
32e54513ed3ca3692e829c8047a776cd.ppt
- Количество слайдов: 28
A Faster Implementation of Video Compression/Decompression on Multi-core Architectures A Faster Implementation of Video Compression/Decompression on Multicore Architectures Dr. M. J. Nigam and Dr. R. Niyogi Deptt. Of Electronics & Comp. Engg. Indian Institute of Technology, Roorkee Dr. Ankush Mittal College of Engineering Roorkee
Gratitude and acknowledgment for the last project
Project goals l Multi-core implementations of standard algorithms (MPEG, H. 264) l Multi-core implementations of bandwidth preserving recent algorithms (CEZW, MJPEG, content-based compression) l Machine dependent fast decompression l Multi-core implementation of integrated network streaming algorithm
CEZW MPEG 4 (H. 264) MPEG 1/MPEG 2 Uncompressed Video
Types of multi-core PC and Laptop CELL (IBM) CUDA – GPU (n. VIDIA) Workstation
Why Multicore Architecture? l Moore’s Law l Increasing job size and problem complexity l Limitations in clock speed l Power saving requirements l Cost effective
Overall Architecture Traditional system Split video Inter - frame into frames coding DCT based transform coding Save to disk Our system Content based approach Network Transmission End User Uploaded To server Network Transmission MPEG based Frame packaging* End User CEZW+ Bandwidth Estimation Save to disk Uploaded To server SDM
Analysis • Performance requirements met – Resistance to low & fluctuating bandwidth – Perceptible video quality – Scalability • In terms of no. of videos • In terms of no. of users – Simple & real-time decoder – Encode once & decode at multiple bitrates – Robustness
Analysis • Comparison with present systems – RSVP & related systems • Expensive for the end-user – Realmedia, Quicktime, Windows Media • Low, Medium, High bandwidth profiles • Do not deal with educational videos – Youtube & Google Videos • Large waiting time & buffering requirements – Extensive research in this field – No end-to-end framework exists
MPEG Encoding Process
Parallelization of Encoder • Encoders can be parallelized at several level – Data per GOP – Data per Frame – Data per Macro Block
The Cell B. E. Processor • A joint venture by IBM, SONY and TOSHIBA • Rated as super computer on a chip
Architecture
Time based deliverables • 1 -3 months: Multi-Core Implementation (MCI) of MPEG-4 (H. 264) along with documentation • 4 -6 months: MCI of MJPEG • 7 -9 months: MCI of CEZW • 10 -12 months: MCI of Content based • 13 -15 months: Decompression optimization • 16 -18 months: MCI of integrated streaming • 19 -24 months: Through testing, documentation
Technologies Fall-outs • Developing a new e-learning video CODEC, • Providing a video interface for using the CODEC, • Developing standards for the streaming protocol, • Providing framework for network bandwidth optimized content delivery solutions.
Expected Outcomes • Standardized development tools for Elearning courseware. • Software for content-based delivery and distribution of E-learning courseware. • Documentation and dissemination in terms of research papers of the methodologies and framework for enhanced E-learning.
Total Budget outlay S. No. Item Estimated cost 1. Equipment/Software/ Services 50 lakhs 2. Research/Project Staff 15 Lakhs 3. Contingency, Consumables and Travel 15 lakhs Total 80 lakhs The budget will be shared by the two participating institutes equally (i. e. , IIT Roorkee and College of Engineering Roorkee).
S. No. Item Year 1 Year II Total for years 1. Equipment/Softwar e/Services 40 Lakhs 50 lakhs 2. Research/Project Staff 7 lakhs 8 lakhs 15 Lakhs 3. Contingency, Consumables and Travel 7 lakhs 8 lakhs 15 lakhs 54 lakhs 80 lakhs Total 10 lakhs 26 lakhs
About College of Engg. Roorkee l Established in 1998 l Admissions through AIEEE l Headed by ex-Vice Chancellor of IIT Roorkee, Prof. Gopal Ranjan l Ranked top in the northern region
Work done • Dishant Ailawadi, Milan Kumar Mohapatra, Ankush Mittal, Frame. Based Parallelization of MPEG-4 on Compute Unified Device Architecture (CUDA), IEEE Int. Conference on Advance Computing (IACC’ 10), Feb. 2010 (accepted) • N. Parakh, A. Mittal, and R. Niyogi, Optimization of MPEG 2 Encoder on Cell B. E. Processor. Proceedings of the IEEE International Conference on Advance Computing (IACC 09), 6 -7 March, 2009 pp 2945 --2949 • Amit Pande, Amit Verma, Ankush Mittal, Ashish Agrawal, Network Aware Content-based Resource Allocation for E-learning Systems, International Journal of Signal and Imaging Systems Engineering, Inderscience (accepted)
Publications related to the project International journal paper • A. Mittal, S. Gupta, S. Jain, and A. Jain, Content-based Adaptive Compression of Educational Videos using Phase Correlation Techniques, ACM Journal of Multimedia Systems (Springer), vol. 11, no. 3, pp. 249 -259, March 2006 • A. Mittal, C. Gupta and A. Gupta, Addressing the Bandwidth Efficiency, Control and Evaluation issues in Software Remote Laboratory, IEEE Transactions on Industrial Electronics, Special Issue on E-Learning and Remote Laboratories within Engineering Education, vol. 55 (6), pp. 2326 -2333, June 2008
M. J. NIGAM Professor Electronics & Computer Engineering Department Indian Institute of Technology Roorkee
M. J. Nigam • B. Tech. degree in Electronics and Communication Engineering from Regional Engineering College, Warangal (AP. ), India 1976. • M. E. degree in Electronics and Communication Engineering with specialization in Control & Guidance in 1978 • Ph. D. degree in Electronics and Computer Engineering in 1992 from University of Roorkee, India. • Presently Professor • Research interests: Digital Image Processing and Intelligent Control of Robot Manipulators. • Number of research articles in these areas have been published/ presented in various journals and Conferences.
Overall Research Activities • Research publications till date • Ph. D. guidance – completed - 50+ - 03 • Ph. D. guidance – in progress - 02 • M. Tech. guidance- completed - 45 • B. Tech. Projects guidance - completed - 14 • B. Tech. (SURA)Projects guidance - completed - 03 • B. Tech. Summer. Projects guidance - completed -04
Dr. R. Niyogi l B. Tech. (Jadavpur Univ) l Ph. D (IIT Kharagpur) l From 2006, assistant professor at IIT-R
Dr. Ankush Mittal l l l B. Tech. Computer Sc. & Engg. (IIT-Delhi) Ph. D (The National Univ. of Singapore) Worked in NUS as faculty for two years From Jan. 2007 – Dec. . ‘ 09, associate professor; now Prof. at COER Young scientist awards from NASI, ISCA, etc. Outstanding teacher award, 2008, IITR IBM Faculty award, US $10000
Research and Thesis Supervision (Dr. Ankush Mittal) Two Books 170 Research papers (60+ in Journals) IEEE Transactions: 5 Springer: 9 Elsevier: 8 Thesis Completed 6 Ph. D. 23 M. Tech. 12 B. Tech. Research Grants 3 Completed 1 in progress
THANK YOU