source: authorstream.com

Content

Guidelines for Selecting Practical MPEG Group of Pictures:  Guidelines for Selecting Practical MPEG Group of Pictures The IASTED International Conference on Internet and Multimedia Systems and Applications (EuroIMSA 2006) February 13-15, 2006 Innsbruck, Austria Huahui Wu, Mark Claypool and Robert Kinicki Worcester Polytechnic Institute Computer Science Department Worcester, MA, 01609 USA Presenter - Bob Kinicki Outline:  Outline Introduction GOP Choices for Static MPEG GOP Choices for Streaming MPEG Conclusions MPEG:  MPEG Group Of Pictures (GOP) IBBPBBPBBPBBIBBPBBPBBPBBI… Important Definitions NP: number of P frames in one GOP NBP : number of B frames between two reference frames (I or P frame) GOP Choice:  GOP Choice GOP choice determines MPEG properties Frame size, picture quality Streaming data rate, perceived quality Our exhaustive search approach needs practical limits for the GOP length. Current GOP choice is mostly intuitive! 300 frames long by default in VirtualDub 10 seconds long in NOSSDAV 2005 online video. Mayer-Patel [6] searches a large GOP range (i.e, 35 P frames). [MLC02] Video Quality Metrics:  Playable Frame Rate Counting frames Temporal domain Appropriate when every frame does not change its content. Peak Signal to Noise Ratio (PSNR) Counting pixel differences for each frame Quality domain Appropriate when there is no frame loss. VQM Temporal + Quality domain Handles temporal loss and quality distortion. Video Quality Metrics Outline:  Outline Introduction GOP Choices for Static MPEG GOP Choices for Streaming MPEG Conclusions Static GOP Tradeoffs:  Static GOP Tradeoffs More B frames Advantage: smallest frame size reduces the average frame size Disadvantage: decreases inter-frame similarity, increase B frame size More P frames Advantages: smaller than I frame, reduces the average frame size However, supporting VCR functions is better with short GOPs. Static MPEG Methodology:  Static MPEG Methodology Use the Berkeley MPEG encoder and decoder. Choose varying combinations of NBP and NP. Encode raw images into video clips. Study the impact of NBP and NP on MPEG file size (MB) and frame quality measured by PSNR. 9 Video Clips:  9 Video Clips Each video is 300 raw images with CIF picture size (352x288) encoded at 30 fps. Quantization level is set at 3. NBP Choice :  NBP Choice Impact of NBP on Foreman Video Slide11:  High Motion Low Motion Medium Motion NBP Impact on Eight Videos NP Choice:  NP Choice Impact of NP on Foreman Video NP Impact on Eight Videos :  NP Impact on Eight Videos NBP = 2 High Motion Low Motion Medium Motion Outline:  Outline Introduction GOP Choices for Static MPEG GOP Choices for Streaming MPEG Conclusions Streaming GOP Tradeoffs:  Streaming GOP Tradeoffs More B frames Advantages Low data rate Easy to get through a lossy network Disadvantages: Low inter-frame similarity Add decoding latency More P frames Advantage low data rate Disadvantages Increases propagation errors Repair techniques need to be used Streaming MPEG Methodology:  Streaming MPEG Methodology Develop a streaming MPEG model with packet loss and a capacity constraint Forward Error Correction (FEC) Pre-Encoding Temporal Scaling (PETS) Estimate the video quality (measured by playable frame rate). [WCK03 Choose varying combinations of NBP and NP. Study the impact of NBP and NP on streaming performance. Forward Error Correction (FEC):  Forward Error Correction (FEC) Pre-Encoding Temporal Scaling (PETS):  Pre-Encoding Temporal Scaling (PETS) GOP rate Frame Size Streaming MPEG Optimization Model:  Streaming MPEG Optimization Model where: R is playable frame rate SI, SP and SB are the frame sizes SIF, SPF and SBF are the FEC frame sizes NP and NB are the number of P and B frames G is the GOP rate T is the capacity constraint [WCK03] Streaming MPEG Performance Analysis:  Streaming MPEG Performance Analysis Three FEC choices Non-FEC 5% Fixed FEC Adjusted FEC via the optimization model Use PETS to respond to the capacity constraint. Study varying combinations of NBP and NP. Streaming Foreman Video:  Streaming Foreman Video 2 percent packet loss 1.5 Mbps capacity constraint Eight Streaming MPEG Videos:  Eight Streaming MPEG Videos 2 percent packet loss 1.5 Mbps capacity constraint Low Motion High Motion Medium Motion Outline:  Outline Introduction GOP Choices for Static MPEG GOP Choices for Streaming MPEG Conclusions Conclusions:  Conclusions For Static MPEG Videos The number of B frames between two references frames should not exceed 2. The number of P frames in a GOP need not be more than 5. For Streaming MPEG Videos The number of B frames between two reference frames should be kept at 2 except when dealing with a delay constraint. The number of P frames in a GOP need not be more than 5. Guidelines for Selecting Practical MPEG Group of Pictures:  Guidelines for Selecting Practical MPEG Group of Pictures Huahui Wu, Mark Claypool and Robert Kinicki Worcester Polytechnic Institute Computer Science Department Worcester, MA, 01609 USA Thank You! Questions?

