typical network traffic using redundancy and interleaving
play

Typical Network Traffic Using Redundancy and Interleaving to - PDF document

Typical Network Traffic Using Redundancy and Interleaving to Ameliorate the Majority is text-based Effects of Packet Loss in a Video File transfer, Email, Web Reliability is critical Stream Latency is not critical


  1. Typical Network Traffic Using Redundancy and Interleaving to Ameliorate the • Majority is text-based Effects of Packet Loss in a Video – File transfer, Email, Web… • Reliability is critical Stream • Latency is not critical • Transmitted Using TCP Yali Zhu, Mark Claypool and Yanlin Liu – Provide reliable service where all bytes arrive Department of Computer Science Worcester Polytechnic Institute CS Technical Report TR-01-03 Multimedia Network Traffic Multimedia Over Internet • High bandwidth • Often suffer from delay, loss and jitter – Can induce congestion � packet loss – degrading multimedia quality • Latency is critical • Loss has the most severe effects on quality • Loss is not critical • Use loss recovery techniques to – Can tolerate some – Improve multimedia quality • Transmitted using UDP – Keep latency low – Provide unreliable service where some packets may be lost Media Specific FEC Multimedia Repair Taxonomy Repair Sender Based Receiver Based Forward Error Interpolation Insertion Regeneration Retransmission Interleaving Correction Repetition • Multiple copies of data • Lower quality of secondary frames 1

  2. Idea of Interleaving Audio Interleaving • Without Interleaving • With Interleaving Encode Interleave WorcesterPolytechnicInstitute otlhnuWsocItreynstcrtiteePeci Transmit sterPolytechnicInstitute WrceserPoytecnicIstitute Decode Goal Groupwork • Above techniques have been done primarily • What are the issues with applying video to audio only redundancy to video vs. audio? • Our goal: • What are the issues with applying interleaving to video vs. audio? – Apply Interleaving and Redundancy to Video • What would be a methodology for evaluating – Evaluate effects on perceptual quality – Evaluate system overhead the benefit? • What performance metrics should you have? MPEG Encoding Techniques Coding Dependency within GOP • Intra-frame encoding – Based on current frame only • Inter-frame encoding I B B P B B P B B I – Based on similarity among frames • Frame types – I-frame (Intra-coded frame) •I frames ( Intra-coded frames) – P-frame (Predictive-coded frame) •P frames ( Predictive-coded frames) – B-frame (Bi-directionally predictive frames) Require information on previous I- or P- frames •B frames ( Bi-directionally predictive-coded frames) Require information on frames before and after 2

  3. Loss Propagation Outline • Loss of one single frame result in multiple • Introduction • Background losses • Approach – Loss of P-frame – Loss of I-frame • System Overhead – B-frame loss has no propagation • Perceptual Quality • Conclusions I B B P B B P B B I Loss Propagation of Second P-frame Repairing a Video Stream Approach • Apply interleaving and redundancy to video .mpg file – Hypothesize will improve perceptual quality mpeg decoder – Measure system overhead 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Original Stream (interleave, redundancy) Interleaver • Build movies (next slide) 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 Interleaved Stream mpeg encoder – With loss, interleaved repair, redundancy repair I B B P B B P B B I B B P B B P B B Encoded Stream • Evaluate with user study (apply loss) Transmitted over Network • Measure system overhead mpeg decoder • Analysis Decoded Stream 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 (apply repair) Reconstructor Reconstructed 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Stream Effect of Two Frames with Different Video Redundancy Compression Rates 1 1 2 2 3 3 4 Encode Transmit 1 3 4 1 1 3 4 Decode Repetition in the case of consecutive loss - (if I frame, bad news) Two Frames with Different Compression Rates Propagation in the case of I, P frame loss - of quality or previous frame 3

