Improving Video Streaming and File Compression Efficiency without - PowerPoint PPT Presentation

Improving Video Streaming and File Compression Efficiency without Affecting Quality By Yves Faroudja

Demand Exceeds Supply = Opportunity BITRATE/ % of Internet DEMAND 8/90% COMPRESSION HEVC 4/75% EFFICIENCY Demand for video bandwidth is • doubling every 3 years H264 2/50% However, network compression • schemes efficiency doubles every MPEG 2 10 years 1/25% 1990 2000 2010 2020 2030 TIME

Realized Technology Goal • Preserve image quality at a bitrate reduction potential of 35% to 50% with existing compression systems via the use of novel pre- and post- processing. • Solution is video compression/codec agnostic.

System Overview OUTPUT(0) LOWER RESOLUTION Conventional Conventional Compression Decompression OUTPUT(n) Main Channel VIDEO Pre- Additional lower Post- INPUT Transmission Processor RESOLUTIONS Processor or Storage OUTPUT(1) Support Channel(s) FULL RESOLUTION MPEG TS

Application Details: Pre-Processor Main Channel H & V Conventional mux Decimator LPFs V IN Compression Reduced Bitrate Send Delay Expand Decompression Match Motion Data Support Layer Pre-Processor Compression

Application Details: Post-Processor High Quality Low Resolution Output To Display or Distribution Receive demux Conventional Expander Full Decompression Main 1/n Resolution Non-linear Pixels Output + Bandwidth Expansion Decompression Support Layer Storage Post-Processor

System Overview OUTPUT(0) 4K LOWER RESOLUTION Conventional Conventional Conventional De- Conventional De- Compression Compression Compression Compression OUTPUT(n) Main Channel Main Channel VIDEO Additional Lower Pre- Post- Transmission Transmission INPUT RESOLUTIONS Processor Processor or or OUTPUT(1) 1080p Storage Storage FULL RESOLUTION 720p Support Channel(s) DVD MPEG TS MPEG TS VCD

Application Example: 1080p Satellite Back-end Transmission High Quality Reduced Output Reduced Bitrate Stream(480p/540p/720p ) Conventional Main Channel Decompression mux demux Pre- Post- Processor Conventional Processor INPUT 1080P Compression 1080p Transmission of reduced bitrate Reduced stream allows for more streams Bitrate per transponder Output Small Support Channel Satellite Uplink Facility Satellite Downlink Facility Second stream at no additional bandwidth cost reduces system complexity

Application Example: 1080p Server Storage High Quality Reduced Output Reduced Bitrate Stream(480p/540p/720p ) Conventional Main Channel Decompression mux demux Pre- Post- Processor Conventional Processor INPUT 1080P Compression 1080p Storage of reduced bitrate Reduced streams increases available space Bitrate 35% -50% Output to Small Support Channel CDN Satellite Uplink Facility Satellite Downlink Facility Second stream at no additional bandwidth cost reduces system complexity

Application Example: 4K UHD Transmission on 1080P Network Very High Quality Reduced Main Channel Reduced Output Stream(1080p/720p ) Conventional Bitrate 1080p Decompression mux demux Pre- Post- Processor Conventional Processor INPUT 4K UHD Compression 4K Deliver the UHD experience now Reduced Bitrate Output Small Support Channel Distribution CDN Second stream at no additional bandwidth cost reduces system complexity

Application Example: 1080P HD Telepresence High Quality Reduced Output Reduced Bitrate Stream(480p/540p/720p ) Conventional Main Channel Decompression mux demux Pre- Post- Processor Conventional Processor Compression Transmission of reduced bitrate streams ensures consistent quality even on crowded Small Support Channel INPUT 1080p Output networks 1080P A pre- and post- processor system is located at each site

Practical Implementation

Computer Simulation • Validate and Measure the algorithms and approach -- Matlab -- C Model -- Executable (Linux) • Process real clips to gauge objective correctness (PSNR, SSIM, etc). • Process real clips to gauge perceptual quality (i.e.. “Beauty”) • Automate • Iterate

Real Time GPU Implementation • HPC-CUDA/OpenCL -- Well suited to speedily process parallel compute and serial compute elements = Real time -- Well suited for rapid prototyping and implementation = Quick changes -- Well suited for server head end = No specialized HW required

Mobile Post-Processing Implementation • Tradeoff of power, processing & speed . -- Should not consume more power than traditional decompression, render, and display -- Validate that decompression, render and display are processing capable at the client -- Validate that decompression, render and display can keep up with real time and low latency playback

Original 2 Mbps/1080p Original 4 Mbps/1080p

Original 2 Mbps/1080p Original 4 Mbps/1080p Processed 2 Mbps/1080p

Original 4 Mbps/1080p Original 2 Mbps/1080p Original 4 Mbps/1080p

Original 4 Mbps/1080p Processed 4 Mbps/1080p Original 2 Mbps/1080p Original 4 Mbps/1080p

In Conclusion • Codec technology cannot keep up with the voracious demand for video bandwidth • The coming clogging of the internet pipes seems inevitable • We are proposing an out-of- the compression ‘box’ solution to improve the situation

Thank you!

FIN