Related presentations


Other presentations created by Callia

GIS presentation
29. 11. 2007
0 views

GIS presentation

Cryptography
05. 01. 2008
0 views

Cryptography

chpt14e
04. 10. 2007
0 views

chpt14e

optie seagull
28. 09. 2007
0 views

optie seagull

Household Waste Water Systems
08. 11. 2007
0 views

Household Waste Water Systems

ipcn leopold
01. 12. 2007
0 views

ipcn leopold

lecture notes 14
02. 11. 2007
0 views

lecture notes 14

Brenda
06. 11. 2007
0 views

Brenda

20050627 SciDAC Straatsma
29. 10. 2007
0 views

20050627 SciDAC Straatsma

arts royalties
16. 11. 2007
0 views

arts royalties

EID zoonoses
19. 11. 2007
0 views

EID zoonoses

Trip Info PPT
20. 11. 2007
0 views

Trip Info PPT

Gothenburg
23. 11. 2007
0 views

Gothenburg

Reiki Lecture
18. 12. 2007
0 views

Reiki Lecture

Cold War what is a cold war
19. 12. 2007
0 views

Cold War what is a cold war

DO YOU NEED A FRIEND
23. 12. 2007
0 views

DO YOU NEED A FRIEND

Mesquite biomass web site
02. 01. 2008
0 views

Mesquite biomass web site

rothberg
01. 10. 2007
0 views

rothberg

jpl zoo1
13. 11. 2007
0 views

jpl zoo1

nguyenngan
24. 02. 2008
0 views

nguyenngan

Gettysburg
29. 02. 2008
0 views

Gettysburg

Sachs Cyber TA ThreatOps
05. 03. 2008
0 views

Sachs Cyber TA ThreatOps

Sperling1762 00
14. 03. 2008
0 views

Sperling1762 00

EN15038 Berlin
18. 03. 2008
0 views

EN15038 Berlin

p2p
02. 10. 2007
0 views

p2p

booklet
30. 03. 2008
0 views

booklet

attr 103788 115
15. 11. 2007
0 views

attr 103788 115

BK12e Ch03 basic
13. 04. 2008
0 views

BK12e Ch03 basic

VTParcelization ButlerII
07. 01. 2008
0 views

VTParcelization ButlerII

lis618n03a 04
06. 12. 2007
0 views

lis618n03a 04

ans11298
17. 12. 2007
0 views

ans11298

07 Lutes
29. 12. 2007
0 views

07 Lutes

Elgg SITE 2007
27. 12. 2007
0 views

Elgg SITE 2007

Network
28. 11. 2007
0 views

Network

Ontologies and friends
10. 12. 2007
0 views

Ontologies and friends

wars ncss 4
05. 11. 2007
0 views

wars ncss 4

DFN2004 aso
21. 11. 2007
0 views

DFN2004 aso

Activities Bran
05. 11. 2007
0 views

Activities Bran

Math Marvels Sample Problems
26. 11. 2007
0 views

Math Marvels Sample Problems

bugtrack
03. 10. 2007
0 views

bugtrack

9 10 saha
21. 12. 2007
0 views

9 10 saha

euphoriamix
05. 11. 2007
0 views

euphoriamix