Real-time interactive video streaming over lossy networks: high performance low delay error resilient algorithms

Xiao, Jimin Real-time interactive video streaming over lossy networks: high performance low delay error resilient algorithms. Doctor of Philosophy thesis, University of Liverpool.

	PDF thesis_jimin_xiao_final_Nov_2013.pdf - Submitted version Access to this file is embargoed until Unspecified. Available under License Creative Commons Attribution No Derivatives. Download (2MB)
	PDF XiaoJim_Nov2013_14293.pdf - Author Accepted Manuscript Available under License Creative Commons Attribution No Derivatives. Download (2MB)

Abstract

According to Cisco's latest forecast, two-thirds of the world's mobile data traffic and 62 percent of the consumer Internet traffic will be video data by the end of 2016. However, the wireless networks and Internet are unreliable, where the video traffic may undergo packet loss and delay. Thus robust video streaming over unreliable networks, i.e., Internet, wireless networks, is of great importance in facing this challenge. Specifically, for the real-time interactive video streaming applications, such as video conference and video telephony, the allowed end-to-end delay is limited, which makes the robust video streaming an even more difficult task. In this thesis, we are going to investigate robust video streaming for real-time interactive applications, where the tolerated end-to-end delay is limited. Intra macroblock refreshment is an effective tool to stop error propagations in the prediction loop of video decoder, whereas redundant coding is a commonly used method to prevent error from happening for video transmission over lossy networks. In this thesis two schemes that jointly use intra macroblock refreshment and redundant coding are proposed. In these schemes, in addition to intra coding, we proposed to add two redundant coding methods to enhance the transmission robustness of the coded bitstreams. The selection of error resilient coding tools, i.e., intra coding and/or redundant coding, and the parameters for redundant coding are determined using the end-to-end rate-distortion optimization. Another category of methods to provide error resilient capacity is using forward error correction (FEC) codes. FEC is widely studied to protect streamed video over unreliable networks, with Reed-Solomon (RS) erasure codes as its commonly used implementation method. As a block-based error correcting code, on the one hand, enlarging the block size can enhance the performance of the RS codes; on the other hand, large block size leads to long delay which is not tolerable for real-time video applications. In this thesis two sub-GOP (Group of Pictures, formed by I-frame and all the following P/B-frames) based FEC schemes are proposed to improve the performance of Reed-Solomon codes for real-time interactive video applications. The first one, named DSGF (Dynamic sub-GOP FEC Coding), is designed for the ideal case, where no transmission network delay is taken into consideration. The second one, named RVS-LE (Real-time Video Streaming scheme exploiting the Late- and Early-arrival packets), is more practical, where the video transmission network delay is considered, and the late- and early-arrival packets are fully exploited. Of the two approaches, the sub-GOP, which contains more than one video frame, is dynamically tuned and used as the RS coding block to get the optimal performance. For the proposed DSGF approach, although the overall error resilient performance is higher than the conventional FEC schemes, that protect the streamed video frame by frame, its video quality fluctuates within the Sub-GOP. To mitigate this problem, in this thesis, another real-time video streaming scheme using randomized expanding Reed-Solomon code is proposed. In this scheme, the Reed-Solomon coding block includes not only the video packets of the current frame, but also all the video packets of previous frames in the current group of pictures (GOP). At the decoding side, the parity-check equations of the current frameare jointly solved with all the parity-check equations of the previous frames. Since video packets of the following frames are not encompassed in the RS coding block, no delay will be caused for waiting for the video or parity packets of the following frames both at encoding and decoding sides. The main contribution of this thesis is investigating the trade-off between the video transmission delay caused by FEC encoding/decoding dependency, the FEC error-resilient performance, and the computational complexity. By leveraging the methods proposed in this thesis, proper error-resilient tools and system parameters could be selected based on the video sequence characteristics, the application requirements, and the available channel bandwidth and computational resources. For example, for the applications that can tolerate relatively long delay, sub-GOP based approach is a suitable solution. For the applications where the end-to-end delay is stringent and the computational resource is sufficient (e.g. CPU is fast), it could be a wise choice to use the randomized expanding Reed-Solomon code.

Item Type:	Thesis (Doctor of Philosophy)
Additional Information:	Date: 2013-11-18 (completed)
Uncontrolled Keywords:	video, streaming, error, packet loss, resilient, delay
Subjects:	?? TK ??
Divisions:	Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	17 Feb 2014 17:16
Last Modified:	16 Dec 2022 04:40
DOI:	10.17638/00014293
Supervisors:	Tillo, Tammam Smith, Jeremy Al-Nuaimy, Waleed
URI:	https://livrepository.liverpool.ac.uk/id/eprint/14293