  4. Interleaving Approaches Partial Video Interleaving • Partial-Interleaving approach A1 A2 B1 B2 C1 C2 D1 D2 Original Stream – Use sub-frame as basic interleaving unit A3 A4 B3 B4 C3 C4 D3 D4 + sub-frame consists of macro blocks frame A frame B frame C frame D – (Next slide) • Whole-Interleaving approach A1 B1 A2 B2 A3 B3 A4 B4 – Use whole frame as basic interleaving unit Interleaved Stream • Focus on Whole-Interleaving C1 D1 C2 D2 C3 D3 C4 D4 frame A frame B frame C frame D – (Rest of work) Repetition and Partial Video Whole Interleaving Interleaving 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Original Stream 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 Interleaved Stream I B B P B B P B B I B B P B B P B B Encoded Stream (4 repair pictures here) frame loss during Propagation Loss transmission I I-frame loss Reconstructed 1 3 5 7 9 11 13 15 17 Stream 1 1 3 3 5 5 7 7 9 9 11 11 13 13 15 15 17 17 Apply Repetition Effects of lost I-frame with whole-interleaving /distance=2 Effects of Whole-Interleaving Whole Interleaving Distance Propagation Loss Original Stream (GOP Size = 9) I-frame loss 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18... I Reconstructed 1 3 5 7 9 11 13 15 17 Results of Stream Interleaving Distance = 2 Interleaved Streams Apply Repetition 1 1 3 3 5 5 7 7 9 9 11 11 13 13 15 15 17 17 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18... Interleaving Distance = 5 Stream with 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 one I frame loss Results of 1 6 11 16 21 26 31 36 41 2 7 12 17 22 27 32 37 42 … Non-interleaved Streams Apply Repetition 1 2 3 4 5 6 7 8 9 9 9 9 9 9 9 9 9 9 4

  5. A Possible Negative Effect of Outline Whole-Interleaving • Introduction • Background 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 Interleaved Stream • Approach • System Overhead Single Losses • Perceptual Quality 1 2 5 6 7 8 9 10 13 14 15 16 17 18 Reconstructed Stream • Conclusions A Special Case of Single Losses in the Interleaved Stream, distance=2 Redundancy Overhead per Frame MPEG Quality Vs. File Size Redundancy Overhead per Movie Interleaving Overhead Type • About 15% bandwidth overhead Frame Size (Kbytes) 20 • Reason 15 – Intra-frame encoding based on similarity among Primary 10 frames Redundancy 5 – Interleaving 0 + Decrease similarities among consecutive frames n s s m t w o r o i o e + Result in bigger B- and P- frames t c a p N t m i S S i n A Video Clips 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Original Stream 1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 Interleaved Stream 5

  6. Proposed Solution to Bandwidth Outline Overhead • Encode using different MPEG qualities • Introduction • Background – Original stream with MPEG quality 1 • Approach – Stream with added repair with MPEG quality 2 • System Overhead – File size decreases exponentially – Video quality slightly decreased • Perceptual Quality + may be undistinguishable by users • Conclusions Perceptual Quality for User Study for Redundancy Redundancy • We had over 40 users watch 22 video clips • Video are from variety of television shows • A video clip without loss is first displayed • Video clips are of various loss rate and loss pattern. – Loss Rate: 1 10 20 20 20 – Loss Pattern: 1 1 1 2 4 Consecutive Loss and Redundancy User Study for Interleaving • Parameters to be tested – MPEG Quality 1 & 2 – Loss Rate: no loss, 2%, 5%, 10%, 20% – Movie type: hockey game & CNN news + frequency of scene changes and + intensity of object actions – Distance Value: 2 & 5 • Totally 24 movie clips – 20 seconds / clips 6

  7. Perceptual Quality for Interleaving User Study for Interleaving (II) (hockey) 100 Interleaved hockey Clips Non-interleaved hockey Clips • Parameters not to be tested 90 – Frame Rate: 30 frames/sec 80 – Size of movie: 320 x 240 pixels Average Scores 70 – Hardware difference 60 + All tests on one machine 50 + One user each time – Human interaction: one same assistant (me) 40 30 20 5% 10 % 15% 20% 0 Loss Rate in % Percentage (hockey) Perceptual Quality for Interleaving – Perceptual Quality for Interleaving Movie Type (CNN) 100 Interleaved CNN clips Non-interleaved CNN clips 90 16 80 14 12 Average Scores 70 Average Increased Points 10 60 H o c k e y G a m e C l i p s 8 C N N N e w s C l i p s 50 6 4 40 2 30 0 2% 5% 1 0 % 2 0 % 20 0 5% 10% 15% 20% Loss Rate in Percentage Loss Rate in % Pecentage (CNN) Perceptual Quality for Interleaving – Perceptual Quality vs. MPEG Interleaving Distance Quality 1 0 0 80 90 70 80 60 70 50 Non-interleaved Clips 60 Average Score Average Score quality number 1 40 Distance=5 50 quality number 2 40 30 Distance=2 30 20 20 10 10 0 0 5% 2 0 % 0% 1 0 % Loss Rate in Percentage Loss Rate in Percentage (hockey) 7

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend