Redundancy controllable scalable unbalanced multiple description bitstream generation for peer-to-peer video streaming

Abstract

In peer-to-peer video streaming, the robustness to peer and packet losses is regarded very important in order to enjoy a good quality of experience. Multiple description coding (MDC) schemes are known to provide high robustness to packet losses. In peer-to-peer streaming in addition to high robustness, efficient controllability of data rate and redundancy among the data streams from peers is also vital. In this paper, a novel framework for scalable multiple description coding scheme for peer-to-peer video streaming is presented by addressing these requirements. The proposed MDC solution is based on the multiple description scalar quantization (MDSQ) by addressing the joint decoding of unbalanced descriptions and addition of jointly decodable successive refinement layers to side descriptions. Firstly, the design conditions for MDSQ with constrained successive refinement to obtain scalable multiple description streams where the bit streams can be truncated at any point to obtain lower quality spatial temporal descriptions are firstly proposed. Then the design conditions for joint decoding of two or more side descriptions from different multiple description scalar quantizers, originating from various bin spread factors leading to controllable redundancy are proposed. These design conditions enable joint redundancy and data rate control for streams coming from different peers, thereby, enabling high robustness to packet losses as well as peer losses with the adaptability of redundancy levels and truncation of scalable streams. The proposed design constraints are used within the motion compensated temporal filtering (MCTF) framework to demonstrate its advancement in robust peer-to-peer video streaming. The results show significant improvements over conventional MDC based on simple MSDQ and over single description scalable video.