On Combining Temporal Scaling and Quality Scaling for
On Combining Temporal Scaling and Quality Scaling for Streaming MPEG Huahui Wu, Mark Claypool, Robert Kinicki Computer Science, Worcester Polytechnic Institute
Introduction § Streaming video – Bitrate > Network capacity • Media Scaling – Temporal Scaling (TS) or Quality Scaling (QS) – Packet loss • Forward Error Correction (FEC) § Operations Research algorithm – To satisfy capacity constraint – To maximize perceived quality § Previous research – Temporal Scaling [TOMCCAP 05] – Quality Scaling [NOSSDAV 05] § This work – Combines Temporal Scaling and Quality Scaling NOSSDAV 06 05/22/2006
Outline § § Introduction Model – Streaming Bitrate (cost) – Video Quality (benefit) § § § Algorithm Experiments Conclusions NOSSDAV 06 05/22/2006
System Layers and Parameters System Layers Parameters & Variables MPEG Scaling and FEC Network NOSSDAV 06 05/22/2006
Streaming Bitrate § Total streaming bitrate, including video packets and FEC packets: where G is the constant GOP rate NPD and NBD are the numbers of transmitting P and B frames depending on Temporal Scaling level l. TS NOSSDAV 06 05/22/2006
Video Quality - Overview § Two distortion factors – Frame Loss • Caused by Temporal Scaling and network packet loss • Appears jerky in the video playout [Wu+ 05 TOMCCAP] • Measured by Playable Frame Rate – Quantization Distortion • Caused by a high quantization value with Quality Scaling • Appears visually as coarse granularity in every frame [Pinson+ 04] • Measured by ITS-VQM § Overall Quality – Distorted Playable Frame Rate NOSSDAV 06 05/22/2006
Playable Frame Rate (R) § Frame Successful Transmission Probability – Where Frame Size § Frame Dependencies § Total Playable Frame Rate NOSSDAV 06 05/22/2006
Distorted Playable Frame Rate (RD ) § Quality scaling distortion varies exponentially with the quantization level [Frossard+ 01] § Distorted Playable Frame Rate NOSSDAV 06 05/22/2006
Algorithm § For each Repair and Scaling combination • Estimate video frame sizes (SI, SP, SB) – Compute streaming bitrate B and make sure it’s under capacity constraint T – Use frame sizes and FEC amount to get successfully frame transmission rate (q. I, q. P, q. B) • Compute playable frame rate (R) • Estimate quality scaling distortion (D) – Compute distorted playable frame rate (RD) § Exhaustively search all FEC and Scaling combination and look for the optimal quality NOSSDAV 06 05/22/2006
Outline § § § Introduction Model Algorithm Experiments Conclusions NOSSDAV 06 05/22/2006
Methodology § Built a function RD() – Returns the distorted playable frame rate § Built an optimization program – TCP-Friendly Bitrate Constraint (T) – Searches repair and scaling levels for the highest RD § Compared three scaling choices with Adjusted FEC – Temporal Scaling – Quality Scaling – Temporal Scaling + Quality Scaling § Compared four FEC choices with the combination of Temporal Scaling and Quality Scaling – – Non-FEC Small Fixed FEC Large Fixed FEC Adjusted FEC NOSSDAV 06 05/22/2006
System Setting Network Layer t. RTT 50 ms MPEG Layer NP 4 frames per GOP s 1 Kbyte NB p 0. 005 to 0. 08 RF § 10 frames per GOP 30 frames per sec GOP: IBBPBBPBB NOSSDAV 06 05/22/2006
Scaling Comparison with Adjusted FEC Low Motion NOSSDAV 06 High Motion 05/22/2006
FEC Comparison with TS+QS NOSSDAV 06 05/22/2006
Conclusions § Summary – Derives analytical models for streaming video with Temporal plus Quality Scaling and FEC – Uses OR algorithm to optimize the quality – Compares scaling methods and FEC methods § Conclusions – Quality Scaling is more effective than Temporal Scaling • But when bandwidth is low and network loss is high, Quality Scaling should be used with Temporal Scaling • Motion matters – Adjusting FEC improves video streaming quality significantly • Better than fixed FEC and non-FEC NOSSDAV 06 05/22/2006
On Combining Temporal Scaling and Quality Scaling for Streaming MPEG Huahui Wu, Mark Claypool, Robert Kinicki Computer Science, Worcester Polytechnic Institute
- Slides: 16