Multiple Description Coding for Path Diversity Video Streaming

In the current heterogeneous communication environments, the great variety of multimedia systems and applications combined with fast evolution of networking architectures and topologies, give rise to new research problems related to the various elements of the communication chain. This includes, the ever present problem in video communications, which results from the need for coping with transmission errors and losses. In this context, video streaming with path diversity appeared as a novel communication framework, involving di?erent technological ?elds and posing several research challenges. The research work carried out in this thesis is a contribution to robust video coding and adaptation techniques in the ?eld of Multiple Description Coding (MDC) for multipath video streaming. The thesis starts with a thorough study of MDC and its theoretical basis followed by a description of the most important practical implementation aspects currently available in literature. Additionally, a review of Multiple Description (MD) video coding is presented, where the relevant methods recently developed regarding this issue are explained, covering di?erent video coding approaches. In MDC, a video signal is typically encoded into several independent descriptions, i.e., compressed streams, where each one can be delivered over di?erent channels making use of path diversity communication environments. A high quality video representation is achieved when all descriptions are available at the decoder, whereas a lower, but still acceptable quality, is obtained when only one description is received. These interesting features of MDC are investigated in this thesis and compared to classic Single Description Coding (SDC). A research evaluation study of MDC for Advanced Video Coding (AVC) in regard to coding e?ciency, distortion and error resilience is presented. Starting with open-loop MD video coding architectures, this study evaluates the e?ects of distortion propagation that happens when individual decoding of each description is performed. A novel multi-loop architecture for AVC is then proposed, which prevents drift distortion accumulation, by generating a controlled amount of additional information. In particular, the proposed MD video coding architecture is based on Multiple Description Scalar Quantization (MDSQ including a new method for generating descriptions with di?erent rates, i.e. unbalanced descriptions. This research extends the current state-of-the-art methods using balanced MDSQ, developing new MDC capabilities in di?erent application scenarios without losing coding e?ciency neither robustness to transmission losses. This thesis also extends the current concept of multiple description coding (MDC) to the compressed domain, by proposing e?cient splitting of standard single description coded (SDC) video into a multi-stream representation. A novel multiple description video splitting (MDVS) scheme was developed to operate at network edges, for increased robustness in path diversity video streaming across heterogeneous communications chains. The proposed scheme is able to e?ectively control drift distortion in both intra and inter predictive coding, even when only one description reaches the decoder. This is achieved by generating a controlled amount of relevant side information to compensate for drift accumulation, whenever any description is lost in its path, achieving signi?cant quality improvement at reduced redundancy cost. Additionally, and taking into account the new achievements, a novel high-accurate rate control method is proposed for MDSQ video coding. Known models based on linear relation between the output rate and the percentage of zeros of the quantized transform coe?cients are extended to MDSQ video coding. In particular, this research demonstrates that the linear relation is maintained when MDSQ is used, making possible its use in linear rate control methods for MDC. Taking into account the balancing rate between descriptions, i.e. the percentage of the overall rate that is given to each description, this new method has the ability to choose the appropriate coding parameters to accurately achieve a prede?ned target MD bitrate. Overall, the new methods investigated in this thesis along with the good performance obtained from the experimental results, demonstrate the relevancy of its contribution to the ?eld of MDC and practical usefulness in new robust multimedia services and applications using path diversity channels.