- Distributed Video Coding and Processing

Emerging applications such as wireless visual sensor networks (WVSN) and wireless video surveillance are requiring lightweight video encoding with high coding efficiency and error-resilience. The WVSNs have a range of applications including surveillance networks, health care systems, and monitoring systems. The surveillance visual sensors combined with signal processing and computer vision techniques can be used to locate criminals, terrorists, or accidents. The sensor networks can be integrated with other multimedia networks to provide health care services. Remote medical centers are able to perform advanced remote monitoring of their patients via multimedia sensors with remote assistance services. The wireless capsule endoscopy provides visual recordings inside the human body for diagnosis and monitoring. The WVSN is possibly a part of advance health informatics challenge, which is one of the grand challenges enabling a new system of distributed tools to collect medical data. In addition, the monitoring systems using visual sensors are used to monitor natural environment, health of human-made structures, e.g bridges, building, ships, etc., and disasters. Multimedia sensors can be used to monitor and control the industrial processes and systems in critical conditions.

The WVSNs are challenged by requiring advanced video coding and processing techniques in the energy-constrained wireless communications. One of the main design objectives of the WVSNs is a local (on-board) coding and processing technique with high compression efficiency, low-complexity, and error-resilience. The WVSNs also require real-time performance for the process extracting visual information from physical environments (by cameras) to transmit it to control centers (by users). Thus, most camera sensors have embedded processors that only support lightweight processing algorithms. Distributed Video Coding (DVC) is a new coding paradigm which exploits the source statistics at the decoder side offering such benefits for these applications. Although there have been some advanced improvement techniques, improving the DVC coding efficiency is still challenging.

The project addresses this challenge by proposing several iterative algorithms at different working levels, e.g. bitplane, band, and frame levels. The first proposed algorithm applies parallel iterative decoding using multiple LDPC decoders to utilize cross bitplane correlation. To improve Side Information (SI) generation and noise modeling and also learn from the previous decoded Wyner-Ziv (WZ) frames, side information and noise learning (SING) is proposed. The SING scheme introduces an optical flow technique to compensate the weaknesses of the block based SI generation and also utilizes clustering of DCT blocks to capture cross band correlation and increase local adaptivity in noise modeling. During decoding, the updated information is used to iteratively reestimate the motion and reconstruction in the proposed motion and reconstruction reestimation (MORE) scheme. The MORE scheme not only reestimates the motion vectors for improving SI and noise modeling but also compensates the residual motion based on the previously decoded WZ frames. Furthermore, the MORE codec enhances the reconstruction by proposing a generalized reconstruction algorithm to optimize reconstructing with multiple competitive SIs. Finally, an adaptive mode decision is investigated to take advantage of skip and intra mode in DVC by deciding the coding modes based on the quality of key frames and rate of WZ frames. Overall, the proposed algorithms significantly improve the coding efficiency of DVC contributing valuable solutions for the emerging applications.

References

1. Huynh Van Luong, Søren Forchhammer, Jürgen Slowack, Jan De Cock, and Rik Van de Walle, "Adaptive Mode Decision with Residual Motion Compensation for Distributed Video Coding," APSIPA Transactions on Signal and Information Processing, vol. 4, no. e1, pp. 1-11, Jan. 2015.
2. Huynh Van Luong, Lars Lau Rakêt, and Søren Forchhammer, "Re-estimation of Motion and Reconstruction for Distributed Video Coding," IEEE Transactions on Image Processing, vol. 23, no. 7, pp. 2804-2819, Jul. 2014.
3. Ph.D. thesis, “Distributed Video Coding: Iterative Improvements”, Technical University of Denmark, Denmark, 2013.
4. Huynh Van Luong, Lars Lau Rakêt, Xin Huang, and Søren Forchhammer, “Side Information and Noise Learning for Distributed Video Coding using Optical Flow and Clustering,” IEEE Transactions on Image Processing, vol. 21, no. 12, pp. 4782-4796, Dec. 2012.
5. Søren Forchhammer and Huynh Van Luong, “Distributed Source Coding of Video” in Workshop on Information Theoretic Methods in Science and Engineering 2015 (WITMSE 2015), Copenhagen, Denmark, June. 2015.
6. Huynh Van Luong, Søren Forchhammer, Jürgen Slowack, Jan De Cock, and Rik Van de Walle, "Adaptive Mode Decision with Residual Motion Compensation for Distributed Video Coding," in Picture Coding Symposium 2013 (PCS 2013), San Jose, California, Dec. 2013.
7. Huynh Van Luong and Søren Forchhammer, "Noise Residual Learning for Noise Modeling in Distributed Video Coding," in Picture Coding Symposium 2012 (PCS 2012), Krakow, Poland, May 2012.
8. S. Forchhammer, M. Salmistraro, K. Larsen, X. Huang, and H. V. Luong, “Rate-adaptive BCH coding for Slepian-Wolf coding of highly correlated sources,” in Data Compression Conference 2012 (DCC 2012), Snowbird, Utah, Apr. 2012.
9. L. L. Raket, J. Sogaard, M. Salmistraro, H. V. Luong, and S. Forchhammer, “Exploiting the Error-Correcting Capabilities of Low Density Parity Check Codes in Distributed Video Coding using Optical Flow,” in SPIE Optics+ Photonics, Optical Engineering+Applications, San Diego, California, Aug. 2012.
10. Huynh Van Luong, Xin Huang, and Søren Forchhammer, "Multiple LDPC Decoding using Bitplane Correlation for Trainsform Domain Wyner-Ziv Video Coding," in IEEE International Conference on Acoustics, Speech and Signal Processing 2011 (ICASSP2011), Prague, Czech Republic, May 2011.
11. Huynh Van Luong, Xin Huang, and Søren Forchhammer, "Parallel Iterative Decoding of Transform Domain Wyner-Ziv Video using Cross Bitplane Correlation," in IEEE International Conference on Image Processing 2011 (ICIP 2011), Brussels, Belgium, Sep. 2011.
12. Huynh Van Luong, Xin Huang, and Søren Forchhammer, "Adaptive Noise Model for Transform Domain Wyner-Ziv Video using Clustering of DCT Blocks," in IEEE International Workshop on Multimedia Signal Processing 2011 (MMSP 2011), Hangzhou, China, Oct. 2011.
13. Xin Huang, Lars Lau Rakêt, Huynh Van Luong, Mads Nielsen, Francois Lauze, and Søren Forchhammer, "Multi-hypothesis Transform Domain Wyner-Ziv Video Coding including Optical Flow," in IEEE International Workshop on Multimedia Signal Processing 2011 (MMSP 2011), Hangzhou, China, Oct. 2011. (Best 10% paper award)
14. Søren Forchhammer, Huynh Van Luong, and Xin Huang, “Multiple LDPC Decoding for Distributed Source Coding and Video Coding” in Workshop on Information Theoretic Methods in Science and Engineering (WITMSE 2011), Helsinki, Finland, Aug. 2011.

Copyright notice: Copyright of the above material belongs to the author. The material may not be copied or re-used without permission from the author.